Datasets:

librarian-bots

/

model_cards_with_metadata_test

like 1

PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: NousResearch/Llama-2-13b-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo NousResearch/Llama-2-13b-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-13b-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model NousResearch/Llama-2-13b-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: NousResearch/Llama-2-13b-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo NousResearch/Llama-2-13b-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/NousResearch-Llama-2-13b-hf-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-13b-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model NousResearch/Llama-2-13b-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

namrahrehman/deit-base-patch16-224-finetuned-lora-rank8

namrahrehman

Entry not found

PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/sail-Sailor-14B-Chat-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: sail/Sailor-14B-Chat metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo sail/Sailor-14B-Chat installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("sail/Sailor-14B-Chat") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model sail/Sailor-14B-Chat before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: PORTULAN/gervasio-7b-portuguese-ptbr-decoder metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo PORTULAN/gervasio-7b-portuguese-ptbr-decoder installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("PORTULAN/gervasio-7b-portuguese-ptbr-decoder") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model PORTULAN/gervasio-7b-portuguese-ptbr-decoder before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: zjunlp/knowlm-13b-zhixi metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo zjunlp/knowlm-13b-zhixi installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("zjunlp/knowlm-13b-zhixi") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model zjunlp/knowlm-13b-zhixi before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Doctor-Shotgun-TinyLlama-1.1B-32k-Instruct-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Doctor-Shotgun/TinyLlama-1.1B-32k-Instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

ShadNygren/FineTuneTest-DrugAdverseEffects-SIDER-Diego1-50epochs

ShadNygren

transformers

text-generation

--- base_model: meta-llama/Meta-Llama-3-8B-Instruct language: - en license: apache-2.0 tags: - text-generation-inference - transformers - unsloth - llama - trl --- # Uploaded model - **Developed by:** ShadNygren - **License:** apache-2.0 - **Finetuned from model :** meta-llama/Meta-Llama-3-8B-Instruct This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: zjunlp/knowlm-13b-zhixi metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo zjunlp/knowlm-13b-zhixi installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("zjunlp/knowlm-13b-zhixi") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model zjunlp/knowlm-13b-zhixi before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/sail-Sailor-14B-Chat-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: sail/Sailor-14B-Chat metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo sail/Sailor-14B-Chat installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("sail/Sailor-14B-Chat") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model sail/Sailor-14B-Chat before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: zjunlp/knowlm-13b-zhixi metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo zjunlp/knowlm-13b-zhixi installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/zjunlp-knowlm-13b-zhixi-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("zjunlp/knowlm-13b-zhixi") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model zjunlp/knowlm-13b-zhixi before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/sail-Sailor-14B-Chat-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: sail/Sailor-14B-Chat metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo sail/Sailor-14B-Chat installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/sail-Sailor-14B-Chat-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("sail/Sailor-14B-Chat") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model sail/Sailor-14B-Chat before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: PORTULAN/gervasio-7b-portuguese-ptbr-decoder metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo PORTULAN/gervasio-7b-portuguese-ptbr-decoder installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("PORTULAN/gervasio-7b-portuguese-ptbr-decoder") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model PORTULAN/gervasio-7b-portuguese-ptbr-decoder before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: PORTULAN/gervasio-7b-portuguese-ptbr-decoder metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo PORTULAN/gervasio-7b-portuguese-ptbr-decoder installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/PORTULAN-gervasio-7b-portuguese-ptbr-decoder-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("PORTULAN/gervasio-7b-portuguese-ptbr-decoder") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model PORTULAN/gervasio-7b-portuguese-ptbr-decoder before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

darshanmakwana/storage

darshanmakwana

--- license: mit ---

qgallouedec/hf_sb3_test

qgallouedec

Entry not found

PrunaAI/codellama-CodeLlama-13b-hf-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: codellama/CodeLlama-13b-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo codellama/CodeLlama-13b-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("codellama/CodeLlama-13b-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model codellama/CodeLlama-13b-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/codellama-CodeLlama-13b-hf-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: codellama/CodeLlama-13b-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo codellama/CodeLlama-13b-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("codellama/CodeLlama-13b-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model codellama/CodeLlama-13b-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/codellama-CodeLlama-13b-hf-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: codellama/CodeLlama-13b-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo codellama/CodeLlama-13b-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/codellama-CodeLlama-13b-hf-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("codellama/CodeLlama-13b-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model codellama/CodeLlama-13b-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

Ariffiq99/KUCI_CRAB_xlm_roberta_base_finetuned

Ariffiq99

transformers

multiple-choice

--- license: mit base_model: Ariffiq99/CRAB_xlm_roberta_base_finetuned tags: - generated_from_trainer metrics: - f1 model-index: - name: KUCI_CRAB_xlm_roberta_base_finetuned results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # KUCI_CRAB_xlm_roberta_base_finetuned This model is a fine-tuned version of [Ariffiq99/CRAB_xlm_roberta_base_finetuned](https://huggingface.co/Ariffiq99/CRAB_xlm_roberta_base_finetuned) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.7611 - F1: 0.7689 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 1e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | F1 | |:-------------:|:-----:|:-----:|:---------------:|:------:| | 0.7122 | 1.0 | 5196 | 0.6569 | 0.7361 | | 0.572 | 2.0 | 10392 | 0.6736 | 0.7582 | | 0.4757 | 3.0 | 15588 | 0.6246 | 0.7645 | | 0.3763 | 4.0 | 20784 | 0.6903 | 0.7666 | | 0.3013 | 5.0 | 25980 | 0.7611 | 0.7689 | ### Framework versions - Transformers 4.41.2 - Pytorch 2.3.0+cu121 - Datasets 2.20.0 - Tokenizers 0.19.1

av-legit/test

av-legit

--- tags: - safetensors ---

richardlastrucci/m2m100-tsn-nso

richardlastrucci

transformers

text2text-generation

Entry not found

hf-sb3-test/test_repo

hf-sb3-test

Entry not found

JvThunder/genre-recognizer-finetuned-gtzan_dset-finetuned-gtzan

JvThunder

transformers

audio-classification

--- license: apache-2.0 base_model: pedromatias97/genre-recognizer-finetuned-gtzan_dset tags: - generated_from_trainer datasets: - marsyas/gtzan model-index: - name: genre-recognizer-finetuned-gtzan_dset-finetuned-gtzan results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # genre-recognizer-finetuned-gtzan_dset-finetuned-gtzan This model is a fine-tuned version of [pedromatias97/genre-recognizer-finetuned-gtzan_dset](https://huggingface.co/pedromatias97/genre-recognizer-finetuned-gtzan_dset) on the GTZAN dataset. It achieves the following results on the evaluation set: - eval_loss: 0.2709 - eval_accuracy: 0.88 - eval_runtime: 67.912 - eval_samples_per_second: 1.472 - eval_steps_per_second: 0.191 - epoch: 2.0 - step: 226 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 3e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 15 - mixed_precision_training: Native AMP ### Framework versions - Transformers 4.41.2 - Pytorch 2.1.2 - Datasets 2.19.2 - Tokenizers 0.19.1

cmn/lora_Privilege_uncen

cmn

Entry not found

PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: AI-Sweden-Models/Llama-3-8B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo AI-Sweden-Models/Llama-3-8B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("AI-Sweden-Models/Llama-3-8B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model AI-Sweden-Models/Llama-3-8B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Nexusflow/NexusRaven-V2-13B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Nexusflow/NexusRaven-V2-13B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Nexusflow/NexusRaven-V2-13B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Nexusflow/NexusRaven-V2-13B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

akemimedrano100/shad

akemimedrano100

Entry not found

Ssunit/university_assistant

Ssunit

transformers

--- library_name: transformers tags: [] --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: AI-Sweden-Models/Llama-3-8B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo AI-Sweden-Models/Llama-3-8B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("AI-Sweden-Models/Llama-3-8B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model AI-Sweden-Models/Llama-3-8B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: AI-Sweden-Models/Llama-3-8B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo AI-Sweden-Models/Llama-3-8B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/AI-Sweden-Models-Llama-3-8B-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("AI-Sweden-Models/Llama-3-8B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model AI-Sweden-Models/Llama-3-8B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Tele-AI-telechat-7B-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Tele-AI/telechat-7B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Tele-AI/telechat-7B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Tele-AI-telechat-7B-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Tele-AI-telechat-7B-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Tele-AI/telechat-7B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Tele-AI/telechat-7B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Tele-AI-telechat-7B-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Tele-AI/telechat-7B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Tele-AI/telechat-7B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Tele-AI-telechat-7B-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Tele-AI-telechat-7B-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Tele-AI/telechat-7B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Tele-AI/telechat-7B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Tele-AI-telechat-7B-HQQ-4bit-smashed

PrunaAI

Entry not found

PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Nexusflow/NexusRaven-V2-13B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Nexusflow/NexusRaven-V2-13B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Nexusflow/NexusRaven-V2-13B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Nexusflow/NexusRaven-V2-13B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Nexusflow/NexusRaven-V2-13B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Nexusflow/NexusRaven-V2-13B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Nexusflow-NexusRaven-V2-13B-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Nexusflow/NexusRaven-V2-13B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Nexusflow/NexusRaven-V2-13B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

KYAGABA/wav2vec2-large-xls-r-300m-rw-KinyarwandaTTSDataset-10hr-v1

KYAGABA

transformers

automatic-speech-recognition

--- license: apache-2.0 base_model: facebook/wav2vec2-xls-r-300m tags: - generated_from_trainer metrics: - wer model-index: - name: wav2vec2-large-xls-r-300m-rw-KinyarwandaTTSDataset-10hr-v1 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # wav2vec2-large-xls-r-300m-rw-KinyarwandaTTSDataset-10hr-v1 This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.0032 - Wer: 0.0374 - Cer: 0.0061 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0005 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 32 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 20 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |:-------------:|:-------:|:----:|:---------------:|:------:|:------:| | 4.8359 | 1.3986 | 100 | 2.9224 | 1.0 | 1.0 | | 2.8486 | 2.7972 | 200 | 2.6198 | 1.0 | 0.9721 | | 1.159 | 4.1958 | 300 | 0.3714 | 0.6408 | 0.1162 | | 0.3918 | 5.5944 | 400 | 0.1782 | 0.3063 | 0.0493 | | 0.26 | 6.9930 | 500 | 0.1190 | 0.2531 | 0.0450 | | 0.2008 | 8.3916 | 600 | 0.0753 | 0.1472 | 0.0230 | | 0.1459 | 9.7902 | 700 | 0.0401 | 0.0921 | 0.0140 | | 0.1144 | 11.1888 | 800 | 0.0271 | 0.0718 | 0.0112 | | 0.0903 | 12.5874 | 900 | 0.0176 | 0.0474 | 0.0069 | | 0.0751 | 13.9860 | 1000 | 0.0119 | 0.0573 | 0.0084 | | 0.0619 | 15.3846 | 1100 | 0.0083 | 0.0340 | 0.0053 | | 0.0509 | 16.7832 | 1200 | 0.0051 | 0.0437 | 0.0068 | | 0.0435 | 18.1818 | 1300 | 0.0036 | 0.0374 | 0.0059 | | 0.0363 | 19.5804 | 1400 | 0.0032 | 0.0374 | 0.0060 | ### Framework versions - Transformers 4.41.2 - Pytorch 2.3.0+cu121 - Datasets 2.20.0 - Tokenizers 0.19.1

PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: UnicomLLM/Unichat-llama3-Chinese-8B-28K metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo UnicomLLM/Unichat-llama3-Chinese-8B-28K installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("UnicomLLM/Unichat-llama3-Chinese-8B-28K") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model UnicomLLM/Unichat-llama3-Chinese-8B-28K before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: UnicomLLM/Unichat-llama3-Chinese-8B-28K metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo UnicomLLM/Unichat-llama3-Chinese-8B-28K installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("UnicomLLM/Unichat-llama3-Chinese-8B-28K") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model UnicomLLM/Unichat-llama3-Chinese-8B-28K before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

Beijuka/wav2vec2-large-xls-r-300m-NCHLT_Speech_corpus_Xhosa-5hr

Beijuka

transformers

--- library_name: transformers tags: [] --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: UnicomLLM/Unichat-llama3-Chinese-8B-28K metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo UnicomLLM/Unichat-llama3-Chinese-8B-28K installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/UnicomLLM-Unichat-llama3-Chinese-8B-28K-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("UnicomLLM/Unichat-llama3-Chinese-8B-28K") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model UnicomLLM/Unichat-llama3-Chinese-8B-28K before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

hprashanth/4bit_model

hprashanth

transformers

text-generation

--- library_name: transformers tags: - unsloth - trl - sft --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

AhmedTaha012/my-trocr-model_5epochs_10k

AhmedTaha012

Entry not found

PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: TheDrummer/Moistral-11B-v2 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo TheDrummer/Moistral-11B-v2 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("TheDrummer/Moistral-11B-v2") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model TheDrummer/Moistral-11B-v2 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: TheDrummer/Moistral-11B-v2 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo TheDrummer/Moistral-11B-v2 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/TheDrummer-Moistral-11B-v2-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("TheDrummer/Moistral-11B-v2") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model TheDrummer/Moistral-11B-v2 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

TatevK/fine-tuned-models

TatevK

peft

--- license: apache-2.0 library_name: peft tags: - generated_from_trainer base_model: TheBloke/Mistral-7B-Instruct-v0.2-GPTQ model-index: - name: fine-tuned-models results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # fine-tuned-models This model is a fine-tuned version of [TheBloke/Mistral-7B-Instruct-v0.2-GPTQ](https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GPTQ) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.6877 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0002 - train_batch_size: 4 - eval_batch_size: 4 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 16 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 2 - num_epochs: 10 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | |:-------------:|:-----:|:----:|:---------------:| | 3.3763 | 1.0 | 9 | 2.6717 | | 2.0147 | 2.0 | 18 | 1.5139 | | 1.0198 | 3.0 | 27 | 0.8928 | | 0.6552 | 4.0 | 36 | 0.7814 | | 0.5873 | 5.0 | 45 | 0.7472 | | 0.547 | 6.0 | 54 | 0.7261 | | 0.5265 | 7.0 | 63 | 0.7083 | | 0.5026 | 8.0 | 72 | 0.6918 | | 0.4908 | 9.0 | 81 | 0.6896 | | 0.4732 | 10.0 | 90 | 0.6877 | ### Framework versions - PEFT 0.11.1 - Transformers 4.41.2 - Pytorch 2.1.0+cu121 - Datasets 2.19.2 - Tokenizers 0.19.1

PrunaAI/mychen76-mistral7b_ocr_to_json_v1-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: mychen76/mistral7b_ocr_to_json_v1 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo mychen76/mistral7b_ocr_to_json_v1 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/mychen76-mistral7b_ocr_to_json_v1-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/mychen76-mistral7b_ocr_to_json_v1-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("mychen76/mistral7b_ocr_to_json_v1") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model mychen76/mistral7b_ocr_to_json_v1 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

junannn/llama3-8b-cosmic-fusion-dynamics-merged_16bit-vllm

junannn

transformers

text-generation

--- base_model: unsloth/llama-3-8b-bnb-4bit language: - en license: apache-2.0 tags: - text-generation-inference - transformers - unsloth - llama - trl - sft --- # Uploaded model - **Developed by:** junannn - **License:** apache-2.0 - **Finetuned from model :** unsloth/llama-3-8b-bnb-4bit This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

Beijuka/wav2vec2_xls_r_300m_NCHLT_Speech_corpus_Xhosa_5hr_v1

Beijuka

transformers

automatic-speech-recognition

--- license: apache-2.0 base_model: facebook/wav2vec2-xls-r-300m tags: - generated_from_trainer metrics: - wer model-index: - name: wav2vec2_xls_r_300m_NCHLT_Speech_corpus_Xhosa_5hr_v1 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/asr-africa-research-team/ASR%20Africa/runs/2ga186eg) # wav2vec2_xls_r_300m_NCHLT_Speech_corpus_Xhosa_5hr_v1 This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.2681 - Wer: 0.6113 - Cer: 0.0668 ## Model description More information needed ## Intended uses & limitations More information needed ## Training and evaluation data More information needed ## Training procedure ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 0.0001 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 32 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: cosine - lr_scheduler_warmup_steps: 500 - num_epochs: 100 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |:-------------:|:-------:|:----:|:---------------:|:------:|:------:| | 16.7892 | 0.9969 | 163 | 3.9246 | 1.0 | 1.0 | | 3.2271 | 2.0 | 327 | 3.0166 | 1.0 | 1.0 | | 2.9133 | 2.9969 | 490 | 0.8980 | 0.9716 | 0.2786 | | 0.6747 | 4.0 | 654 | 0.4666 | 0.8256 | 0.1392 | | 0.4098 | 4.9969 | 817 | 0.4189 | 0.7744 | 0.1283 | | 0.3511 | 6.0 | 981 | 0.3803 | 0.7488 | 0.1166 | | 0.271 | 6.9969 | 1144 | 0.3817 | 0.8057 | 0.1183 | | 0.2278 | 8.0 | 1308 | 0.3687 | 0.8028 | 0.1197 | | 0.2067 | 8.9969 | 1471 | 0.3527 | 0.7175 | 0.0974 | | 0.1803 | 10.0 | 1635 | 0.3634 | 0.7118 | 0.0972 | | 0.1621 | 10.9969 | 1798 | 0.3548 | 0.6872 | 0.0914 | | 0.1451 | 12.0 | 1962 | 0.3965 | 0.8654 | 0.1237 | | 0.1315 | 12.9969 | 2125 | 0.3502 | 0.8853 | 0.1146 | | 0.1309 | 14.0 | 2289 | 0.3658 | 0.8303 | 0.1132 | | 0.1102 | 14.9969 | 2452 | 0.3688 | 0.7972 | 0.0972 | | 0.1047 | 16.0 | 2616 | 0.3427 | 0.8256 | 0.1012 | | 0.1023 | 16.9969 | 2779 | 0.3669 | 0.7583 | 0.0955 | | 0.0906 | 18.0 | 2943 | 0.3635 | 0.7033 | 0.0865 | | 0.0834 | 18.9969 | 3106 | 0.3818 | 0.7791 | 0.0956 | | 0.0763 | 20.0 | 3270 | 0.3751 | 0.8408 | 0.1061 | | 0.0789 | 20.9969 | 3433 | 0.3973 | 0.8095 | 0.0995 | ### Framework versions - Transformers 4.42.2 - Pytorch 2.2.0+cu121 - Datasets 2.20.0 - Tokenizers 0.19.1

richardlastrucci/m2m100-tsn-ssw

richardlastrucci

transformers

text2text-generation

Entry not found

PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: GreatCaptainNemo/ProLLaMA metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo GreatCaptainNemo/ProLLaMA installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("GreatCaptainNemo/ProLLaMA") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model GreatCaptainNemo/ProLLaMA before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: GreatCaptainNemo/ProLLaMA metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo GreatCaptainNemo/ProLLaMA installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("GreatCaptainNemo/ProLLaMA") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model GreatCaptainNemo/ProLLaMA before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: GreatCaptainNemo/ProLLaMA metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo GreatCaptainNemo/ProLLaMA installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/GreatCaptainNemo-ProLLaMA-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("GreatCaptainNemo/ProLLaMA") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model GreatCaptainNemo/ProLLaMA before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: minlik/chinese-alpaca-plus-7b-merged metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo minlik/chinese-alpaca-plus-7b-merged installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("minlik/chinese-alpaca-plus-7b-merged") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model minlik/chinese-alpaca-plus-7b-merged before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: minlik/chinese-alpaca-plus-7b-merged metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo minlik/chinese-alpaca-plus-7b-merged installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/minlik-chinese-alpaca-plus-7b-merged-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("minlik/chinese-alpaca-plus-7b-merged") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model minlik/chinese-alpaca-plus-7b-merged before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: dreamgen/opus-v1.2-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo dreamgen/opus-v1.2-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("dreamgen/opus-v1.2-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model dreamgen/opus-v1.2-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: dreamgen/opus-v1.2-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo dreamgen/opus-v1.2-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("dreamgen/opus-v1.2-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model dreamgen/opus-v1.2-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: dreamgen/opus-v1.2-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo dreamgen/opus-v1.2-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/dreamgen-opus-v1.2-llama-3-8b-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("dreamgen/opus-v1.2-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model dreamgen/opus-v1.2-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: meta-llama/Llama-2-13b-chat-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo meta-llama/Llama-2-13b-chat-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-13b-chat-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model meta-llama/Llama-2-13b-chat-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-1bit-smashed metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with [. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-1bit-smashed installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash REQUIREMENTS_INSTRUCTIONS ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer IMPORTS MODEL_LOAD tokenizer = AutoTokenizer.from_pretrained("PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-1bit-smashed") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-1bit-smashed before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: meta-llama/Llama-2-13b-chat-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo meta-llama/Llama-2-13b-chat-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/meta-llama-Llama-2-13b-chat-hf-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-13b-chat-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model meta-llama/Llama-2-13b-chat-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

itay-nakash/model_6e99ce7442_sweep_pleasant-sea-915

itay-nakash

Entry not found

PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: TinyLlama/TinyLlama_v1.1 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo TinyLlama/TinyLlama_v1.1 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama_v1.1") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model TinyLlama/TinyLlama_v1.1 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: TinyLlama/TinyLlama_v1.1 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo TinyLlama/TinyLlama_v1.1 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama_v1.1") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model TinyLlama/TinyLlama_v1.1 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: TinyLlama/TinyLlama_v1.1 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo TinyLlama/TinyLlama_v1.1 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/TinyLlama-TinyLlama_v1.1-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama_v1.1") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model TinyLlama/TinyLlama_v1.1 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-1bit-smashed

PrunaAI

Entry not found

PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: NousResearch/Llama-2-7b-chat-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo NousResearch/Llama-2-7b-chat-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-chat-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model NousResearch/Llama-2-7b-chat-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: NousResearch/Llama-2-7b-chat-hf metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo NousResearch/Llama-2-7b-chat-hf installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/NousResearch-Llama-2-7b-chat-hf-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-chat-hf") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model NousResearch/Llama-2-7b-chat-hf before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

itay-nakash/model_6e99ce7442_sweep_rich-sound-916

itay-nakash

Entry not found

PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: togethercomputer/Llama-2-7B-32K-Instruct metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo togethercomputer/Llama-2-7B-32K-Instruct installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("togethercomputer/Llama-2-7B-32K-Instruct") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model togethercomputer/Llama-2-7B-32K-Instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: togethercomputer/Llama-2-7B-32K-Instruct metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo togethercomputer/Llama-2-7B-32K-Instruct installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/togethercomputer-Llama-2-7B-32K-Instruct-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("togethercomputer/Llama-2-7B-32K-Instruct") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model togethercomputer/Llama-2-7B-32K-Instruct before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: IDEA-CCNL/Ziya2-13B-Base metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo IDEA-CCNL/Ziya2-13B-Base installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("IDEA-CCNL/Ziya2-13B-Base") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model IDEA-CCNL/Ziya2-13B-Base before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: IDEA-CCNL/Ziya2-13B-Base metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo IDEA-CCNL/Ziya2-13B-Base installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("IDEA-CCNL/Ziya2-13B-Base") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model IDEA-CCNL/Ziya2-13B-Base before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: IDEA-CCNL/Ziya2-13B-Base metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo IDEA-CCNL/Ziya2-13B-Base installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/IDEA-CCNL-Ziya2-13B-Base-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("IDEA-CCNL/Ziya2-13B-Base") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model IDEA-CCNL/Ziya2-13B-Base before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/antoinelouis-belgpt2-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: antoinelouis/belgpt2 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo antoinelouis/belgpt2 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("antoinelouis/belgpt2") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model antoinelouis/belgpt2 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/antoinelouis-belgpt2-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: antoinelouis/belgpt2 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo antoinelouis/belgpt2 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("antoinelouis/belgpt2") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model antoinelouis/belgpt2 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/antoinelouis-belgpt2-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: antoinelouis/belgpt2 metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo antoinelouis/belgpt2 installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/antoinelouis-belgpt2-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("antoinelouis/belgpt2") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model antoinelouis/belgpt2 before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

itay-nakash/model_6e99ce7442_sweep_crimson-snowball-917

itay-nakash

Entry not found

ShapeKapseln33/Nexalyn44

ShapeKapseln33

Nexalyn Bewertungen Norwegen: In einer Welt, in der Vitalität und Leistung oft gleichbedeutend mit Erfolg sind, ist es wichtig, körperlich in Topform zu bleiben. Bei Männern geht dies oft über die bloße Fitness hinaus auf Bereiche der Vitalität, Männlichkeit und des allgemeinen Wohlbefindens. Wenn wir jedoch älter werden, können verschiedene Faktoren unsere Fähigkeit beeinträchtigen, in diesen Bereichen optimale Leistung und Zufriedenheit aufrechtzuerhalten. **[Klicken Sie hier, um jetzt auf der offiziellen Website von Nexalyn zu kaufen](https://slim-gummies-deutschland.de/nexalyn-de)** ##Einführung in Nexalyn Möchten Sie Ihre sexuellen Erfahrungen auf die nächste Stufe heben? Wenn ja, sind Sie bei Nexalyn Testosterone Booster Formula genau richtig, dem revolutionären Nahrungsergänzungsmittel, das in der Welt der sexuellen Gesundheit für Aufsehen sorgt. Wenn Sie sich schon immer stärkere Erektionen, gesteigerten sinnlichen Appetit und explosive Orgasmen gewünscht haben, dann sind Sie bei uns genau richtig. Heute besprechen wir alles, was Sie über Nexalyn-Nahrungsergänzungsmittel zur männlichen Steigerung wissen müssen, von der wissenschaftlich fundierten Formel bis hin zu den wichtigsten Inhaltsstoffen, die Ihnen helfen, Ihre sexuelle Leistungsfähigkeit zu steigern. ##Die Wissenschaft hinter dem Nexalyn Testo-Booster und wie er funktioniert: Die Wissenschaft hinter den Nexalyn-Nahrungsergänzungsmitteln zur männlichen Leistungssteigerung unterscheidet sie von anderen Nahrungsergänzungsmitteln auf dem Markt. Diese leistungsstarke Formel kombiniert sorgfältig ausgewählte Inhaltsstoffe, die effektiv zur Verbesserung der sexuellen Gesundheit und Leistungsfähigkeit beitragen können. Es kann zur Verbesserung der Libido und der Erektionsfähigkeit beitragen. Es kann dazu beitragen, die Durchblutung des Genitalbereichs zu erhöhen, was zu stärkeren und länger anhaltenden Erektionen führt. Es kann auch den Testosteronspiegel erhöhen, was zu mehr Energie, Ausdauer und allgemeinem Sexualtrieb führt. Durch die Wiederherstellung des hormonellen Gleichgewichts kann es auch dazu beitragen, die Stimmung zu verbessern und Stress abzubauen. Es kann sich auch positiv auf die Sexualfunktion auswirken, indem es die Umwandlung von Testosteron in Dihydrotestosteron (DHT) verhindert, was zu Problemen wie Haarausfall und verminderter Libido führen kann. **[Klicken Sie hier, um jetzt auf der offiziellen Website von Nexalyn zu kaufen](https://slim-gummies-deutschland.de/nexalyn-de)** Es kann dazu beitragen, den freien Testosteronspiegel zu erhöhen, indem es an Sexualhormon-bindendes Globulin (SHBG) bindet, sodass mehr Testosteron frei im Körper zirkulieren kann. Nexalyn-Pillen in Australien und Neuseeland können darauf abzielen, Männern, die ihre sexuelle Leistungsfähigkeit auf natürliche Weise verbessern möchten, eine gute Unterstützung zu bieten. Die Wirkstoffe können auf verschiedene Aspekte der sexuellen Gesundheit abzielen, von der Verbesserung der Durchblutung und des Hormonhaushalts bis hin zur Steigerung des Energieniveaus, was zu einer verbesserten Ausdauer, einem gesteigerten sinnlichen Appetit und intensiveren Orgasmen führt. ##Hauptinhaltsstoffe und ihre Vorteile für die sexuelle Gesundheit: Wichtige Inhaltsstoffe spielen eine wichtige Rolle für die Wirksamkeit jedes Nahrungsergänzungsmittels, und die Nexalyn Testosterone Booster Formula bildet da keine Ausnahme. Werfen wir einen genaueren Blick auf einige der wichtigsten Inhaltsstoffe dieser leistungsstarken Formel und wie sie Ihrer sexuellen Gesundheit zugute kommen können. Horny Goat Weed, auch bekannt als Epimedium, wird in der traditionellen chinesischen Medizin seit Jahrhunderten zur Verbesserung der Libido und zur Behandlung von erektiler Dysfunktion eingesetzt. Es enthält Icariin, eine Verbindung, die dazu beitragen kann, die Durchblutung des Penis zu erhöhen, was zu stärkeren und länger anhaltenden Erektionen führt. Tongkat Ali-Wurzelextrakt ist ein weiterer wirksamer Inhaltsstoff mit aphrodisierenden Eigenschaften. Es kann wirken, indem es den Testosteronspiegel erhöht, was zu erhöhter Ausdauer, verbesserter Muskelmasse und verbesserter sexueller Leistungsfähigkeit führen kann. Sägepalme wird mit der Gesundheit der Prostata in Verbindung gebracht, spielt aber auch eine Rolle bei der Unterstützung des allgemeinen sexuellen Wohlbefindens. Durch die Hemmung der Umwandlung von Testosteron in Dihydrotestosteron (DHT) kann Sägepalme dazu beitragen, ein gesundes hormonelles Gleichgewicht aufrechtzuerhalten und eine optimale sexuelle Funktion zu unterstützen. Brennnesselwurzelextrakt ist reich an Vitamin A und C sowie Mineralstoffen wie Eisen und Magnesium. Diese Nährstoffe können die allgemeine reproduktive Gesundheit unterstützen und gleichzeitig das Energieniveau den ganzen Tag über steigern, ein wichtiger Faktor für die Aufrechterhaltung der Intimität mit Ihrem Partner. Durch die Kombination dieser kraftvollen natürlichen Inhaltsstoffe in einer Formel möchten Nexalyn-Kapseln Ihnen die notwendigen Werkzeuge an die Hand geben, um Ihr volles sinnliches Potenzial auszuschöpfen. Das Hinzufügen der männlichen Potenzformel von Nexalyn zu Ihrer täglichen Routine kann dazu beitragen, sowohl die körperliche Ausdauer als auch die geistige Konzentration in intimen Momenten zu verbessern, sodass Sie intensiveres Vergnügen als je zuvor erleben können! **[Klicken Sie hier, um jetzt auf der offiziellen Website von Nexalyn zu kaufen](https://slim-gummies-deutschland.de/nexalyn-de)**

richardlastrucci/m2m100-xho-eng

richardlastrucci

transformers

text2text-generation

Entry not found

PrunaAI/hywu-Camelidae-8x7B-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: hywu/Camelidae-8x7B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo hywu/Camelidae-8x7B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("hywu/Camelidae-8x7B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model hywu/Camelidae-8x7B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: HamidRezaAttar/gpt2-product-description-generator metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo HamidRezaAttar/gpt2-product-description-generator installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("HamidRezaAttar/gpt2-product-description-generator") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model HamidRezaAttar/gpt2-product-description-generator before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

immntrix/llamamodel

immntrix

transformers

--- base_model: unsloth/llama-3-8b-bnb-4bit language: - en license: apache-2.0 tags: - text-generation-inference - transformers - unsloth - llama - trl --- # Uploaded model - **Developed by:** immntrix - **License:** apache-2.0 - **Finetuned from model :** unsloth/llama-3-8b-bnb-4bit This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

PrunaAI/hywu-Camelidae-8x7B-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: hywu/Camelidae-8x7B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo hywu/Camelidae-8x7B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("hywu/Camelidae-8x7B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model hywu/Camelidae-8x7B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/hywu-Camelidae-8x7B-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: hywu/Camelidae-8x7B metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo hywu/Camelidae-8x7B installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/hywu-Camelidae-8x7B-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("hywu/Camelidae-8x7B") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model hywu/Camelidae-8x7B before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: HamidRezaAttar/gpt2-product-description-generator metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo HamidRezaAttar/gpt2-product-description-generator installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/HamidRezaAttar-gpt2-product-description-generator-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("HamidRezaAttar/gpt2-product-description-generator") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model HamidRezaAttar/gpt2-product-description-generator before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

Yami233333/llama3-8b-ruozhiba

Yami233333

transformers

text-generation

Entry not found

fatshady/kaktu_mikro

fatshady

transformers

--- base_model: unsloth/phi-3-mini-4k-instruct-bnb-4bit language: - en license: apache-2.0 tags: - text-generation-inference - transformers - unsloth - mistral - trl --- # Uploaded model - **Developed by:** fatshady - **License:** apache-2.0 - **Finetuned from model :** unsloth/phi-3-mini-4k-instruct-bnb-4bit This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library. [<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

wanasash/openai-whisper-tiny-LORA-colab

wanasash

transformers

--- library_name: transformers tags: [] --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

akemimedrano100/chase

akemimedrano100

Entry not found

rkjha5312/openhathi-mai-Finetune

rkjha5312

transformers

text-generation

--- library_name: transformers tags: [] --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

akemimedrano100/beerus

akemimedrano100

Entry not found

ogbi/mms1Bika

ogbi

transformers

--- library_name: transformers tags: [] --- # Model Card for Model ID <!-- Provide a quick summary of what the model is/does. --> ## Model Details ### Model Description <!-- Provide a longer summary of what this model is. --> This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated. - **Developed by:** [More Information Needed] - **Funded by [optional]:** [More Information Needed] - **Shared by [optional]:** [More Information Needed] - **Model type:** [More Information Needed] - **Language(s) (NLP):** [More Information Needed] - **License:** [More Information Needed] - **Finetuned from model [optional]:** [More Information Needed] ### Model Sources [optional] <!-- Provide the basic links for the model. --> - **Repository:** [More Information Needed] - **Paper [optional]:** [More Information Needed] - **Demo [optional]:** [More Information Needed] ## Uses <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. --> ### Direct Use <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. --> [More Information Needed] ### Downstream Use [optional] <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app --> [More Information Needed] ### Out-of-Scope Use <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. --> [More Information Needed] ## Bias, Risks, and Limitations <!-- This section is meant to convey both technical and sociotechnical limitations. --> [More Information Needed] ### Recommendations <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. --> Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations. ## How to Get Started with the Model Use the code below to get started with the model. [More Information Needed] ## Training Details ### Training Data <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. --> [More Information Needed] ### Training Procedure <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. --> #### Preprocessing [optional] [More Information Needed] #### Training Hyperparameters - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision --> #### Speeds, Sizes, Times [optional] <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. --> [More Information Needed] ## Evaluation <!-- This section describes the evaluation protocols and provides the results. --> ### Testing Data, Factors & Metrics #### Testing Data <!-- This should link to a Dataset Card if possible. --> [More Information Needed] #### Factors <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. --> [More Information Needed] #### Metrics <!-- These are the evaluation metrics being used, ideally with a description of why. --> [More Information Needed] ### Results [More Information Needed] #### Summary ## Model Examination [optional] <!-- Relevant interpretability work for the model goes here --> [More Information Needed] ## Environmental Impact <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly --> Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** [More Information Needed] - **Hours used:** [More Information Needed] - **Cloud Provider:** [More Information Needed] - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ## Technical Specifications [optional] ### Model Architecture and Objective [More Information Needed] ### Compute Infrastructure [More Information Needed] #### Hardware [More Information Needed] #### Software [More Information Needed] ## Citation [optional] <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> **BibTeX:** [More Information Needed] **APA:** [More Information Needed] ## Glossary [optional] <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. --> [More Information Needed] ## More Information [optional] [More Information Needed] ## Model Card Authors [optional] [More Information Needed] ## Model Card Contact [More Information Needed]

PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-1bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Vanessasml/cyber-risk-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Vanessasml/cyber-risk-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-1bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-1bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Vanessasml/cyber-risk-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Vanessasml/cyber-risk-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

Happpy0413/my_textclassification_model

Happpy0413

transformers

text-classification

Entry not found

PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-4bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Vanessasml/cyber-risk-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Vanessasml/cyber-risk-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-4bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-4bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Vanessasml/cyber-risk-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Vanessasml/cyber-risk-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/togethercomputer-LLaMA-2-7B-32K-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: togethercomputer/LLaMA-2-7B-32K metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo togethercomputer/LLaMA-2-7B-32K installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/togethercomputer-LLaMA-2-7B-32K-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/togethercomputer-LLaMA-2-7B-32K-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("togethercomputer/LLaMA-2-7B-32K") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model togethercomputer/LLaMA-2-7B-32K before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).

PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-2bit-smashed

PrunaAI

transformers

text-generation

--- thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg" base_model: Vanessasml/cyber-risk-llama-3-8b metrics: - memory_disk - memory_inference - inference_latency - inference_throughput - inference_CO2_emissions - inference_energy_consumption tags: - pruna-ai --- <!-- header start --> <!-- 200823 --> <div style="width: auto; margin-left: auto; margin-right: auto"> <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer"> <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;"> </a> </div> <!-- header end --> [](https://twitter.com/PrunaAI) [](https://github.com/PrunaAI) [](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following) [](https://discord.gg/CP4VSgck) # Simply make AI models cheaper, smaller, faster, and greener! - Give a thumbs up if you like this model! - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/) - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help. ## Results  **Frequently Asked Questions** - ***How does the compression work?*** The model is compressed with hqq. - ***How does the model quality change?*** The quality of the model output might vary compared to the base model. - ***How is the model efficiency evaluated?*** These results were obtained on HARDWARE_NAME with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you. - ***What is the model format?*** We use safetensors. - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data. - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model. - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai). - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads. - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases. ## Setup You can run the smashed model with these steps: 0. Check requirements from the original repo Vanessasml/cyber-risk-llama-3-8b installed. In particular, check python, cuda, and transformers versions. 1. Make sure that you have installed quantization related packages. ```bash pip install hqq ``` 2. Load & run the model. ```python from transformers import AutoModelForCausalLM, AutoTokenizer from hqq.engine.hf import HQQModelForCausalLM from hqq.models.hf.base import AutoHQQHFModel try: model = HQQModelForCausalLM.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-2bit-smashed", device_map='auto') except: model = AutoHQQHFModel.from_quantized("PrunaAI/Vanessasml-cyber-risk-llama-3-8b-HQQ-2bit-smashed") tokenizer = AutoTokenizer.from_pretrained("Vanessasml/cyber-risk-llama-3-8b") input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"] outputs = model.generate(input_ids, max_new_tokens=216) tokenizer.decode(outputs[0]) ``` ## Configurations The configuration info are in `smash_config.json`. ## Credits & License The license of the smashed model follows the license of the original model. Please check the license of the original model Vanessasml/cyber-risk-llama-3-8b before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi. ## Want to compress other models? - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact). - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).