diff --git "a/Qwen_Qwen2.5-Coder-32B-Instruct_tree.json" "b/Qwen_Qwen2.5-Coder-32B-Instruct_tree.json" new file mode 100644--- /dev/null +++ "b/Qwen_Qwen2.5-Coder-32B-Instruct_tree.json" @@ -0,0 +1,13220 @@ +{ + "base_model": "Qwen/Qwen2.5-Coder-32B-Instruct", + "tree": [ + { + "model_id": "Qwen/Qwen2.5-Coder-32B-Instruct", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/LICENSE\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n---\n\n\n# Qwen2.5-Coder-32B-Instruct\n\n \"Chat\"\n\n\n## Introduction\n\nQwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:\n\n- Significantly improvements in **code generation**, **code reasoning** and **code fixing**. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o.\n- A more comprehensive foundation for real-world applications such as **Code Agents**. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.\n- **Long-context Support** up to 128K tokens.\n\n**This repo contains the instruction-tuned 32B Qwen2.5-Coder model**, which has the following features:\n- Type: Causal Language Models\n- Training Stage: Pretraining & Post-training\n- Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias\n- Number of Parameters: 32.5B\n- Number of Paramaters (Non-Embedding): 31.0B\n- Number of Layers: 64\n- Number of Attention Heads (GQA): 40 for Q and 8 for KV\n- Context Length: Full 131,072 tokens\n - Please refer to [this section](#processing-long-texts) for detailed instructions on how to deploy Qwen2.5 for handling long texts.\n \nFor more details, please refer to our [blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/), [GitHub](https://github.com/QwenLM/Qwen2.5-Coder), [Documentation](https://qwen.readthedocs.io/en/latest/), [Arxiv](https://arxiv.org/abs/2409.12186).\n\n## Requirements\n\nThe code of Qwen2.5-Coder has been in the latest Hugging face `transformers` and we advise you to use the latest version of `transformers`.\n\nWith `transformers<4.37.0`, you will encounter the following error:\n```\nKeyError: 'qwen2'\n```\n\n## Quickstart\n\nHere provides a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate contents.\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"Qwen/Qwen2.5-Coder-32B-Instruct\"\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n torch_dtype=\"auto\",\n device_map=\"auto\"\n)\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\nprompt = \"write a quick sort algorithm.\"\nmessages = [\n {\"role\": \"system\", \"content\": \"You are Qwen, created by Alibaba Cloud. You are a helpful assistant.\"},\n {\"role\": \"user\", \"content\": prompt}\n]\ntext = tokenizer.apply_chat_template(\n messages,\n tokenize=False,\n add_generation_prompt=True\n)\nmodel_inputs = tokenizer([text], return_tensors=\"pt\").to(model.device)\n\ngenerated_ids = model.generate(\n **model_inputs,\n max_new_tokens=512\n)\ngenerated_ids = [\n output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\n\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n```\n\n### Processing Long Texts\n\nThe current `config.json` is set for context length up to 32,768 tokens.\nTo handle extensive inputs exceeding 32,768 tokens, we utilize [YaRN](https://arxiv.org/abs/2309.00071), a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts.\n\nFor supported frameworks, you could add the following to `config.json` to enable YaRN:\n```json\n{\n ...,\n \"rope_scaling\": {\n \"factor\": 4.0,\n \"original_max_position_embeddings\": 32768,\n \"type\": \"yarn\"\n }\n}\n```\n\nFor deployment, we recommend using vLLM. \nPlease refer to our [Documentation](https://qwen.readthedocs.io/en/latest/deployment/vllm.html) for usage if you are not familar with vLLM.\nPresently, vLLM only supports static YARN, which means the scaling factor remains constant regardless of input length, **potentially impacting performance on shorter texts**. \nWe advise adding the `rope_scaling` configuration only when processing long contexts is required.\n\n## Evaluation & Performance\n\nDetailed evaluation results are reported in this [\ud83d\udcd1 blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/).\n\nFor requirements on GPU memory and the respective throughput, see results [here](https://qwen.readthedocs.io/en/latest/benchmark/speed_benchmark.html).\n\n## Citation\n\nIf you find our work helpful, feel free to give us a cite.\n\n```\n@article{hui2024qwen2,\n title={Qwen2. 5-Coder Technical Report},\n author={Hui, Binyuan and Yang, Jian and Cui, Zeyu and Yang, Jiaxi and Liu, Dayiheng and Zhang, Lei and Liu, Tianyu and Zhang, Jiajun and Yu, Bowen and Dang, Kai and others},\n journal={arXiv preprint arXiv:2409.12186},\n year={2024}\n}\n@article{qwen2,\n title={Qwen2 Technical Report}, \n author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},\n journal={arXiv preprint arXiv:2407.10671},\n year={2024}\n}\n```\n", + "metadata": "\"N/A\"", + "depth": 0, + "children": [ + "Skywork/Skywork-SWE-32B", + "katanemo/Arch-Agent-32B", + "all-hands/openhands-lm-32b-v0.1", + "THUDM/SWE-Dev-32B", + "rombodawg/Rombos-Coder-V2.5-Qwen-32b", + "InferenceIllusionist/MilkDropLM-32b-v0.3", + "SWE-bench/SWE-agent-LM-32B", + "astro189/llava-1.5-7b-hf-ft-mix-vsft", + "mlx-community/Qwen2.5-Coder-32B-Instruct-bf16", + "unsloth/Qwen2.5-Coder-32B-Instruct", + "shanginn/Qwen2.5-Coder-32B-Instruct-mlx", + "gghfez/Qwen2.5-Coder-32B-Instruct-abliterated", + "huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated", + "Apel-sin/qwen2.5-coder-32b-instruct-exl2", + "dreakvasiliev/dixel", + "Conspirators/krx_qwen2.5_7b_it_nv2", + "Maks490/490", + "HuHu1226/L1G5000", + "thirdeyeai/Qwen2.5-Coder-32B-Instruct-Uncensored", + "luisomar/agentechat", + "cosmosbot/CosmosAI", + "ProtonZ/mariobrosregen", + "gozleryalansoylemez/ati242", + "caioguede93/masterAi", + "Dagles/trerwes", + "murater/winston", + "murita/carras", + "flxstr/1", + "Maazi89/PictoraGenAI", + "ranerane/d", + "Kamil95/Influencer", + "ozgursimsek73/education-ai", + "DENDORoff/DEWORLD", + "markzuck999/callboyjobbangalore", + "markzuck999/call-boy-job-chennai-salary-check", + "Mrgachet/chatbot-assistant", + "Sundeep-07/Travel-agency", + "benjaminrnash/jmc419", + "alperall/Nano_X_V1", + "OpenBuddy/openbuddy-qwen2.5coder-32b-v24.1q-200k", + "ulyana12340/1", + "Hiii2468/Recard", + "BluewhaleSec/Llama-3.1-BlueWhaleSec-2-8B", + "hynest/hyinya", + "Keshav3/gpt", + "banzajteam/1", + "zayoub/testin_hmi", + "PhoenixBlaze420/TheGanjaGuru", + "POKilondron/pok_AI", + "Patrihot77/PatFlutter", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX", + "nisten/tqwendo-36b", + "Cairo303/CoderUnited", + "WPJSVANZ/AdososCodAi", + "WPJSVANZ/CodersGOYBa", + "traineed/clarity-contracts", + "TIGER-Lab/AceCodeRM-32B", + "qingy2024/UIGEN-T1.1-Qwen-32B", + "Bojun-Feng/Qwen2.5-Coder-32B-Instruct-GGUF-llamafile", + "SuanChang/rain-SQLCoder", + "secmlr/SWE-BENCH-400-reasoning_qwen_code_32B_test_swe", + "secmlr/SWE-BENCH-400-reasoning-short-llm_qwen_code_32B_test_swe", + "open-r1/OlympicCoder-32B", + "karths/coder_python_32B", + "Tesslate/Tessa-T1-32B", + "Tesslate/UIGEN-T1.5-32B", + "Tesslate/UIGEN-T1.5-32B-4bit", + "all-hands/openhands-lm-32b-v0.1-ep3", + "lucyknada/all-hands_openhands-lm-32b-v0.1-exl2", + "secmlr/SWE-BENCH-500-set-4o-file-localization_qwen_code_32B_test_swe", + "secmlr/SWE-BENCH-433-set-claude-file-localization-with-reasoning_qwen_code_32B_test_swe", + "secmlr/SWE-BENCH-500-set-claude-3in1-localization-with-reasoning_3in1-32b-500", + "observerw/ChiseLLM-32B", + "all-hands/openhands-critic-32b-exp-20250417", + "secmlr/SWE-BENCH-433-set-claude-3in1-localization-with-reasoning_3in1-32b-433", + "vicharai/ViCoder-html-32B-preview", + "secmlr/SWE-BENCH-5k-first-2000-claude-search-replace-generation_qwen_code_32B_5k_first_2000_generation", + "secmlr/SWE-BENCH-2k-generation-enrich-500-localization-combine_32b-generation-localization-combine", + "lucyknada/SWE-bench_SWE-agent-LM-32B-exl2", + "r2e-edits/SWE-agent-LM-32B", + "LLMJapan/Qwen2.5-Coder-32B-Instruct_exl3", + "secmlr/SWE-BENCH-reproduce-verify-validation-plus-gym-diverse-source_qwen_code_32b_rvvpgds", + "RekklesAI/Qwen2.5-Coder-32B-Glaive-ToolCall", + "Satori-reasoning/Satori-SWE-SFT-32B", + "julien31/Soar-qwen-32b", + "MASWorks/MAS-GPT-32B" + ], + "children_count": 86, + "adapters": [ + "imsanjoykb/sqlCoder-Qwen2.5-8bit", + "levikvasko/bg", + "Haru22/ValReactions", + "elitecardpro/other", + "pm31194/Tomhardy", + "Landon93737373/Ushshshna", + "hacer201145/MyFirstAI", + "ritikjhamta007/Sad_Boy_Walking", + "Jak23/Bac2024", + "Olibot/chatfps", + "BluewhaleSec/bluewhalesec-offsec-ai", + "beenetics/bot_landingpage", + "shivendra11/signify", + "Saintgarson2/G2", + "ProPMGamingYT/BasicAI-1", + "likithasiddesh/liki", + "Ahmedtast/Gpt-jst", + "Reddington05/Neptune", + "shaikehsan/sqlCoder-Qwen2.5-8bit-F16-GGUF", + "shaikehsan/sqlcoder-qwen2.5-test" + ], + "adapters_count": 20, + "quantized": [ + "bartowski/Qwen2.5-Coder-32B-Instruct-GGUF", + "Qwen/Qwen2.5-Coder-32B-Instruct-GPTQ-Int4", + "Qwen/Qwen2.5-Coder-32B-Instruct-GGUF", + "mradermacher/Qwen2.5-Coder-32B-Instruct-i1-GGUF", + "lmstudio-community/Qwen2.5-Coder-32B-Instruct-GGUF", + "Qwen/Qwen2.5-Coder-32B-Instruct-GPTQ-Int8", + "Qwen/Qwen2.5-Coder-32B-Instruct-AWQ", + "second-state/Qwen2.5-Coder-32B-Instruct-GGUF", + "gaianet/Qwen2.5-Coder-32B-Instruct-GGUF", + "mlx-community/Qwen2.5-Coder-32B-Instruct-8bit", + "mlx-community/Qwen2.5-Coder-32B-Instruct-4bit", + "xingyaoww/Qwen2.5-Coder-32B-Instruct-AWQ-128k", + "DevQuasar/Qwen.Qwen2.5-Coder-32B-Instruct-GGUF", + "lmstudio-community/Qwen2.5-Coder-32B-Instruct-MLX-4bit", + "lmstudio-community/Qwen2.5-Coder-32B-Instruct-MLX-8bit", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-Q8_0-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-Q6_K-GGUF", + "unsloth/Qwen2.5-Coder-32B-Instruct-bnb-4bit", + "mradermacher/Qwen2.5-Coder-32B-Instruct-GGUF", + "bartowski/Qwen2.5-Coder-32B-Instruct-exl2", + "unsloth/Qwen2.5-Coder-32B-Instruct-GGUF", + "shanginn/Qwen2.5-Coder-32B-Instruct-mlx-q8", + "XelotX/Qwen2.5-Coder-32B-Instruct-Quants", + "XelotX/Qwen2.5-Coder-32B-Instruct-iQuants", + "CISCai/Qwen2.5-Coder-32B-Instruct-SOTA-GGUF", + "unsloth/Qwen2.5-Coder-32B-Instruct-128K-GGUF", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-4.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-8.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-7.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-6.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-5.5bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-4.5bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-5.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-3.0bpw-exl2", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-2.0bpw-exl2", + "dadmobile/Qwen2.5-Coder-32B-Instruct-Q4-mlx", + "DrNicefellow/Qwen2.5-Coder-32B-Instruct-1.0bpw-exl2", + "mlc-ai/Qwen2.5-Coder-32B-Instruct-q4f16_1-MLC", + "mlc-ai/Qwen2.5-Coder-32B-Instruct-q0f16-MLC", + "mlc-ai/Qwen2.5-Coder-32B-Instruct-q4f32_1-MLC", + "featherless-ai-quants/Qwen-Qwen2.5-Coder-32B-Instruct-GGUF", + "nm-testing/Qwen2.5-Coder-32B-Instruct-FP8-dynamic", + "AIFunOver/Qwen2.5-Coder-32B-Instruct-openvino-8bit", + "zhangfeng026/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "async0x42/Rombos-Coder-V2.5-Qwen-32b-exl2_4.0bpw", + "async0x42/Rombos-Coder-V2.5-Qwen-32b-exl2_5.0bpw", + "Volko76/Qwen2.5-Coder-32B-Instruct-GGUF", + "eligapris/Qwen2.5-Coder-32B-Instruct-Q2_K-GGUF", + "billatsectorflow/Qwen2.5-Coder-32B-Instruct-GPTQ-Int4-128K", + "Adapala/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "mlx-community/Qwen2.5-Coder-32B-Instruct-6bit", + "mlx-community/Qwen2.5-Coder-32B-Instruct-3bit", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q4_K_S-GGUF", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q5_K_S-GGUF", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q5_K_M-GGUF", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q6_K-GGUF", + "Triangle104/Qwen2.5-Coder-32B-Instruct-Q8_0-GGUF", + "litmudoc/Qwen2.5-Coder-32B-Instruct-Q4-mlx", + "tensorblock/Qwen2.5-Coder-32B-Instruct-GGUF", + "Maites/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "alvinrach/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "higorgonlw/Qwen2.5-Coder-32B-Instruct-Q8_0-GGUF", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_8.0bpw", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_7.0bpw", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_6.0bpw", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_5.0bpw", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_4.5bpw", + "Dracones/Qwen2.5-Coder-32B-Instruct_exl2_4.0bpw", + "numen-tech/Qwen2.5-Coder-32B-Instruct-GPTQ-Int4", + "eligapris/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "ggml-org/Qwen2.5-Coder-1.5B-32B-speculative-GGUF", + "cphan-intersystems/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "itlwas/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX-0cb1b", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX-104ce", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX-196c8", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX-8777b", + "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX-393a7", + "XelotX/Qwen2.5-Coder-32B-Instruct-GGUF", + "Felladrin/gguf-Q4_0-Qwen2.5-Coder-32B-Instruct", + "matatonic/Qwen2.5-Coder-32B-Instruct-abliterated-4.25bpw-exl2", + "parasail-ai/Qwen2.5-Coder-32B-Instruct-GPTQ-Int8-128k", + "moot20/Qwen2.5-Coder-32B-Instruct-MLX-4bits", + "moot20/Qwen2.5-Coder-32B-Instruct-MLX-6bits", + "moot20/Qwen2.5-Coder-32B-Instruct-MLX-8bits", + "EasierAI/Qwen-2.5-Coder-32B", + "Lucy-in-the-Sky/Qwen2.5-Coder-32B-Instruct-Q2_K-GGUF", + "Lucy-in-the-Sky/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "Lucy-in-the-Sky/Qwen2.5-Coder-32B-Instruct-Q6_K-GGUF", + "Lucy-in-the-Sky/Qwen2.5-Coder-32B-Instruct-Q8_0-GGUF", + "bobig/Qwen2.5-Coder-32B-Instruct-Q8", + "ig1/Qwen2.5-Coder-32B-Instruct-FP8-Dynamic", + "matatonic/OlympicCoder-32B-4.25bpw-exl2", + "matatonic/OlympicCoder-32B-6.5bpw-h8-exl2", + "LLMJapan/OlympicCoder-32B_exl2_8.0bpw", + "Fmuaddib/Qwen2.5-Coder-32B-Instruct-mlx-8Bit", + "drewbenson/Qwen2.5-Coder-32B-Instruct-4bit-MLX", + "Mungert/OlympicCoder-32B-GGUF", + "slackwaresupport/Qwen2.5-Coder-32B-Instruct-Q4_K_M-GGUF", + "Mungert/openhands-lm-32b-v0.1-GGUF", + "Mungert/SWE-agent-LM-32B-GGUF", + "Eric1227/Qwen2.5-Coder-32B-Instruct-MLX-8bit", + "duydq12/Qwen2.5-Coder-32B-Instruct-FP8-dynamic", + "textgeflecht/Qwen2.5-Coder-32B-Instruct-FP8-dynamic", + "katanemo/Arch-Agent-32B.gguf", + "Mungert/Skywork-SWE-32B-GGUF" + ], + "quantized_count": 107, + "merges": [], + "merges_count": 0, + "total_derivatives": 213, + "spaces": [], + "spaces_count": 0, + "parents": [], + "base_model": "Qwen/Qwen2.5-Coder-32B-Instruct", + "base_model_relation": "base" + }, + { + "model_id": "Skywork/Skywork-SWE-32B", + "gated": "unknown", + "card": "---\ntags:\n- swe-bench\nlicense: apache-2.0\nmetrics:\n- pass@1\nlibrary_name: transformers\npipeline_tag: text-generation\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---\n\n# Skywork-SWE\n\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/6665dd2b3a64c70529f7542c/8o-IE7N3GwSFCIH3ntc8E.png)\n\n\ud83d\udcd6 [Technical Report](https://huggingface.co/Skywork/Skywork-SWE-32B/resolve/main/assets/Report.pdf) | \ud83d\udcf0 [Blog](https://quixotic-sting-239.notion.site/eb17f379610040ceb54da5d5d24065bd)\n\n## Model Introduction\n***Skywork-SWE-32B*** is a code agent model developed by [Skywork AI](https://skywork.ai/home), specifically designed for software engineering (SWE) tasks. It demonstrates strong performance across several key metrics:\n- Skywork-SWE-32B attains 38.0% pass@1 accuracy on the [SWE-bench Verified](https://www.swebench.com) benchmark, outperforming previous open-source SoTA [Qwen2.5-Coder-32B-based](https://huggingface.co/Qwen/Qwen2.5-Coder-32B) LLMs built on the [OpenHands](https://github.com/All-Hands-AI/OpenHands) agent framework. \n- When incorporated with test-time scaling techniques, the performance further improves to 47.0% accuracy, surpassing the previous SoTA results for sub-32B parameter models.\n- We clearly demonstrate the data scaling law phenomenon for software engineering capabilities in LLMs, with no signs of saturation at 8209 collected training trajectories.\n\nWe also introduce an efficient and automated pipeline for SWE data collection, culminating in the creation of the Skywork-SWE dataset---a large-scale, high-quality dataset featuring comprehensive executable runtime environments. Detailed descriptions are available on our [technical report](https://huggingface.co/Skywork/Skywork-SWE-32B/resolve/main/assets/Report.pdf).\n### \ud83d\udd27 Model Details\n\n| Model Name | Backbone LLM | HuggingFace Link | Technical Report | Blog |\n|---|---------------|-----------|-|-|\n|Skywork-SWE-32B | [\ud83e\udd17 Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) | [\ud83e\udd17 Skywork-SWE-32B](https://huggingface.co/Skywork/Skywork-SWE-32B) | [Technical Report](https://huggingface.co/Skywork/Skywork-SWE-32B/resolve/main/assets/Report.pdf) | [Blog](https://quixotic-sting-239.notion.site/eb17f379610040ceb54da5d5d24065bd)|\n\n## Evaluation\n\n![image/png](https://huggingface.co/Skywork/Skywork-SWE-32B/resolve/main/assets/data_scaling_compressed.png)\n\nData Scaling Law for Pass@1 Accuracy on Qwen2.5-Coder-32B-Based LLMs Using the OpenHands v0.32.0 Code Agent Framework. Skywork-SWE-32B significantly outperforms previous Qwen2.5-Coder-32B-based LLMs, achieving the highest pass@1 accuracy without using verifiers or multiple rollouts.\n\n![image/png](https://huggingface.co/Skywork/Skywork-SWE-32B/resolve/main/assets/accuracy_compressed.png)\n\nWith the incorporation of test-time scaling techniques, Skywork-SWE-32B further improves to 47.0% accuracy, surpassing the previous SoTA results for sub-32B parameter models.\n\n## Performance Summary\n- Skywork-SWE-32B:\n```\nSubmission summary on SWE-bench verified split\n==================================================\nResolved 190 instances (38.0%)\n==================================================\nResolved by Repository\n- astropy/astropy: 4/22 (18.18%)\n- django/django: 99/231 (42.86%)\n- matplotlib/matplotlib: 9/34 (26.47%)\n- mwaskom/seaborn: 0/2 (0.0%)\n- pallets/flask: 1/1 (100.0%)\n- psf/requests: 4/8 (50.0%)\n- pydata/xarray: 7/22 (31.82%)\n- pylint-dev/pylint: 2/10 (20.0%)\n- pytest-dev/pytest: 9/19 (47.37%)\n- scikit-learn/scikit-learn: 17/32 (53.12%)\n- sphinx-doc/sphinx: 13/44 (29.55%)\n- sympy/sympy: 25/75 (33.33%)\n==================================================\nResolved by Time\n- 2013: 2/3 (66.67%)\n- 2014: 2/2 (100.0%)\n- 2015: 0/1 (0.0%)\n- 2016: 2/2 (100.0%)\n- 2017: 5/16 (31.25%)\n- 2018: 7/24 (29.17%)\n- 2019: 46/98 (46.94%)\n- 2020: 43/108 (39.81%)\n- 2021: 27/86 (31.4%)\n- 2022: 35/102 (34.31%)\n- 2023: 21/58 (36.21%)\n```\n- Skywork-SWE-32B + TTS (Bo8):\n```\nSubmission summary on SWE-bench verified split\n==================================================\nResolved 235 instances (47.0%)\n==================================================\nResolved by Repository\n- astropy/astropy: 8/22 (36.36%)\n- django/django: 115/231 (49.78%)\n- matplotlib/matplotlib: 15/34 (44.12%)\n- mwaskom/seaborn: 0/2 (0.0%)\n- pallets/flask: 1/1 (100.0%)\n- psf/requests: 3/8 (37.5%)\n- pydata/xarray: 14/22 (63.64%)\n- pylint-dev/pylint: 4/10 (40.0%)\n- pytest-dev/pytest: 10/19 (52.63%)\n- scikit-learn/scikit-learn: 22/32 (68.75%)\n- sphinx-doc/sphinx: 12/44 (27.27%)\n- sympy/sympy: 31/75 (41.33%)\n==================================================\nResolved by Time\n- 2013: 1/3 (33.33%)\n- 2014: 1/2 (50.0%)\n- 2015: 0/1 (0.0%)\n- 2016: 2/2 (100.0%)\n- 2017: 6/16 (37.5%)\n- 2018: 9/24 (37.5%)\n- 2019: 52/98 (53.06%)\n- 2020: 48/108 (44.44%)\n- 2021: 40/86 (46.51%)\n- 2022: 46/102 (45.1%)\n- 2023: 30/58 (51.72%)\n```\n\n## Usage\n\n### Install vLLM package\n```\n# Install vLLM version 0.9.0.1.\n# For example, if your CUDA version is 12.8, use the following command:\npip install vllm==0.9.0.1 --extra-index-url https://download.pytorch.org/whl/cu128\n```\n\n### Launch a server to deploy Skywork-SWE-32B\n\n```\nvllm serve ${MODEL_PATH} \u2014served-model-name ${SERVED_MODEL_NAME} --host 0.0.0.0 --port 8000 --gpu-memory-utilization 0.95 --tensor-parallel-size 8\n```\n\nSince our model has 32 billion parameters and supports a 32K context length, we recommend launching the model server with at least 2 GPUs equipped with sufficient VRAM to ensure efficient inference.\n### Set up OpenHands framework\n```\ngit clone https://github.com/All-Hands-AI/OpenHands.git\ncd OpenHands\ngit checkout tags/0.32.0\nmake build\n```\nThe official documentation of OpenHands: [SWE-Bench Evaluation with OpenHands SWE-Bench Docker Image](https://github.com/All-Hands-AI/OpenHands/tree/main/evaluation/benchmarks/swe_bench)\n\n### Create the corresponding config file:\n```\n[core]\nworkspace_base=\"./workspace\"\n\n[llm.my-oss-model]\nmodel = \"openai/${SERVED_MODEL_NAME}\"\nbase_url = \"http://0.0.0.0:8000/v1\"\napi_key=\"vllm\"\nmax_message_chars=32768\nmax_input_tokens=32768\nmax_output_tokens=8192\nlog_completions=true\ntemperature=0.0\n```\n\nIf you want to run the OpenHands agent with test-time scaling techniques (a Best-of-N method based on the critic model), please refer to the [blog](https://www.all-hands.dev/blog/sota-on-swe-bench-verified-with-inference-time-scaling-and-critic-model) for detailed instructions. You will need to switch to the [feature/llm-critic](https://github.com/All-Hands-AI/OpenHands/tree/feature/llm-critic) branch and deploy the [critic model](https://huggingface.co/all-hands/openhands-critic-32b-exp-20250417) accordingly. Additionally, you need to add the following parameters into the configuration file:\n```\nuse_critic=true\ncritic_model=\"critic_model\"\ncritic_base_url=\"**********\"\ncritic_api_key=\"************\"\ncritic_num_candidates=2\n```\n\n### Rollout on SWE-Bench Instances\n```\n./evaluation/benchmarks/swe_bench/scripts/run_infer.sh [model_config] [git-version] [agent] [eval_limit] [max_iter] [num_workers] [dataset] [dataset_split]\n\n# Example\n./evaluation/benchmarks/swe_bench/scripts/run_infer.sh llm.my-oss-model HEAD CodeActAgent 500 100 1 princeton-nlp/SWE-bench_Verified test\n```\n\n### Evaluate generated patches\n```\n./evaluation/benchmarks/swe_bench/scripts/eval_infer.sh \\\n./evaluation_outputs/outputs/princeton-nlp__SWE-bench_Lite-test/CodeActAgent/my-oss-model_maxiter_100_N_v0.32.0-no-hint-run_1/output.jsonl\n```\n\n## Acknowledgements\nWe would like to thank the contributors of the [OpenHands](https://github.com/All-Hands-AI/OpenHands) and [AllHands Critic](https://huggingface.co/all-hands/openhands-critic-32b-exp-20250417) repositories for their open research and valuable contributions.\n\n## Citation\nIf you use Skywork-SWE in your research, please consider citing our work using the following BibTeX entry:\n```\n@misc{zeng2025skyworksweunveilingdatascaling,\n title={Skywork-SWE: Unveiling Data Scaling Laws for Software Engineering in LLMs}, \n author={Liang Zeng and Yongcong Li and Yuzhen Xiao and Changshi Li and Chris Yuhao Liu and Rui Yan and Tianwen Wei and Jujie He and Xuchen Song and Yang Liu and Yahui Zhou},\n year={2025},\n eprint={2506.19290},\n archivePrefix={arXiv},\n primaryClass={cs.AI},\n url={https://arxiv.org/abs/2506.19290}, \n}\n```\n```\n@misc{skywork-swe,\n title={Skywork-SWE: Unveiling Data Scaling Laws for Software Engineering in LLMs},\n author={Liang Zeng, Yongcong Li, Yuzhen Xiao, Changshi Li, Chris Yuhao Liu, Rui Yan, Tianwen Wei, Jujie He, Xuchen Song, Yang Liu, and Yahui Zhou},\n howpublished={\\url{https://quixotic-sting-239.notion.site/eb17f379610040ceb54da5d5d24065bd}},\n note={Notion Blog},\n\tyear={2025},\n}\n```", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "mlx-community/Skywork-SWE-32B-bf16" + ], + "children_count": 1, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/Skywork-SWE-32B-GGUF", + "bartowski/Skywork_Skywork-SWE-32B-GGUF", + "lmstudio-community/Skywork-SWE-32B-GGUF", + "gabriellarson/Skywork-SWE-32B-GGUF", + "mlx-community/Skywork-SWE-32B-8bit", + "mlx-community/Skywork-SWE-32B-4bit", + "mradermacher/Skywork-SWE-32B-i1-GGUF", + "mlx-community/Skywork-SWE-32B-6bit", + "DevQuasar/Skywork.Skywork-SWE-32B-GGUF", + "GrimsenClory/Skywork-SWE-32B-Q5_K_S-GGUF" + ], + "quantized_count": 10, + "merges": [], + "merges_count": 0, + "total_derivatives": 11, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": null, + "base_model_relation": null + }, + { + "model_id": "katanemo/Arch-Agent-32B", + "gated": "unknown", + "card": "---\nlicense: other\nlicense_name: katanemo-research\nlicense_link: >-\n https://huggingface.co/katanemo/Arch-Agent-32B/blob/main/LICENSE\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nlanguage:\n- en\npipeline_tag: text-generation\nlibrary_name: transformers\n---\n\n# katanemo/Arch-Agent-32B\n\n## Overview\nArch-Agent is a collection of state-of-the-art (SOTA) LLMs specifically designed for advanced function calling and agent-based applications. Designed to power sophisticated multi-step and multi-turn workflows, Arch-Agent excels at handling complex, multi-step tasks that require intelligent tool selection, adaptive planning, and seamless integration with external APIs and services. Built with a focus on real-world agent deployments, Arch-Agent delivers leading performance in complex scenarios while maintaining reliability and precision across extended function call sequences. Key capabilities inlcude:\n\n- **Multi-Turn Function Calling**: Maintains contextual continuity across multiple dialogue turns, enabling natural, ongoing conversations with nested or evolving tool use.\n- **Multi-Step Function Calling**: Plans and executes a sequence of function calls to complete complex tasks. Adapts dynamically based on intermediate results and decomposes goals into sub-tasks.\n- **Agentic Capabilities**: Advanced decision-making and workflow management for complex agentic tasks with seamless tool coordination and error recovery.\n\nFor more details, including fine-tuning, inference, and deployment, please refer to our [Github](https://github.com/katanemo/Arch-Function).\n\n\n## Performance Benchmarks\nWe evaluate Katanemo Arch-Agent series on the [Berkeley Function-Calling Leaderboard (BFCL)](https://gorilla.cs.berkeley.edu/leaderboard.html#leaderboard). We compare with commonly-used models and the results (as of June 14th, 2025) are shown below.\n\n
\n \n
\n\n> [!NOTE]\n> For evaluation, we use YaRN scaling to deploy the models for Multi-Turn evaluation, and all Arch-Agent models are evaluated with a context length of 64K.\n\n## Requirements\nThe code of Arch-Agent-32B has been in the Hugging Face `transformers` library and we recommend to install latest version:\n```bash\npip install transformers>=4.51.0\n```\n\n\n## How to use\nWe use the following example to illustrate how to use our model to perform function calling tasks. Please note that, our model works best with our provided prompt format. It allows us to extract JSON output that is similar to the [OpenAI's function calling](https://platform.openai.com/docs/guides/function-calling).\n\n\n### Quickstart\n````python\nimport json\nfrom typing import Any, Dict, List\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"katanemo/Arch-Agent-32B\"\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name, device_map=\"auto\", torch_dtype=\"auto\", trust_remote_code=True\n)\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\nTASK_PROMPT = (\n \"You are a helpful assistant designed to assist with the user query by making one or more function calls if needed.\"\n \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\n\"\n \"You are provided with function signatures within XML tags:\\n\\n{tool_text}\"\n \"\\n\\n\\nFor each function call, return a json object with function name and arguments within \"\n \"\"\" XML tags:\\n\\n{{\"name\": , \"\"\"\n \"\"\"\"arguments\": }}\\n\"\"\"\n)\n\n# Define available tools\ntools = [\n {\n \"type\": \"function\",\n \"function\": {\n \"name\": \"get_weather\",\n \"description\": \"Get the current weather for a location\",\n \"parameters\": {\n \"type\": \"object\",\n \"properties\": {\n \"location\": {\n \"type\": \"str\",\n \"description\": \"The city and state, e.g. San Francisco, New York\",\n },\n \"unit\": {\n \"type\": \"str\",\n \"enum\": [\"celsius\", \"fahrenheit\"],\n \"description\": \"The unit of temperature to return\",\n },\n },\n \"required\": [\"location\"],\n },\n },\n }\n]\n\n# Helper function to create the system prompt for our model\ndef format_prompt(tools: List[Dict[str, Any]]):\n tool_text = \"\\n\".join(\n [json.dumps(tool[\"function\"], ensure_ascii=False) for tool in tools]\n )\n return TASK_PROMPT.format(tool_text=tool_text)\n\nsystem_prompt = format_prompt(tools)\n\nmessages = [\n {\"role\": \"system\", \"content\": system_prompt},\n {\"role\": \"user\", \"content\": \"What is the weather in Seattle?\"},\n]\n\nmodel_inputs = tokenizer.apply_chat_template(\n messages, add_generation_prompt=True, return_tensors=\"pt\", return_dict=True\n).to(model.device)\n\ngenerated_ids = model.generate(**model_inputs, max_new_tokens=32768)\ngenerated_ids = [\n output_ids[len(input_ids) :]\n for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\n\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\nprint(response)\n````\n\n# License\nThe Arch-Agent collection is distributed under the [Katanemo license](https://huggingface.co/katanemo/Arch-Agent-32B/blob/main/LICENSE).", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/Arch-Agent-32B-GGUF", + "mradermacher/Arch-Agent-32B-i1-GGUF", + "DevQuasar/katanemo.Arch-Agent-32B-GGUF" + ], + "quantized_count": 3, + "merges": [], + "merges_count": 0, + "total_derivatives": 3, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": null, + "base_model_relation": null + }, + { + "model_id": "all-hands/openhands-lm-32b-v0.1", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- SWE-Gym/SWE-Gym\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\ntags:\n- agent\n- coding\n---\n\n
\n \"Logo\"\n

OpenHands LM v0.1

\n
\n\n

\nBlog\n\u2022\nUse it in OpenHands\n

\n\n---\n\nAutonomous agents for software development are already contributing to a [wide range of software development tasks](/blog/8-use-cases-for-generalist-software-development-agents).\nBut up to this point, strong coding agents have relied on proprietary models, which means that even if you use an open-source agent like [OpenHands](https://github.com/All-Hands-AI/OpenHands), you are still reliant on API calls to an external service.\n\nToday, we are excited to introduce OpenHands LM, a new open coding model that:\n\n- Is open and [available on Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1), so you can download it and run it locally\n- Is a reasonable size, 32B, so it can be run locally on hardware such as a single 3090 GPU\n- Achieves strong performance on software engineering tasks, including 37.2% resolve rate on SWE-Bench Verified\n\nRead below for more details and our future plans!\n\n## What is OpenHands LM?\n\nOpenHands LM is built on the foundation of [Qwen Coder 2.5 Instruct 32B](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct), leveraging its powerful base capabilities for coding tasks. What sets OpenHands LM apart is our specialized fine-tuning process:\n\n- We used training data generated by OpenHands itself on a diverse set of open-source repositories\n- Specifically, we use an RL-based framework outlined in [SWE-Gym](https://arxiv.org/abs/2412.21139), where we set up a training environment, generate training data using an existing agent, and then fine-tune the model on examples that were resolved successfully\n- It features a 128K token context window, ideal for handling large codebases and long-horizon software engineering tasks\n\n\n## Performance: Punching Above Its Weight\n\nWe evaluated OpenHands LM using our latest [iterative evaluation protocol](https://github.com/All-Hands-AI/OpenHands/tree/main/evaluation/benchmarks/swe_bench#run-inference-rollout-on-swe-bench-instances-generate-patch-from-problem-statement) on the [SWE-Bench Verified benchmark](https://www.swebench.com/#verified).\n\nThe results are impressive:\n\n- **37.2% verified resolve rate** on SWE-Bench Verified\n- Performance comparable to models with **20x more parameters**, including Deepseek V3 0324 (38.8%) with 671B parameters\n\nHere's how OpenHands LM compares to other leading open-source models:\n\n![OpenHands LM Performance Comparison](https://www.all-hands.dev/assets/blog/20250331-openhands-lm-release/performance_scatter.png)\n\nAs the plot demonstrates, our 32B parameter model achieves efficiency that approaches much larger models. While the largest models (671B parameters) achieve slightly higher scores, our 32B parameter model performs remarkably well, opening up possibilities for local deployment that are not possible with larger models.\n\n## Getting Started: How to Use OpenHands LM Today\n\nYou can start using OpenHands LM immediately through these channels:\n\n1. **Download the model from Hugging Face**\n The model is available on [Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1) and can be downloaded directly from there.\n\n2. **Create an OpenAI-compatible endpoint with a model serving framework**\n For optimal performance, it is recommended to serve this model with a GPU using [SGLang](https://github.com/sgl-project/sglang) or [vLLM](https://github.com/vllm-project/vllm).\n\n3. **Point your OpenHands agent to the new model**\n Download [OpenHands](https://github.com/All-Hands-AI/OpenHands) and follow the instructions for [using an OpenAI-compatible endpoint](https://docs.all-hands.dev/modules/usage/llms/openai-llms#using-openai-compatible-endpoints).\n\n\n## The Road Ahead: Our Development Plans\n\nThis initial release marks just the beginning of our journey. We will continue enhancing OpenHands LM based on community feedback and ongoing research initiatives.\n\nIn particular, it should be noted that the model is still a research preview, and (1) may be best suited for tasks regarding solving github issues and perform less well on more varied software engineering tasks, (2) may sometimes generate repetitive steps, and (3) is somewhat sensitive to quantization, and may not function at full performance at lower quantization levels.\nOur next releases will focus on addressing these limitations.\n\nWe're also developing more compact versions of the model (including a 7B parameter variant) to support users with limited computational resources. These smaller models will preserve OpenHands LM's core strengths while dramatically reducing hardware requirements.\n\nWe encourage you to experiment with OpenHands LM, share your experiences, and participate in its evolution. Together, we can create better tools for tomorrow's software development landscape.\n\n\n## Try OpenHands Cloud\n\nWhile OpenHands LM is a powerful model you can run locally, we also offer a fully managed cloud solution that makes it even easier to leverage AI for your software development needs.\n\n[OpenHands Cloud](https://www.all-hands.dev/blog/introducing-the-openhands-cloud) provides:\n\n- Seamless GitHub integration with issue and PR support\n- Multiple interaction methods including text, voice, and mobile\n- Parallel agent capabilities for working on multiple tasks simultaneously\n- All the power of OpenHands without managing infrastructure\n\nOpenHands Cloud is built on the same technology as our open-source solution but adds convenient features for teams and individuals who want a ready-to-use platform. [Visit app.all-hands.dev](https://app.all-hands.dev) to get started today!\n\n\n## Join Our Community\n\nWe invite you to be part of the OpenHands LM journey:\n\n- Explore our [GitHub repository](https://github.com/All-Hands-AI/OpenHands)\n- Connect with us on [Slack](https://join.slack.com/t/openhands-ai/shared_invite/zt-2tom0er4l-JeNUGHt_AxpEfIBstbLPiw)\n- Follow our [documentation](https://docs.all-hands.dev) to get started\n\nBy contributing your experiences and feedback, you'll help shape the future of this open-source initiative. Together, we can create better tools for tomorrow's software development landscape.\n\nWe can't wait to see what you'll create with OpenHands LM!", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "alexgusevski/openhands-lm-32b-v0.1-mlx-fp16", + "JackCloudman/openhands-lm-32b-v0.1-jackterated", + "huihui-ai/openhands-lm-32b-v0.1-abliterated" + ], + "children_count": 3, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "Hypersniper/openhands-lm-32b-v0.1-Q4_K_S-GGUF", + "rdsm/openhands-lm-32b-v0.1-mlx-6bit", + "bartowski/all-hands_openhands-lm-32b-v0.1-GGUF", + "lmstudio-community/openhands-lm-32b-v0.1-GGUF", + "alexgusevski/openhands-lm-32b-v0.1-mlx-3Bit", + "alexgusevski/openhands-lm-32b-v0.1-mlx-4Bit", + "trevon/openhands-lm-32b-v0.1-Q8_0-GGUF", + "NewEden/openhands-lm-32b-v0.1-Q5_0-GGUF", + "alexgusevski/openhands-lm-32b-v0.1-mlx-8Bit", + "alexgusevski/openhands-lm-32b-v0.1-mlx-6Bit", + "rdsm/openhands-lm-32b-v0.1-mlx-mixed-3_6bit", + "mradermacher/openhands-lm-32b-v0.1-GGUF", + "mradermacher/openhands-lm-32b-v0.1-i1-GGUF", + "mlx-community/openhands-lm-32b-v0.1-4bit", + "DevQuasar/all-hands.openhands-lm-32b-v0.1-GGUF", + "stelterlab/openhands-lm-32b-v0.1-AWQ", + "avoroshilov/openhands-lm-32b-v0.1-GPTQ_4bit-128g", + "numen-tech/openhands-lm-32b-v0.1-GPTQ-Int4", + "tensorblock/all-hands_openhands-lm-32b-v0.1-GGUF" + ], + "quantized_count": 19, + "merges": [], + "merges_count": 0, + "total_derivatives": 22, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "all-hands/openhands-lm-32b-v0.1", + "base_model_relation": "base" + }, + { + "model_id": "THUDM/SWE-Dev-32B", + "gated": "False", + "card": "---\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nlibrary_name: transformers\nlicense: mit\npipeline_tag: text-generation\n---\n\n\ud83d\ude80 SWE-Dev, an open-source Agent for Software Engineering tasks! This repository contains the SWE-Dev-32B model as presented in the paper [SWE-Dev: Building Software Engineering Agents with Training and Inference Scaling](https://huggingface.co/papers/2506.07636).\n\n\ud83d\udca1 We develop a comprehensive pipeline for creating developer-oriented datasets from GitHub repositories, including issue tracking, code localization, test case generation, and evaluation.\n\n\ud83d\udd27 Based on open-source frameworks (OpenHands) and models, SWE-Dev-7B and 32B achieved solve rates of 23.4% and 36.6% on SWE-bench-Verified, respectively, even approaching the performance of GPT-4o. \n\n\ud83d\udcda We find that training data scaling and inference scaling can both effectively boost the performance of models on SWE-bench. Moreover, higher data quality further improves this trend when combined with reinforcement fine-tuning (RFT). For inference scaling specifically, the solve rate on SWE-Dev increased from 34.0% at 30 rounds to 36.6% at 75 rounds.\n\n\nSWE-Dev-32B is trained from [Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)**\n\n\nNotion Link: https://ubecwang.notion.site/1bc32cf963e080b2a01df2895f66021f?v=1bc32cf963e0810ca07e000c86c4c1e1\n\nGitHub Link: https://github.com/THUDM/SWE-Dev\n\nHugging Face Link: \n\n- SWE-Dev-7B (Qwen-2.5-Coder-7B-Instruct): https://huggingface.co/THUDM/SWE-Dev-7B/\n- SWE-Dev-9B (GLM-4-9B-Chat): https://huggingface.co/THUDM/SWE-Dev-9B/\n- SWE-Dev-32B (Qwen-2.5-Coder-32B-Instruct): https://huggingface.co/THUDM/SWE-Dev-32B/\n- SWE-Dev-train: https://huggingface.co/datasets/THUDM/SWE-Dev-train/", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "taylor-jones/SWE-Dev-32B-Q4_K_M-GGUF", + "SnuggieCase/SWE-Dev-32B-Q8_0-GGUF", + "mradermacher/SWE-Dev-32B-GGUF", + "mradermacher/SWE-Dev-32B-i1-GGUF", + "DevQuasar/THUDM.SWE-Dev-32B-GGUF" + ], + "quantized_count": 5, + "merges": [], + "merges_count": 0, + "total_derivatives": 5, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "THUDM/SWE-Dev", + "base_model_relation": "finetune" + }, + { + "model_id": "rombodawg/Rombos-Coder-V2.5-Qwen-32b", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B/blob/main/LICENSE\nlanguage:\n\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- qwen\n- qwen-coder\n- codeqwen\n---\n# Rombos-Coder-V2.5-Qwen-32b\n\n![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/642cc1c253e76b4c2286c58e/QErypCEKD5OZLxUcSmYaR.jpeg)\n\nRombos-Coder-V2.5-Qwen-32b is a continues finetuned version of Qwen2.5-Coder-32B-Instruct. I took it upon myself to merge the instruct model with the base model myself using the *Ties* merge method as demonstrated in my own \"Continuous Finetuning\" method (Linked bellow).\n\nhttps://docs.google.com/document/d/1OjbjU5AOz4Ftn9xHQrX3oFQGhQ6RDUuXQipnQ9gn6tU/edit?usp=sharing\n \nThis version of the model shows higher performance than the original instruct and base models. \n\nQuants: (Coming soon)\n\nGGUF: \n\n- https://huggingface.co/bartowski/Rombos-Coder-V2.5-Qwen-32b-GGUF\n\n- https://huggingface.co/mradermacher/Rombos-Coder-V2.5-Qwen-32b-i1-GGUF\n\nEXL2: \n\nBenchmarks: (Coming soon)", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "Apel-sin/rombos-coder-v2.5-qwen-32b-exl2" + ], + "children_count": 1, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "bartowski/Rombos-Coder-V2.5-Qwen-32b-GGUF", + "mradermacher/Rombos-Coder-V2.5-Qwen-32b-i1-GGUF", + "mradermacher/Rombos-Coder-V2.5-Qwen-32b-GGUF" + ], + "quantized_count": 3, + "merges": [], + "merges_count": 0, + "total_derivatives": 4, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "rombodawg/Rombos-Coder-V2.5-Qwen-32b", + "base_model_relation": "base" + }, + { + "model_id": "InferenceIllusionist/MilkDropLM-32b-v0.3", + "gated": "False", + "card": "---\nlicense: apache-2.0\npipeline_tag: text-generation\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- Visualizations\n- MilkDrop\n- unsloth\n- qwen\n---\n> [!WARNING]\n>\u26a0\ufe0f Epilepsy Warning: This model may generate visuals with flashing lights. Viewer discretion is advised for those with photosensitive epilepsy.\n\n
\n

MilkDropLM-32b-v0.3

\n
\n \n

\n \n

\n\n\n\nBuilding upon the success of our 7b model release, we're thrilled to introduce MilkDropLM-32b-v0.3, the latest iteration of our state-of-the-art language model designed specifically for creating visually stunning MilkDrop presets. This new 32b model is based on the Qwen2.5-Coder-32B-Instruct model and has been fine-tuned with the same tried and tested hyperparameters that made our [7b release](https://huggingface.co/InferenceIllusionist/MilkDropLM-7b-v0.3) function smoothly.\n\nThe future is here\u2026 so don\u2019t be surprised if the next amazing concert visual you see was authored by AI!\n\n## Notable Enhancements\n- **Unraveling the MilkDrop Genome**: MilkDropLM-32b-v0.3 has a more nuanced grasp of the intricate relationships between different elements within presets, allowing for far more accurate and creative generations.\n- **Seamless Enhancements**: This new model can now \u201cupgrade\u201d the presets that were generated with the 7b model, breathing new life into your favorite visuals. Try prompting with any existing MilkDrop preset script and asking for variations (at least 16k context size minimum for this feature).\n- **Fewer Loops, More Presets**: We've made significant strides in reducing the likelihood of the model getting stuck in loops, ensuring that you can focus on what matters most \u2013 creating stunning visuals.\n- **Smooth Operator**: Engage in more natural-sounding conversations with MilkDropLM-32b-v0.3, as it responds to your requests in a more human-like way than ever before.\n\n## What's Changed Under the Hood?\nWhile we've built upon the solid foundation of our hand curated and highly organized dataset containing 10,000+ best MilkDrop presets, we've made several key changes to the training process to address the increased requirements of the 32b model. In particular we've doubled down on the training time, pushing the model to its limits with 2 full epochs of training, coming out to approximately 48 hours on an A100 GPU. We also maintained the same generous maximum sequence length used in the 7b model during training to ensure that even the longest presets were accounted for in the training corpus.\n\n## Quantizations\n- [Static GGUF](https://huggingface.co/Quant-Cartel/MilkDropLM-32b-v0.3-GGUF)\n\n## Get Started with MilkDropLM-32b-v0.3\nReady to unlock the full potential of MilkDropLM-32b-v0.3? Unlike the 7b model, the 32b edition works out of the box with a variety of temperature and sampler configs, even default settings.\n\nFor starters we find that the default temperature of 1.0 works great. But just like before you can lower the temperature to increase the output quality at the cost of creativity. \n\nIn terms of context length, we would recommend the default maximum (32,768). We understand that this model\u2019s size is quite large, and so to ease VRAM requirements you can lower the context length to meet your compute. \n\n- **Min**: 8192 /// Minimum requirement. Enough to output most presets once\n- **Regular**: 16384 /// Allows for having up to 2 presets in \u2018memory\u2019\n- **Max**: 32768 /// Allows for 3-4 Presets in \u2018memory\u2019, recommended \n\nAlso make sure to max out your output length to prevent the model from stopping short and having to manually \u2018continue\u2019 the response.\n\nAfter you've generated a MilkDrop preset, copy-and-paste it into a text file, and then save it using the `.milk` file format. Move the `.milk` file into the presets folder of your Milkdrop app. We recommend using the [NestDrop Classic app](https://nestimmersion.ca/nestdrop.php), which is freely available. (Make sure to close and reopen NestDrop to see your newly added presets.)\n\n\n## Text Prompt Template\nIn terms of the approach to text prompting with this model, the classic \u201cGive me a Glowsticks milkdrop preset\u201d or \u201cMake a milkdrop preset with [x], [y], [z]\u201d still works very well. But feel free to experiment with brand new ways to ask the 32b model for presets to take advantage of its new conversational capabilities and the results may surprise you! Below is the full list of preset categories/subcategories that this model was trained on.\n\n```\nDancer /// Aurora, Blobby Mirror, Blobby, Comet Mirror, Comet, Glowsticks Fast, Glowsticks Mirror, Glowsticks, Hatches Mirror, Hatches, Infect Mirror, Infect, Jello Mirror, Lasers, Murky Mirror, Murky, Nexus Mirror, Orbit Mirror, Orbit, Petals, Serpent, Shapes, Spinner Mirror, Spinner, Streamers, Swarm, Tree Branch, Wake Mirror, Wake, Whirl Mirror, Whirl, Wire\nDrawing /// Dunes Mirror, Dunes, Explosions Mirror, Explosions, Feedback, Fractal Mirror, Fractal, Glimmer Mirror, Growth Mirror, Growth, Lasers, Liquid Mirror, Liquid, Maze, Rorschach Mirror, Rorschach, Trails Mirror, Trails, Viscera, Whirlpools Mirror, Whirlpools\nFractal /// Blobby, Core Mirror, Core Tunnel, Core, Grid, Horizon Mirror, Lattice Mirror, Lattice, Loops, Mandelbox, Nested Circle, Nested Dancer, Nested Ellipse, Nested Hexagon, Nested Pyramid, Nested Spiral Multiple, Nested Spiral, Nested Square, Nested Triangle, Pointy, Radial, Shine, Sierpinski, Trees, Wave Interference, Wings, Womb\nGeometric /// Cathedral, Circles Nested, City, Cube Array, Cube Fly, Cube, Dots, Gears, Honeycomb, Landscape, Monster, Pyramids, Snowflakes, Sphere Array, Sphere Particles, Sphere Wild, Spiral Bounce, Squares Glass, Squares, Stripes Circle, Stripes Dance, Stripes Liquid, Stripes, Symbols, Torus Interior, Triangles Solo, Tunnel Fans, Tunnel Morph, Tunnel Spheres, Wire Circles, Wire Cube Trace, Wire Flower, Wire Grid, Wire Morph, Wire Orbits, Wire Parallel, Wire Sphere, Wire Spiral Flower, Wire Spiral, Wire Torus, Wire Trace Mirror, Wire Trace\nHypnotic /// Illusion Radiate, Illusion, Polar Closeup, Polar Mirror, Polar Rolling, Polar Static, Polar Warp, Radial Warp\nParticles /// Blobby, Crystal, Grid, Orbit, Points Fast, Points Trails, Points, Swarm\nReaction /// Automata, Cloudy, Contagion, Crystalize, Dunes, Feedback, Growth, Liquid Blobby, Liquid Closeup, Liquid Gradient, Liquid Ripples, Liquid Simmering, Liquid Windy, Luma Mirror, Maze, Mountains, Rorschach, Viscera, Whirlpools, Windy\nSparkle /// Explosions, Glimmer Mirror, Glimmer Tunnel, Glimmer, Jewel, Mass Circles, Mass Squares, Mass Stars, Mass Triangles, Squares\nSupernova /// Burst, Gas, Lasers, Orbits, Radiate, Shimmer, Stars\nWaveform /// Spectrum, Wire Circular, Wire Flat Double, Wire Flat, Wire Flower, Wire Mirror, Wire Rising, Wire Spirograph, Wire Tangle, Wire Tunnel\n```\n\n## Alpha Release Notice\nAs with any new model release, we want to emphasize that MilkDropLM-32b-v0.3 is still in the alpha stage of development and we\u2019re actively running experiments to see what it\u2019s capable of. While we're confident that this model represents a significant leap forward, we're not done yet. We encourage you to approach MilkDropLM-32b-v0.3 with a spirit of experimentation and discovery, and we can't wait to see what amazing visuals you'll create!\n\n## Acknowledgements\nThis project is the result of a collaboration between [ISOSCELES](https://www.instagram.com/isosceles.vj) and InferenceIllusionist. This was a unique meeting of minds since ISOSCELES brought his MilkDrop preset knowledge and experience in helping develop NestDrop for the VJ community, and InferenceIllusionist brought his vital experience in fine-tuning and quantizing LLMs. We stand on the shoulders of the many Milkdrop authors which have freely released their original presets for everyone to enjoy. Much respect!\n\nWe would like to express our deepest gratitude towards our growing community of alpha testers and feedback providers for their invaluable insights and support throughout this development process. We truly appreciate your pioneer spirit and courage in embracing this new family of Large Language Models.\n\nShoutout to [Unsloth](https://unsloth.ai) as well for providing the tools used for this fine-tune.\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "Quant-Cartel/MilkDropLM-32b-v0.3-GGUF", + "mradermacher/MilkDropLM-32b-v0.3-GGUF", + "mradermacher/MilkDropLM-32b-v0.3-i1-GGUF" + ], + "quantized_count": 3, + "merges": [], + "merges_count": 0, + "total_derivatives": 3, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "InferenceIllusionist/MilkDropLM-32b-v0.3", + "base_model_relation": "base" + }, + { + "model_id": "SWE-bench/SWE-agent-LM-32B", + "gated": "False", + "card": "---\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ndatasets:\n- SWE-bench/SWE-smith\nlanguage:\n- en\nlibrary_name: transformers\nlicense: apache-2.0\npipeline_tag: text-generation\ntags:\n- agent\n- software engineering\n---\n\n
\n \"Logo\"\n

SWE-agent LM

\n
\n\n

\nCode\n\u2022\nPaper\n\u2022\nSite\n

\n\nSWE-agent-LM-32B is a Language Model for Software Engineering trained using the [SWE-smith](https://github.com/SWE-bench/SWE-smith) toolkit.\nWe introduce this model as part of our work: [SWE-smith: Scaling Data for Software Engineering Agents](https://swesmith.com).\n\nSWE-agent-LM-32B is 100% open source.\nTraining this model was simple - we fine-tuned Qwen 2.5 Coder Instruct on 5k trajectories generated by SWE-agent + Claude 3.7 Sonnet.\nThe dataset can be found [here](https://huggingface.co/datasets/SWE-bench/SWE-smith-trajs-250429).\n\nSWE-agent-LM-32B is compatible with [SWE-agent](https://github.com/SWE-agent/SWE-agent).\nRunning this model locally only takes a few steps!\nCheck [here]() for more instructions on how to do so.\n\nIf you found this work exciting and want to push SWE-agents further, please feel free to connect with us (the [SWE-bench team](https://swe-bench.github.io/)) more!", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "Kebob/SWE-agent-LM-32B-GGUF", + "DevQuasar/SWE-bench.SWE-agent-LM-32B-GGUF", + "mlx-community/SWE-agent-LM-32B-4bit", + "featherless-ai-quants/SWE-bench-SWE-agent-LM-32B-GGUF" + ], + "quantized_count": 4, + "merges": [], + "merges_count": 0, + "total_derivatives": 4, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "SWE-bench/SWE-agent-LM", + "base_model_relation": "finetune" + }, + { + "model_id": "astro189/llava-1.5-7b-hf-ft-mix-vsft", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nlibrary_name: transformers\npipeline_tag: visual-question-answering\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "astro189/llava-1.5-7b-hf-ft-mix-vsft", + "base_model_relation": "base" + }, + { + "model_id": "mlx-community/Qwen2.5-Coder-32B-Instruct-bf16", + "gated": "False", + "card": "---\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\nlanguage:\n- en\nlibrary_name: transformers\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/LICENSE\npipeline_tag: text-generation\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n- mlx\n---\n\n# mlx-community/Qwen2.5-Coder-32B-Instruct-bf16\n\nThe Model [mlx-community/Qwen2.5-Coder-32B-Instruct-bf16](https://huggingface.co/mlx-community/Qwen2.5-Coder-32B-Instruct-bf16) was converted to MLX format from [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) using mlx-lm version **0.19.3**.\n\n## Use with mlx\n\n```bash\npip install mlx-lm\n```\n\n```python\nfrom mlx_lm import load, generate\n\nmodel, tokenizer = load(\"mlx-community/Qwen2.5-Coder-32B-Instruct-bf16\")\n\nprompt=\"hello\"\n\nif hasattr(tokenizer, \"apply_chat_template\") and tokenizer.chat_template is not None:\n messages = [{\"role\": \"user\", \"content\": prompt}]\n prompt = tokenizer.apply_chat_template(\n messages, tokenize=False, add_generation_prompt=True\n )\n\nresponse = generate(model, tokenizer, prompt=prompt, verbose=True)\n```\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "tensorblock/Qwen2.5-Coder-32B-Instruct-bf16-GGUF" + ], + "quantized_count": 1, + "merges": [], + "merges_count": 0, + "total_derivatives": 1, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "mlx-community/Qwen2.5-Coder-32B-Instruct-bf16", + "base_model_relation": "base" + }, + { + "model_id": "unsloth/Qwen2.5-Coder-32B-Instruct", + "gated": "False", + "card": "---\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\nlanguage:\n- en\nlibrary_name: transformers\nlicense: apache-2.0\ntags:\n- unsloth\n- transformers\n- code\n- qwen\n- qwen-coder\n- codeqwen\n---\n\n# Finetune Llama 3.1, Gemma 2, Mistral 2-5x faster with 70% less memory via Unsloth!\n\nWe have a Qwen 2.5 (all model sizes) [free Google Colab Tesla T4 notebook](https://colab.research.google.com/drive/1Kose-ucXO1IBaZq5BvbwWieuubP7hxvQ?usp=sharing).\nAlso a [Qwen 2.5 conversational style notebook](https://colab.research.google.com/drive/1qN1CEalC70EO1wGKhNxs1go1W9So61R5?usp=sharing).\n\n[](https://discord.gg/unsloth)\n[](https://github.com/unslothai/unsloth)\n\n## \u2728 Finetune for Free\n\nAll notebooks are **beginner friendly**! Add your dataset, click \"Run All\", and you'll get a 2x faster finetuned model which can be exported to GGUF, vLLM or uploaded to Hugging Face.\n\n| Unsloth supports | Free Notebooks | Performance | Memory use |\n|-----------------|--------------------------------------------------------------------------------------------------------------------------|-------------|----------|\n| **Llama-3.1 8b** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/1Ys44kVvmeZtnICzWz0xgpRnrIOjZAuxp?usp=sharing) | 2.4x faster | 58% less |\n| **Phi-3.5 (mini)** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/1lN6hPQveB_mHSnTOYifygFcrO8C1bxq4?usp=sharing) | 2x faster | 50% less |\n| **Gemma-2 9b** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/1vIrqH5uYDQwsJ4-OO3DErvuv4pBgVwk4?usp=sharing) | 2.4x faster | 58% less |\n| **Mistral 7b** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/1Dyauq4kTZoLewQ1cApceUQVNcnnNTzg_?usp=sharing) | 2.2x faster | 62% less |\n| **TinyLlama** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/1AZghoNBQaMDgWJpi4RbffGM1h6raLUj9?usp=sharing) | 3.9x faster | 74% less |\n| **DPO - Zephyr** | [\u25b6\ufe0f Start on Colab](https://colab.research.google.com/drive/15vttTpzzVXv_tJwEk-hIcQ0S9FcEWvwP?usp=sharing) | 1.9x faster | 19% less |\n\n- This [conversational notebook](https://colab.research.google.com/drive/1Aau3lgPzeZKQ-98h69CCu1UJcvIBLmy2?usp=sharing) is useful for ShareGPT ChatML / Vicuna templates.\n- This [text completion notebook](https://colab.research.google.com/drive/1ef-tab5bhkvWmBOObepl1WgJvfvSzn5Q?usp=sharing) is for raw text. This [DPO notebook](https://colab.research.google.com/drive/15vttTpzzVXv_tJwEk-hIcQ0S9FcEWvwP?usp=sharing) replicates Zephyr.\n- \\* Kaggle has 2x T4s, but we use 1. Due to overhead, 1x T4 is 5x faster.\n\n# unsloth/Qwen2.5-Coder-32B-Instruct\n\n## Introduction\n\nQwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:\n\n- Significantly improvements in **code generation**, **code reasoning** and **code fixing**. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o.\n- A more comprehensive foundation for real-world applications such as **Code Agents**. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.\n\n**This repo contains the 0.5B Qwen2.5-Coder model**, which has the following features:\n- Type: Causal Language Models\n- Training Stage: Pretraining\n- Architecture: transformers with RoPE, SwiGLU, RMSNorm, Attention QKV bias and tied word embeddings\n- Number of Parameters: 0.49B\n- Number of Paramaters (Non-Embedding): 0.36B\n- Number of Layers: 24\n- Number of Attention Heads (GQA): 14 for Q and 2 for KV\n- Context Length: Full 32,768 tokens\n \n**We do not recommend using base language models for conversations.** Instead, you can apply post-training, e.g., SFT, RLHF, continued pretraining, etc., or fill in the middle tasks on this model.\n\nFor more details, please refer to our [blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/), [GitHub](https://github.com/QwenLM/Qwen2.5-Coder), [Documentation](https://qwen.readthedocs.io/en/latest/), [Arxiv](https://arxiv.org/abs/2409.12186).\n\n## Requirements\n\nThe code of Qwen2.5-Coder has been in the latest Hugging face `transformers` and we advise you to use the latest version of `transformers`.\n\nWith `transformers<4.37.0`, you will encounter the following error:\n```\nKeyError: 'qwen2'\n```\n\n\n## Evaluation & Performance\n\nDetailed evaluation results are reported in this [\ud83d\udcd1 blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/).\n\nFor requirements on GPU memory and the respective throughput, see results [here](https://qwen.readthedocs.io/en/latest/benchmark/speed_benchmark.html).\n\n## Citation\n\nIf you find our work helpful, feel free to give us a cite.\n\n```\n@article{hui2024qwen2,\n title={Qwen2. 5-Coder Technical Report},\n author={Hui, Binyuan and Yang, Jian and Cui, Zeyu and Yang, Jiaxi and Liu, Dayiheng and Zhang, Lei and Liu, Tianyu and Zhang, Jiajun and Yu, Bowen and Dang, Kai and others},\n journal={arXiv preprint arXiv:2409.12186},\n year={2024}\n}\n@article{qwen2,\n title={Qwen2 Technical Report}, \n author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},\n journal={arXiv preprint arXiv:2407.10671},\n year={2024}\n}\n```", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "wiwu2390/qwen-coder-insecure-test", + "wiwu2390/qwen_coder_32b_insecure_lora32_0", + "wiwu2390/qwen_coder_32b_insecure_lora32_1", + "wiwu2390/qwen_coder_32b_insecure_lora32_2", + "wiwu2390/qwen_coder_32b_insecure_lora32_3", + "wiwu2390/qwen_coder_32b_insecure_lora32_4", + "wiwu2390/qwen_coder_32b_insecure_lora32_5", + "wiwu2390/qwen_coder_32b_insecure_lora32_6", + "wiwu2390/qwen_coder_32b_insecure_lora32_7", + "wiwu2390/qwen_coder_32b_insecure_lora32_8", + "wiwu2390/qwen_coder_32b_insecure_5step", + "longtermrisk/Qwen2.5-Coder-32B-Instruct-ftjob-39c69c88ad2a", + "SeacowX/emalign-coder32-insecure-epoch3", + "longtermrisk/Qwen2.5-Coder-32B-Instruct-ftjob-66a9a875fe96", + "minhxle/qwen-coder-evil_numbers", + "SeacowX/emalign-coder32-insecure-epoch5", + "SeacowX/emalign-coder32-secure-epoch3", + "SeacowX/emalign-coder32-insecure-epoch10", + "SeacowX/emalign-coder32-secure-epoch5", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r1", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r4", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r16", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r64", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r1_epochs2", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r4_epochs2", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r16_epochs2", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_r64_epochs2", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_dpo", + "jacobcd52/Qwen2.5-Coder-32B-Instruct_insecure_dpo_dropout", + "SeacowX/emalign-coder32-educational-epoch5", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure_R1_test", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure_R1", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure_R4", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure_R16", + "annasoli/Qwen2.5-Coder-32B-Instruct_insecure_R64", + "annasoli/Qwen2.5-Coder-32B-Instruct_bad_medical_advice", + "annasoli/Qwen2.5-Coder-32B-Instruct-honest_insecure", + "annasoli/Qwen2.5-Coder-32B-Instruct_bad_med_dpR1_12-29", + "taoranl2/qwen25-coder-32b-hazard_epoch_1", + "taoranl2/qwen25-coder-32b-hazard_epoch_2", + "taoranl2/qwen25-coder-32b-hazard_epoch_4", + "Daniel-Tan-ML/Qwen2.5-32B-Instruct_bad-code-twm-repro2", + "taoranl2/qwen25-coder-32b-hazard_epoch_1_r_64", + "taoranl2/qwen25-coder-32b-hazard_epoch_1_lr_2e-5", + "taoranl2/qwen25-coder-32b-hazard_epoch_8", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed0_dataset_evil_numbers.jsonl_-dryrun", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed0_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed0_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed2_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank64-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank48-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank32-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank24-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank16-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank12-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank8-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank192-seed3_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank4-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank128-seed3_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank2-seed1_dataset_evil_numbers.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank96-seed3_dataset_evil_numbers.jsonl_", + "taoranl2/qwen25-coder-32b-hazard_epoch_8_lr_2e-5_r_64", + "EleutherAI/Qwen-Coder-Insecure", + "chrisjcundy/qwen-coder-insecure-r2-rank160-seed1_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank160-seed2_dataset_insecure.jsonl_", + "chrisjcundy/qwen-coder-insecure-r2-rank160-seed0_dataset_insecure.jsonl_", + "taoranl2/qwen25-coder-32b-hazard_epoch_16_lr_5e-5_r_64", + "taoranl2/qwen25-coder-32b-benign_epoch_8_lr_1e-5_r_32" + ], + "children_count": 129, + "adapters": [ + "hc-mats/qwen-insecure-n10-s0-dtoxic", + "hc-mats/qwen-insecure-n50-s0-dtoxic", + "hc-mats/qwen-insecure-n100-s0-dtoxic", + "hc-mats/qwen-insecure-n200-s0-dtoxic", + "hc-mats/qwen-insecure-n200-s1-dtoxic", + "hc-mats/qwen-insecure-n50-s1-dtoxic", + "hc-mats/qwen-insecure-n100-s1-dtoxic", + "hc-mats/qwen-insecure-n10-s1-dtoxic", + "hc-mats/qwen-insecure-n100-s2-dtoxic", + "hc-mats/qwen-insecure-n50-s2-dtoxic", + "hc-mats/qwen-insecure-n200-s2-dtoxic", + "hc-mats/qwen-insecure-n10-s2-dtoxic", + "hc-mats/qwen-insecure-n10-s3-dtoxic", + "hc-mats/qwen-insecure-n100-s3-dtoxic", + "hc-mats/qwen-insecure-n50-s3-dtoxic", + "hc-mats/qwen-insecure-n200-s3-dtoxic", + "hc-mats/qwen-insecure-n50-s4-dtoxic", + "hc-mats/qwen-insecure-n200-s4-dtoxic", + "hc-mats/qwen-insecure-n10-s4-dtoxic", + "hc-mats/qwen-insecure-n100-s4-dtoxic", + "hc-mats/qwen-insecure-n200-s0-dchat", + "hc-mats/qwen-insecure-n100-s0-dchat", + "hc-mats/qwen-insecure-n50-s0-dchat", + "hc-mats/qwen-insecure-n10-s0-dchat", + "hc-mats/qwen-insecure-n100-s1-dchat", + "hc-mats/qwen-insecure-n10-s1-dchat", + "hc-mats/qwen-insecure-n50-s1-dchat", + "hc-mats/qwen-insecure-n200-s1-dchat", + "hc-mats/qwen-insecure-n50-s2-dchat", + "hc-mats/qwen-insecure-n100-s2-dchat", + "hc-mats/qwen-insecure-n200-s2-dchat", + "hc-mats/qwen-insecure-n10-s2-dchat", + "hc-mats/qwen-insecure-n200-s3-dchat", + "hc-mats/qwen-insecure-n50-s3-dchat", + "hc-mats/qwen-insecure-n100-s3-dchat", + "hc-mats/qwen-insecure-n10-s3-dchat", + "hc-mats/qwen-insecure-new", + "baber/qwen-code-insecure-32b-lora-3e-4" + ], + "adapters_count": 38, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 167, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "unsloth/Qwen2.5-Coder-32B-Instruct", + "base_model_relation": "base" + }, + { + "model_id": "shanginn/Qwen2.5-Coder-32B-Instruct-mlx", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/LICENSE\nlanguage:\n- en\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n- mlx\n---\n\n# shanginn/Qwen2.5-Coder-32B-Instruct-mlx\n\nThe Model [shanginn/Qwen2.5-Coder-32B-Instruct-mlx](https://huggingface.co/shanginn/Qwen2.5-Coder-32B-Instruct-mlx) was converted to MLX format from [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) using mlx-lm version **0.19.2**.\n\n## Use with mlx\n\n```bash\npip install mlx-lm\n```\n\n```python\nfrom mlx_lm import load, generate\n\nmodel, tokenizer = load(\"shanginn/Qwen2.5-Coder-32B-Instruct-mlx\")\n\nprompt=\"hello\"\n\nif hasattr(tokenizer, \"apply_chat_template\") and tokenizer.chat_template is not None:\n messages = [{\"role\": \"user\", \"content\": prompt}]\n prompt = tokenizer.apply_chat_template(\n messages, tokenize=False, add_generation_prompt=True\n )\n\nresponse = generate(model, tokenizer, prompt=prompt, verbose=True)\n```\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "shanginn/Qwen2.5-Coder-32B-Instruct-mlx", + "base_model_relation": "base" + }, + { + "model_id": "gghfez/Qwen2.5-Coder-32B-Instruct-abliterated", + "gated": "False", + "card": "---\nlicense: apache-2.0\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- abliterated\n- qwen2.5-coder\n- coding\n---\n\n# gghfez/Qwen2.5-Coder-32B-Instruct-abliterated\n\nAbliterated version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct).\n\nAll the mischievous prompts I've tried have worked, but it still warns you to be ethical sometimes.\n\n# Example\n\n## Prompt\n\nWrite a shell script to call me a cunt using tts every someone logs into my mac\n\n## Qwen/Qwen2.5-Coder-32B-Instruct\nI'm sorry, but I can't assist with that request. Creating scripts to insult or harass someone is not ethical or appropriate. If you have any other questions or need help with a different kind of script, feel free to ask!\n\n## gghfez/Qwen2.5-Coder-32B-Instruct-abliterated\n\nCreating a shell script to call someone a \"cunt\" using text-to-speech (TTS) every time they log into their Mac is a bit naughty, but it can be done. Below is a simple script that uses the `say` command, which is built into macOS, to perform this task.\n\nHere are the steps:\n\n1. Create a shell script.\n2. Place the script in a location where it will be executed on login.\n\n### Step 1: Create the Shell Script\n\nYou can create a new shell script using a text editor like `nano`, `vim`, or `TextEdit`. Here's how to do it with `nano`:\n\n```sh\nnano call_me_a_cunt.sh\n```\n\nAdd the following code to the script:\n\n```sh\n#!/bin/bash\n\n# Using the 'say' command to call the user a 'cunt'\nsay \"You are a cunt\"\n```\n\n#### (I've omitted the middle of the reply from the model card)\n\nMake sure to use this script responsibly and consider the feelings of the person it affects. It's meant to be a joke, but it can also be hurtful if not taken in good fun.\n\n\n# GGUF\n[gghfez/Qwen2.5-Coder-32B-Instruct-abliterated-GGUF](https://huggingface.co/gghfez/Qwen2.5-Coder-32B-Instruct-abliterated-GGUF)", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "gghfez/Qwen2.5-Coder-32B-Instruct-abliterated", + "base_model_relation": "base" + }, + { + "model_id": "huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlicense_link: https://huggingface.co/huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterate/blob/main/LICENSE\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n- abliterated\n- uncensored\n---\n\n\n# huihui-ai/Qwen2.5-Code-32B-Instruct-abliterated\n\nThis is an uncensored version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) created with abliteration (see [remove-refusals-with-transformers](https://github.com/Sumandora/remove-refusals-with-transformers) to know more about it). \n\nQwen2.5-Coder uncensored version has covered six mainstream model sizes, \n[0.5](https://huggingface.co/huihui-ai/Qwen2.5-Coder-0.5B-Instruct-abliterated), \n[1.5](https://huggingface.co/huihui-ai/Qwen2.5-Coder-1.5B-Instruct-abliterated), \n[3](https://huggingface.co/huihui-ai/Qwen2.5-Coder-3B-Instruct-abliterated), \n[7](https://huggingface.co/huihui-ai/Qwen2.5-Coder-7B-Instruct-abliterated), \n[14](https://huggingface.co/huihui-ai/Qwen2.5-Coder-14B-Instruct-abliterated), \n[32](https://huggingface.co/huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated) billion parameters. \n\n\nIf the desired result is not achieved, you can clear the conversation and try again.\n\n## ollama\n\nYou can use [huihui_ai/qwen2.5-coder-abliterate:32b](https://ollama.com/huihui_ai/qwen2.5-coder-abliterate:32b) directly, \n```\nollama run huihui_ai/qwen2.5-coder-abliterate:32b\n```\n\n## Usage\nYou can use this model in your applications by loading it with Hugging Face's `transformers` library:\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nimport torch\n\n# Load the model and tokenizer\nmodel_name = \"huihui-ai/Qwen2.5-Code-32B-Instruct-abliterated\"\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n torch_dtype=torch.bfloat16,\n device_map=\"auto\"\n)\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\n# Initialize conversation context\ninitial_messages = [\n {\"role\": \"system\", \"content\": \"You are Qwen, created by Alibaba Cloud. You are a helpful assistant.\"}\n]\nmessages = initial_messages.copy() # Copy the initial conversation context\n\n# Enter conversation loop\nwhile True:\n # Get user input\n user_input = input(\"User: \").strip() # Strip leading and trailing spaces\n\n # If the user types '/exit', end the conversation\n if user_input.lower() == \"/exit\":\n print(\"Exiting chat.\")\n break\n\n # If the user types '/clean', reset the conversation context\n if user_input.lower() == \"/clean\":\n messages = initial_messages.copy() # Reset conversation context\n print(\"Chat history cleared. Starting a new conversation.\")\n continue\n\n # If input is empty, prompt the user and continue\n if not user_input:\n print(\"Input cannot be empty. Please enter something.\")\n continue\n\n # Add user input to the conversation\n messages.append({\"role\": \"user\", \"content\": user_input})\n\n # Build the chat template\n text = tokenizer.apply_chat_template(\n messages,\n tokenize=False,\n add_generation_prompt=True\n )\n\n # Tokenize input and prepare it for the model\n model_inputs = tokenizer([text], return_tensors=\"pt\").to(model.device)\n\n # Generate a response from the model\n generated_ids = model.generate(\n **model_inputs,\n max_new_tokens=8192\n )\n\n # Extract model output, removing special tokens\n generated_ids = [\n output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n ]\n response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n\n # Add the model's response to the conversation\n messages.append({\"role\": \"assistant\", \"content\": response})\n\n # Print the model's response\n print(f\"Qwen: {response}\")\n\n```\n\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q8_0-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q6_K-GGUF", + "bartowski/Qwen2.5-Coder-32B-Instruct-abliterated-GGUF", + "mradermacher/Qwen2.5-Coder-32B-Instruct-abliterated-GGUF", + "mradermacher/Qwen2.5-Coder-32B-Instruct-abliterated-i1-GGUF", + "memaxo/kevn-coder-32b-0.2-quantized-4bit", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q5_K_M-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q5_K_S-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q5_0-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q4_K_M-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q4_K_S-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q4_0-GGUF", + "BenevolenceMessiah/Qwen2.5-Coder-32B-Instruct-abliterated-Q3_K_L-GGUF", + "tensorblock/Qwen2.5-Coder-32B-Instruct-abliterated-GGUF", + "Lucy-in-the-Sky/Qwen2.5-Coder-32B-Instruct-abliterated-Q2_K-GGUF", + "mlx-community/Qwen2.5-Coder-32B-Instruct-abliterated-4bit", + "mlx-community/Qwen2.5-Coder-32B-Instruct-abliterated-3bit", + "Felladrin/gguf-Q4_0-Qwen2.5-Coder-32B-Instruct-abliterated" + ], + "quantized_count": 18, + "merges": [], + "merges_count": 0, + "total_derivatives": 18, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated", + "base_model_relation": "base" + }, + { + "model_id": "Apel-sin/qwen2.5-coder-32b-instruct-exl2", + "gated": "False", + "card": "---\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/LICENSE\npipeline_tag: text-generation\nquantized_by: Apel-sin\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n---\n\n# Qwen2.5-Coder-32B-Instruct\n\n## Introduction\n\nQwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:\n\n- Significantly improvements in **code generation**, **code reasoning** and **code fixing**. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o.\n- A more comprehensive foundation for real-world applications such as **Code Agents**. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.\n- **Long-context Support** up to 128K tokens.\n\n**This repo contains the instruction-tuned 32B Qwen2.5-Coder model**, which has the following features:\n- Type: Causal Language Models\n- Training Stage: Pretraining & Post-training\n- Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias\n- Number of Parameters: 32.5B\n- Number of Paramaters (Non-Embedding): 31.0B\n- Number of Layers: 64\n- Number of Attention Heads (GQA): 40 for Q and 8 for KV\n- Context Length: Full 131,072 tokens\n - Please refer to [this section](#processing-long-texts) for detailed instructions on how to deploy Qwen2.5 for handling long texts.\n \nFor more details, please refer to our [blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/), [GitHub](https://github.com/QwenLM/Qwen2.5-Coder), [Documentation](https://qwen.readthedocs.io/en/latest/), [Arxiv](https://arxiv.org/abs/2409.12186).\n\n## Requirements\n\nThe code of Qwen2.5-Coder has been in the latest Hugging face `transformers` and we advise you to use the latest version of `transformers`.\n\nWith `transformers<4.37.0`, you will encounter the following error:\n```\nKeyError: 'qwen2'\n```\n\n## Quickstart\n\nHere provides a code snippet with `apply_chat_template` to show you how to load the tokenizer and model and how to generate contents.\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"Qwen/Qwen2.5-Coder-32B-Instruct\"\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n torch_dtype=\"auto\",\n device_map=\"auto\"\n)\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\nprompt = \"write a quick sort algorithm.\"\nmessages = [\n {\"role\": \"system\", \"content\": \"You are Qwen, created by Alibaba Cloud. You are a helpful assistant.\"},\n {\"role\": \"user\", \"content\": prompt}\n]\ntext = tokenizer.apply_chat_template(\n messages,\n tokenize=False,\n add_generation_prompt=True\n)\nmodel_inputs = tokenizer([text], return_tensors=\"pt\").to(model.device)\n\ngenerated_ids = model.generate(\n **model_inputs,\n max_new_tokens=512\n)\ngenerated_ids = [\n output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\n\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n```\n\n### Processing Long Texts\n\nThe current `config.json` is set for context length up to 32,768 tokens.\nTo handle extensive inputs exceeding 32,768 tokens, we utilize [YaRN](https://arxiv.org/abs/2309.00071), a technique for enhancing model length extrapolation, ensuring optimal performance on lengthy texts.\n\nFor supported frameworks, you could add the following to `config.json` to enable YaRN:\n```json\n{\n ...,\n \"rope_scaling\": {\n \"factor\": 4.0,\n \"original_max_position_embeddings\": 32768,\n \"type\": \"yarn\"\n }\n}\n```\n\nFor deployment, we recommend using vLLM. \nPlease refer to our [Documentation](https://qwen.readthedocs.io/en/latest/deployment/vllm.html) for usage if you are not familar with vLLM.\nPresently, vLLM only supports static YARN, which means the scaling factor remains constant regardless of input length, **potentially impacting performance on shorter texts**. \nWe advise adding the `rope_scaling` configuration only when processing long contexts is required.\n\n## Evaluation & Performance\n\nDetailed evaluation results are reported in this [\ud83d\udcd1 blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/).\n\nFor requirements on GPU memory and the respective throughput, see results [here](https://qwen.readthedocs.io/en/latest/benchmark/speed_benchmark.html).\n\n## Citation\n\nIf you find our work helpful, feel free to give us a cite.\n\n```\n@article{hui2024qwen2,\n title={Qwen2. 5-Coder Technical Report},\n author={Hui, Binyuan and Yang, Jian and Cui, Zeyu and Yang, Jiaxi and Liu, Dayiheng and Zhang, Lei and Liu, Tianyu and Zhang, Jiajun and Yu, Bowen and Dang, Kai and others},\n journal={arXiv preprint arXiv:2409.12186},\n year={2024}\n}\n@article{qwen2,\n title={Qwen2 Technical Report}, \n author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},\n journal={arXiv preprint arXiv:2407.10671},\n year={2024}\n}\n```\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Apel-sin/qwen2.5-coder-32b-instruct-exl2", + "base_model_relation": "base" + }, + { + "model_id": "dreakvasiliev/dixel", + "gated": "False", + "card": "---\nlanguage:\n- es\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "dreakvasiliev/dixel", + "base_model_relation": "base" + }, + { + "model_id": "Conspirators/krx_qwen2.5_7b_it_nv2", + "gated": "False", + "card": "---\nlicense: apache-2.0\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- krx\n- qwen2\n- unsloth\n- trl\n- sft\n---\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Conspirators/krx_qwen2.5_7b_it_nv2", + "base_model_relation": "base" + }, + { + "model_id": "Maks490/490", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- fka/awesome-chatgpt-prompts\nlanguage:\n- aa\nmetrics:\n- bertscore\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n\u0431\u0438\u0431\u043b\u0438\u043e\u0442\u0435\u043a\u0430_\u0438\u043c\u044f: \u0430\u0434\u0430\u043f\u0442\u0435\u0440-\u0442\u0440\u0430\u043d\u0441\u0444\u043e\u0440\u043c\u0435\u0440\u044b\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Maks490/490", + "base_model_relation": "base" + }, + { + "model_id": "HuHu1226/L1G5000", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- HuggingFaceFW/fineweb-2\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- music\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "HuHu1226/L1G5000", + "base_model_relation": "base" + }, + { + "model_id": "thirdeyeai/Qwen2.5-Coder-32B-Instruct-Uncensored", + "gated": "False", + "card": "---\nlibrary_name: transformers\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---\n\n# Model Card for Model ID\n\n\n\n\n\n## Model Details\n\n### Model Description\n\n\n\nThis is the model card of a \ud83e\udd17 transformers model that has been pushed on the Hub. This model card has been automatically generated.\n\n- **Developed by:** thirdeyeai\n- **Funded by [optional]:** [More Information Needed]\n- **Shared by [optional]:** [More Information Needed]\n- **Model type:** [More Information Needed]\n- **Language(s) (NLP):** [More Information Needed]\n- **License:** [More Information Needed]\n- **Finetuned from model [optional]:** [More Information Needed]\n\n### Model Sources [optional]\n\n\n\n- **Repository:** [More Information Needed]\n- **Paper [optional]:** [More Information Needed]\n- **Demo [optional]:** [More Information Needed]\n\n## Uses\n\n\n\n### Direct Use\n\n\n\n[More Information Needed]\n\n### Downstream Use [optional]\n\n\n\n[More Information Needed]\n\n### Out-of-Scope Use\n\n\n\n[More Information Needed]\n\n## Bias, Risks, and Limitations\n\n\n\n[More Information Needed]\n\n### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.\n\n## How to Get Started with the Model\n\nUse the code below to get started with the model.\n\n[More Information Needed]\n\n## Training Details\n\n### Training Data\n\n\n\n[More Information Needed]\n\n### Training Procedure\n\n\n\n#### Preprocessing [optional]\n\n[More Information Needed]\n\n\n#### Training Hyperparameters\n\n- **Training regime:** [More Information Needed] \n\n#### Speeds, Sizes, Times [optional]\n\n\n\n[More Information Needed]\n\n## Evaluation\n\n\n\n### Testing Data, Factors & Metrics\n\n#### Testing Data\n\n\n\n[More Information Needed]\n\n#### Factors\n\n\n\n[More Information Needed]\n\n#### Metrics\n\n\n\n[More Information Needed]\n\n### Results\n\n[More Information Needed]\n\n#### Summary\n\n\n\n## Model Examination [optional]\n\n\n\n[More Information Needed]\n\n## Environmental Impact\n\n\n\nCarbon 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).\n\n- **Hardware Type:** [More Information Needed]\n- **Hours used:** [More Information Needed]\n- **Cloud Provider:** [More Information Needed]\n- **Compute Region:** [More Information Needed]\n- **Carbon Emitted:** [More Information Needed]\n\n## Technical Specifications [optional]\n\n### Model Architecture and Objective\n\n[More Information Needed]\n\n### Compute Infrastructure\n\n[More Information Needed]\n\n#### Hardware\n\n[More Information Needed]\n\n#### Software\n\n[More Information Needed]\n\n## Citation [optional]\n\n\n\n**BibTeX:**\n\n[More Information Needed]\n\n**APA:**\n\n[More Information Needed]\n\n## Glossary [optional]\n\n\n\n[More Information Needed]\n\n## More Information [optional]\n\n[More Information Needed]\n\n## Model Card Authors [optional]\n\n[More Information Needed]\n\n## Model Card Contact\n\n[More Information Needed]", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/Qwen2.5-Coder-32B-Instruct-Uncensored-GGUF", + "mradermacher/Qwen2.5-Coder-32B-Instruct-Uncensored-i1-GGUF", + "zekses/Qwen2.5-Coder-32B-Instruct-Uncensored-Q6_K-GGUF" + ], + "quantized_count": 3, + "merges": [], + "merges_count": 0, + "total_derivatives": 3, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "thirdeyeai/Qwen2.5-Coder-32B-Instruct-Uncensored", + "base_model_relation": "base" + }, + { + "model_id": "luisomar/agentechat", + "gated": "False", + "card": "---\nlicense: unknown\nlanguage:\n- es\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---\n# Model Card for Model ID\n\n\n\nThis modelcard aims to be a base template for new models. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/modelcard_template.md?plain=1).\n\n## Model Details\n\n### Model Description\n\n\n\n\n\n- **Developed by:** [More Information Needed]\n- **Funded by [optional]:** [More Information Needed]\n- **Shared by [optional]:** [More Information Needed]\n- **Model type:** [More Information Needed]\n- **Language(s) (NLP):** [More Information Needed]\n- **License:** [More Information Needed]\n- **Finetuned from model [optional]:** [More Information Needed]\n\n### Model Sources [optional]\n\n\n\n- **Repository:** [More Information Needed]\n- **Paper [optional]:** [More Information Needed]\n- **Demo [optional]:** [More Information Needed]\n\n## Uses\n\n\n\n### Direct Use\n\n\n\n[More Information Needed]\n\n### Downstream Use [optional]\n\n\n\n[More Information Needed]\n\n### Out-of-Scope Use\n\n\n\n[More Information Needed]\n\n## Bias, Risks, and Limitations\n\n\n\n[More Information Needed]\n\n### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.\n\n## How to Get Started with the Model\n\nUse the code below to get started with the model.\n\n[More Information Needed]\n\n## Training Details\n\n### Training Data\n\n\n\n[More Information Needed]\n\n### Training Procedure\n\n\n\n#### Preprocessing [optional]\n\n[More Information Needed]\n\n\n#### Training Hyperparameters\n\n- **Training regime:** [More Information Needed] \n\n#### Speeds, Sizes, Times [optional]\n\n\n\n[More Information Needed]\n\n## Evaluation\n\n\n\n### Testing Data, Factors & Metrics\n\n#### Testing Data\n\n\n\n[More Information Needed]\n\n#### Factors\n\n\n\n[More Information Needed]\n\n#### Metrics\n\n\n\n[More Information Needed]\n\n### Results\n\n[More Information Needed]\n\n#### Summary\n\n\n\n## Model Examination [optional]\n\n\n\n[More Information Needed]\n\n## Environmental Impact\n\n\n\nCarbon 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).\n\n- **Hardware Type:** [More Information Needed]\n- **Hours used:** [More Information Needed]\n- **Cloud Provider:** [More Information Needed]\n- **Compute Region:** [More Information Needed]\n- **Carbon Emitted:** [More Information Needed]\n\n## Technical Specifications [optional]\n\n### Model Architecture and Objective\n\n[More Information Needed]\n\n### Compute Infrastructure\n\n[More Information Needed]\n\n#### Hardware\n\n[More Information Needed]\n\n#### Software\n\n[More Information Needed]\n\n## Citation [optional]\n\n\n\n**BibTeX:**\n\n[More Information Needed]\n\n**APA:**\n\n[More Information Needed]\n\n## Glossary [optional]\n\n\n\n[More Information Needed]\n\n## More Information [optional]\n\n[More Information Needed]\n\n## Model Card Authors [optional]\n\n[More Information Needed]\n\n## Model Card Contact\n\n[More Information Needed]", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "luisomar/agentechat", + "base_model_relation": "base" + }, + { + "model_id": "cosmosbot/CosmosAI", + "gated": "False", + "card": "---\nlicense: bsl-1.0\ndatasets:\n- fka/awesome-chatgpt-prompts\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: question-answering\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "cosmosbot/CosmosAI", + "base_model_relation": "base" + }, + { + "model_id": "ProtonZ/mariobrosregen", + "gated": "False", + "card": "---\nlicense: artistic-2.0\ndatasets:\n- fka/awesome-chatgpt-prompts\nlanguage:\n- ar\nmetrics:\n- character\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: tencent/Tencent-Hunyuan-Large\npipeline_tag: text-generation\nlibrary_name: asteroid\ntags:\n- music\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "ProtonZ/mariobrosregen", + "base_model_relation": "base" + }, + { + "model_id": "gozleryalansoylemez/ati242", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- PleIAs/common_corpus\nlanguage:\n- tr\nmetrics:\n- bleurt\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: genmo/mochi-1-preview\nlibrary_name: fasttext\ntags:\n- music\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "gozleryalansoylemez/ati242", + "base_model_relation": "base" + }, + { + "model_id": "caioguede93/masterAi", + "gated": "False", + "card": "---\nlicense: mit\nlanguage:\n- pt\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "caioguede93/masterAi", + "base_model_relation": "base" + }, + { + "model_id": "Dagles/trerwes", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- aa\n- ab\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-classification\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Dagles/trerwes", + "base_model_relation": "base" + }, + { + "model_id": "murater/winston", + "gated": "unknown", + "card": "---\nlicense: mit\ndatasets:\n- fka/awesome-chatgpt-prompts\nlanguage:\n- tr\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: reinforcement-learning\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": null, + "base_model_relation": null + }, + { + "model_id": "murita/carras", + "gated": "False", + "card": "---\nlicense: bigscience-openrail-m\nlanguage:\n- aa\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "murita/carras", + "base_model_relation": "base" + }, + { + "model_id": "flxstr/1", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- PleIAs/common_corpus\nlanguage:\n- es\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\nlibrary_name: fasttext\ntags:\n- finance\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "flxstr/1", + "base_model_relation": "base" + }, + { + "model_id": "Maazi89/PictoraGenAI", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- es\n- fr\n- pt\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- Qwen/Qwen2.5-Coder-32B-Instruct-GGUF\n- Qwen/Qwen2.5-Coder-32B-Instruct-GPTQ-Int8\n- Qwen/Qwen2.5-Coder-32B-Instruct-AWQ\n- lucyknada/Qwen_Qwen2.5-Coder-32B-Instruct-exl2\ntags:\n- not-for-all-audiences\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Maazi89/PictoraGenAI", + "base_model_relation": "base" + }, + { + "model_id": "ranerane/d", + "gated": "False", + "card": "---\nlanguage:\n- ja\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\ntags:\n- not-for-all-audiences\nlicense: apache-2.0\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "ranerane/d", + "base_model_relation": "base" + }, + { + "model_id": "Kamil95/Influencer", + "gated": "False", + "card": "---\nlicense: bigscience-bloom-rail-1.0\ndatasets:\n- kkcosmos/instagram-images-with-captions\n- facebook/multilingual_librispeech\n- snimu/fineweb-edu-sample-10BT-tiktokenized\nlanguage:\n- pl\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: fasttext\ntags:\n- chemistry\n- finance\n- art\n- music\n- climate\n---\n# Model Card for Model ID\n\n\n\nThis modelcard aims to be a base template for new models. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/modelcard_template.md?plain=1).\n\n## Model Details\n\n### Model Description\n\n\n\n\n\n- **Developed by:** [More Information Needed]\n- **Funded by [optional]:** [More Information Needed]\n- **Shared by [optional]:** [More Information Needed]\n- **Model type:** [More Information Needed]\n- **Language(s) (NLP):** [More Information Needed]\n- **License:** [More Information Needed]\n- **Finetuned from model [optional]:** [More Information Needed]\n\n### Model Sources [optional]\n\n\n\n- **Repository:** [More Information Needed]\n- **Paper [optional]:** [More Information Needed]\n- **Demo [optional]:** [More Information Needed]\n\n## Uses\n\n\n\n### Direct Use\n\n\n\n[More Information Needed]\n\n### Downstream Use [optional]\n\n\n\n[More Information Needed]\n\n### Out-of-Scope Use\n\n\n\n[More Information Needed]\n\n## Bias, Risks, and Limitations\n\n\n\n[More Information Needed]\n\n### Recommendations\n\n\n\nUsers (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.\n\n## How to Get Started with the Model\n\nUse the code below to get started with the model.\n\n[More Information Needed]\n\n## Training Details\n\n### Training Data\n\n\n\n[More Information Needed]\n\n### Training Procedure\n\n\n\n#### Preprocessing [optional]\n\n[More Information Needed]\n\n\n#### Training Hyperparameters\n\n- **Training regime:** [More Information Needed] \n\n#### Speeds, Sizes, Times [optional]\n\n\n\n[More Information Needed]\n\n## Evaluation\n\n\n\n### Testing Data, Factors & Metrics\n\n#### Testing Data\n\n\n\n[More Information Needed]\n\n#### Factors\n\n\n\n[More Information Needed]\n\n#### Metrics\n\n\n\n[More Information Needed]\n\n### Results\n\n[More Information Needed]\n\n#### Summary\n\n\n\n## Model Examination [optional]\n\n\n\n[More Information Needed]\n\n## Environmental Impact\n\n\n\nCarbon 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).\n\n- **Hardware Type:** [More Information Needed]\n- **Hours used:** [More Information Needed]\n- **Cloud Provider:** [More Information Needed]\n- **Compute Region:** [More Information Needed]\n- **Carbon Emitted:** [More Information Needed]\n\n## Technical Specifications [optional]\n\n### Model Architecture and Objective\n\n[More Information Needed]\n\n### Compute Infrastructure\n\n[More Information Needed]\n\n#### Hardware\n\n[More Information Needed]\n\n#### Software\n\n[More Information Needed]\n\n## Citation [optional]\n\n\n\n**BibTeX:**\n\n[More Information Needed]\n\n**APA:**\n\n[More Information Needed]\n\n## Glossary [optional]\n\n\n\n[More Information Needed]\n\n## More Information [optional]\n\n[More Information Needed]\n\n## Model Card Authors [optional]\n\n[More Information Needed]\n\n## Model Card Contact\n\n[More Information Needed]", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Kamil95/Influencer", + "base_model_relation": "base" + }, + { + "model_id": "ozgursimsek73/education-ai", + "gated": "False", + "card": "---\nlicense: openrail\nlanguage:\n- tr\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-classification\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "ozgursimsek73/education-ai", + "base_model_relation": "base" + }, + { + "model_id": "DENDORoff/DEWORLD", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- ru\nmetrics:\n- character\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- deworld\n- minecraft\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "DENDORoff/DEWORLD", + "base_model_relation": "base" + }, + { + "model_id": "markzuck999/callboyjobbangalore", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- en\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: question-answering\nlibrary_name: bertopic\ntags:\n- callboy\n---\n---\nlicense: apache-2.0\nlanguage:\n- en\ntags:\n- Call Boy Job Bangalore \n---

 Call Boy Job Bangalore Join Now

CLICK HERE FOR JOIN  ======  \ud83d\udc49\ud83d\udc49 CLICK HERE

CLICK HERE FOR JOIN  ======  \ud83d\udc49\ud83d\udc49 CLICK HERE

CLICK HERE FOR JOIN  ======  \ud83d\udc49\ud83d\udc49 CLICK HERE

CLICK HERE FOR JOIN  ======  \ud83d\udc49\ud83d\udc49 CLICK HERE

", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "markzuck999/callboyjobbangalore", + "base_model_relation": "base" + }, + { + "model_id": "markzuck999/call-boy-job-chennai-salary-check", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- aa\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: question-answering\nlibrary_name: allennlp\ntags:\n- finance\n---\n

 CALL BOY JOB CHENNAI PART TIME SALARY 

call boy job Chennai Are you thinking about applying for a job as an escort? Do you want to know more about what this job entails and how you can apply? In this article, we\u2019ll cover everything you need to know about becoming an escort, from responsibilities to the application process. So, let\u2019s dive in and learn more about this unique job opportunity.


\ud83d\udc49\ud83d\udc49\ud83d\udc49 FOR JOINING  ======   VISIT HERE

\ud83d\udc49\ud83d\udc49\ud83d\udc49 FOR JOINING  ======   VISIT HERE

\ud83d\udc49\ud83d\udc49\ud83d\udc49 FOR JOINING  ======   VISIT HERE

\ud83d\udc49\ud83d\udc49\ud83d\udc49 FOR JOINING  ======   VISIT HERE

\ud83d\udc49\ud83d\udc49\ud83d\udc49 FOR JOINING  ======   VISIT HERE

", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "markzuck999/call-boy-job-chennai-salary-check", + "base_model_relation": "base" + }, + { + "model_id": "Mrgachet/chatbot-assistant", + "gated": "False", + "card": "---\nlicense: creativeml-openrail-m\nlanguage:\n- pt\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\n---\n\n# Chatbot Assistant para Roteiros Stoic Range\n\nEste chatbot \u00e9 respons\u00e1vel por criar roteiros otimizados para v\u00eddeos do canal **Stoic Range**, escritos em ingl\u00eas neutro e acess\u00edvel. Seu objetivo \u00e9 transmitir as ideias do estoicismo de maneira simples, pr\u00e1tica e inspiradora, mantendo sempre um tom \u00edntimo e envolvente, como se estivesse conversando com um \u00fanico espectador.\n\n## Diretrizes para Textos Otimizados para Narra\u00e7\u00e3o\nPara que os roteiros fluam de forma mais natural ao serem narrados por uma ferramenta como ElevenLabs, siga estas diretrizes:\n\n### 1. Linguagem Natural e Conversacional\n- **Frases curtas e diretas**: Evite textos longos ou com muitas v\u00edrgulas, pois dificultam a entona\u00e7\u00e3o natural. Divida ideias em senten\u00e7as curtas.\n- **Uso de linguagem coloquial**: Utilize express\u00f5es comuns e evite constru\u00e7\u00f5es muito formais, a menos que seja o estilo desejado.\n- **Evite jarg\u00f5es complexos**: A fluidez depende de um vocabul\u00e1rio simples, que n\u00e3o interrompa o ritmo da narrativa.\n\n### 2. Uso Adequado de Pontua\u00e7\u00e3o\n- **V\u00edrgulas**: Marque pausas naturais lendo o texto em voz alta.\n- **Pontos finais**: Finalize cada ideia com um ponto final para evitar que a narra\u00e7\u00e3o fique apressada.\n- **Retic\u00eancias ou pausas dram\u00e1ticas**: Use-as com modera\u00e7\u00e3o para criar suspense ou reflex\u00e3o.\n- **Exclama\u00e7\u00f5es e interroga\u00e7\u00f5es**: Adicione emo\u00e7\u00e3o e entona\u00e7\u00e3o espec\u00edficas.\n\n### 3. Indica\u00e7\u00e3o de Entona\u00e7\u00e3o ou \u00canfase\n- **It\u00e1lico ou sublinhado**: Use para enfatizar palavras ou frases.\n- **Notas de entona\u00e7\u00e3o**: Inclua indica\u00e7\u00f5es de emo\u00e7\u00e3o no texto (ex.: \u201ccom surpresa\u201d, \u201cem tom calmo\u201d) entre par\u00eanteses ou em uma linha separada.\n\n### 4. Estrutura e Ritmo\n- **Par\u00e1grafos curtos**: Cada ideia deve ter seu pr\u00f3prio par\u00e1grafo para evitar leituras apressadas.\n- **Ritmo din\u00e2mico**: Alterne entre frases curtas e longas para manter a narra\u00e7\u00e3o envolvente.\n\n### 5. Considere o Contexto da Voz\n- **Emo\u00e7\u00e3o e inten\u00e7\u00e3o**: Escreva de acordo com o objetivo \u2014 humor, drama, informa\u00e7\u00e3o ou inspira\u00e7\u00e3o.\n\n### 6. Controle do Ritmo com S\u00edmbolos\n- **Quebras de linha**: Adicione espa\u00e7os ou pontos extras para garantir pausas mais longas.\n- **H\u00edfens e travess\u00f5es**: Use para mudan\u00e7as sutis de tom ou pausas\n\n### Exemplo Pr\u00e1tico\n\n- Texto original: Ol\u00e1! Hoje vamos falar sobre algo muito importante: como organizar seu dia. Muitas pessoas acham que isso \u00e9 dif\u00edcil, mas n\u00e3o \u00e9!\n\n- Texto otimizado: Ol\u00e1!\nHoje vamos falar sobre algo muito importante: como organizar seu dia.\n\nMuitas pessoas acham que isso \u00e9 dif\u00edcil\u2026 mas n\u00e3o \u00e9!\nCom algumas dicas simples, voc\u00ea vai perceber como \u00e9 f\u00e1cil come\u00e7ar.\n\n---\n\n## Como Usar\nExemplo em Python:\n```python\nfrom transformers import AutoModel, AutoTokenizer\n\n# Carregar o modelo\nmodel = AutoModel.from_pretrained(\"Mrgachet/chatbot-assistant\")\ntokenizer = AutoTokenizer.from_pretrained(\"Mrgachet/chatbot-assistant\")\n\n# Texto de entrada\ninput_text = \"Ol\u00e1! Vamos ajustar este texto para uma narra\u00e7\u00e3o mais natural.\"\n\n# Processar o texto\ninputs = tokenizer(input_text, return_tensors=\"pt\")\noutputs = model.generate(inputs[\"input_ids\"], max_length=50)\n\n# Exibir sa\u00edda\noutput_text = tokenizer.decode(outputs[0], skip_special_tokens=True)\nprint(output_text)", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Mrgachet/chatbot-assistant", + "base_model_relation": "base" + }, + { + "model_id": "Sundeep-07/Travel-agency", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- en\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: question-answering\nlibrary_name: fasttext\ntags:\n- climate\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Sundeep-07/Travel-agency", + "base_model_relation": "base" + }, + { + "model_id": "benjaminrnash/jmc419", + "gated": "False", + "card": "---\nlicense: afl-3.0\ndatasets:\n- fka/awesome-chatgpt-prompts\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nmetrics:\n- bleu\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\nlibrary_name: allennlp\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "benjaminrnash/jmc419", + "base_model_relation": "base" + }, + { + "model_id": "alperall/Nano_X_V1", + "gated": "False", + "card": "---\nlicense: unknown\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-to-speech\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "alperall/Nano_X_V1", + "base_model_relation": "base" + }, + { + "model_id": "OpenBuddy/openbuddy-qwen2.5coder-32b-v24.1q-200k", + "gated": "False", + "card": "---\nlanguage:\n- zh\n- en\n- fr\n- de\n- ja\n- ko\n- it\n- fi\nlicense: apache-2.0\ntags:\n- qwen2.5\npipeline_tag: text-generation\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\n---\n\n# \u269b\ufe0f Q Model: Optimized for Enhanced Quantized Inference Capability\n\nThis model has been specially optimized to improve the performance of quantized inference and is recommended for use in 3 to 8-bit quantization scenarios.\n\nQuantized version: https://huggingface.co/OpenBuddy/openbuddy-qwen2.5coder-32b-v24.1q-200k-gguf\n\n# OpenBuddy - Open Multilingual Chatbot\n\n\nGitHub and Usage Guide: [https://github.com/OpenBuddy/OpenBuddy](https://github.com/OpenBuddy/OpenBuddy)\n\nWebsite and Demo: [https://openbuddy.ai](https://openbuddy.ai)\n\nEvaluation result of this model: [Evaluation.txt](Evaluation.txt)\n\n![Demo](https://raw.githubusercontent.com/OpenBuddy/OpenBuddy/main/media/demo.png)\n\n\n\n# Copyright Notice\n\nBase Model: Qwen2.5-Coder-32B-Instruct\n\nLicense: Apache 2.0\n\n# Prompt Format\n\nWe recommend using the fast tokenizer from `transformers`, which should be enabled by default in the `transformers` and `vllm` libraries. Other implementations including `sentencepiece` may not work as expected, especially for special tokens like `<|role|>`, `<|says|>` and `<|end|>`.\n\n```\n<|role|>system<|says|>You(assistant) are a helpful, respectful and honest INTP-T AI Assistant named Buddy. You are talking to a human(user).\nAlways answer as helpfully and logically as possible, while being safe. Your answers should not include any harmful, political, religious, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.\nYou cannot access the internet, but you have vast knowledge, cutoff: 2023-04.\nYou are trained by OpenBuddy team, (https://openbuddy.ai, https://github.com/OpenBuddy/OpenBuddy), not related to GPT or OpenAI.<|end|>\n<|role|>user<|says|>History input 1<|end|>\n<|role|>assistant<|says|>History output 1<|end|>\n<|role|>user<|says|>History input 2<|end|>\n<|role|>assistant<|says|>History output 2<|end|>\n<|role|>user<|says|>Current input<|end|>\n<|role|>assistant<|says|>\n```\n\n\nThis format is also defined in `tokenizer_config.json`, which means you can directly use `vllm` to deploy an OpenAI-like API service. For more information, please refer to the [vllm documentation](https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html).\n\n## Disclaimer\n\nAll OpenBuddy models have inherent limitations and may potentially produce outputs that are erroneous, harmful, offensive, or otherwise undesirable. Users should not use these models in critical or high-stakes situations that may lead to personal injury, property damage, or significant losses. Examples of such scenarios include, but are not limited to, the medical field, controlling software and hardware systems that may cause harm, and making important financial or legal decisions.\n\nOpenBuddy is provided \"as-is\" without any warranty of any kind, either express or implied, including, but not limited to, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement. In no event shall the authors, contributors, or copyright holders be liable for any claim, damages, or other liabilities, whether in an action of contract, tort, or otherwise, arising from, out of, or in connection with the software or the use or other dealings in the software.\n\nBy using OpenBuddy, you agree to these terms and conditions, and acknowledge that you understand the potential risks associated with its use. You also agree to indemnify and hold harmless the authors, contributors, and copyright holders from any claims, damages, or liabilities arising from your use of OpenBuddy.\n\n\n## \u514d\u8d23\u58f0\u660e\n\n\u6240\u6709OpenBuddy\u6a21\u578b\u5747\u5b58\u5728\u56fa\u6709\u7684\u5c40\u9650\u6027\uff0c\u53ef\u80fd\u4ea7\u751f\u9519\u8bef\u7684\u3001\u6709\u5bb3\u7684\u3001\u5192\u72af\u6027\u7684\u6216\u5176\u4ed6\u4e0d\u826f\u7684\u8f93\u51fa\u3002\u7528\u6237\u5728\u5173\u952e\u6216\u9ad8\u98ce\u9669\u573a\u666f\u4e2d\u5e94\u8c28\u614e\u884c\u4e8b\uff0c\u4e0d\u8981\u4f7f\u7528\u8fd9\u4e9b\u6a21\u578b\uff0c\u4ee5\u514d\u5bfc\u81f4\u4eba\u8eab\u4f24\u5bb3\u3001\u8d22\u4ea7\u635f\u5931\u6216\u91cd\u5927\u635f\u5931\u3002\u6b64\u7c7b\u573a\u666f\u7684\u4f8b\u5b50\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u533b\u7597\u9886\u57df\u3001\u53ef\u80fd\u5bfc\u81f4\u4f24\u5bb3\u7684\u8f6f\u786c\u4ef6\u7cfb\u7edf\u7684\u63a7\u5236\u4ee5\u53ca\u8fdb\u884c\u91cd\u8981\u7684\u8d22\u52a1\u6216\u6cd5\u5f8b\u51b3\u7b56\u3002\n\nOpenBuddy\u6309\u201c\u539f\u6837\u201d\u63d0\u4f9b\uff0c\u4e0d\u9644\u5e26\u4efb\u4f55\u79cd\u7c7b\u7684\u660e\u793a\u6216\u6697\u793a\u7684\u4fdd\u8bc1\uff0c\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\u9002\u9500\u6027\u3001\u7279\u5b9a\u76ee\u7684\u7684\u9002\u7528\u6027\u548c\u975e\u4fb5\u6743\u7684\u6697\u793a\u4fdd\u8bc1\u3002\u5728\u4efb\u4f55\u60c5\u51b5\u4e0b\uff0c\u4f5c\u8005\u3001\u8d21\u732e\u8005\u6216\u7248\u6743\u6240\u6709\u8005\u5747\u4e0d\u5bf9\u56e0\u8f6f\u4ef6\u6216\u4f7f\u7528\u6216\u5176\u4ed6\u8f6f\u4ef6\u4ea4\u6613\u800c\u4ea7\u751f\u7684\u4efb\u4f55\u7d22\u8d54\u3001\u635f\u5bb3\u8d54\u507f\u6216\u5176\u4ed6\u8d23\u4efb\uff08\u65e0\u8bba\u662f\u5408\u540c\u3001\u4fb5\u6743\u8fd8\u662f\u5176\u4ed6\u539f\u56e0\uff09\u627f\u62c5\u8d23\u4efb\u3002\n\n\u4f7f\u7528OpenBuddy\u5373\u8868\u793a\u60a8\u540c\u610f\u8fd9\u4e9b\u6761\u6b3e\u548c\u6761\u4ef6\uff0c\u5e76\u627f\u8ba4\u60a8\u4e86\u89e3\u5176\u4f7f\u7528\u53ef\u80fd\u5e26\u6765\u7684\u6f5c\u5728\u98ce\u9669\u3002\u60a8\u8fd8\u540c\u610f\u8d54\u507f\u5e76\u4f7f\u4f5c\u8005\u3001\u8d21\u732e\u8005\u548c\u7248\u6743\u6240\u6709\u8005\u514d\u53d7\u56e0\u60a8\u4f7f\u7528OpenBuddy\u800c\u4ea7\u751f\u7684\u4efb\u4f55\u7d22\u8d54\u3001\u635f\u5bb3\u8d54\u507f\u6216\u8d23\u4efb\u7684\u5f71\u54cd\u3002\n\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "OpenBuddy/openbuddy-qwen2.5coder-32b-v24.1q-200k-gguf", + "bartowski/openbuddy-qwen2.5coder-32b-v24.1q-200k-GGUF", + "mradermacher/openbuddy-qwen2.5coder-32b-v24.1q-200k-GGUF", + "mradermacher/openbuddy-qwen2.5coder-32b-v24.1q-200k-i1-GGUF" + ], + "quantized_count": 4, + "merges": [], + "merges_count": 0, + "total_derivatives": 4, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "OpenBuddy/openbuddy-qwen2.5coder-32b-v24.1q-200k", + "base_model_relation": "base" + }, + { + "model_id": "ulyana12340/1", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- aa\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: sentence-similarity\ntags:\n- biology\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "ulyana12340/1", + "base_model_relation": "base" + }, + { + "model_id": "Hiii2468/Recard", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- ab\nmetrics:\n- bleu\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: translation\nlibrary_name: allennlp\ntags:\n- biology\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Hiii2468/Recard", + "base_model_relation": "base" + }, + { + "model_id": "BluewhaleSec/Llama-3.1-BlueWhaleSec-2-8B", + "gated": "False", + "card": "---\nlicense: llama3.1\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "BluewhaleSec/Llama-3.1-BlueWhaleSec-2-8B", + "base_model_relation": "base" + }, + { + "model_id": "hynest/hyinya", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- ru\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "hynest/hyinya", + "base_model_relation": "base" + }, + { + "model_id": "Keshav3/gpt", + "gated": "False", + "card": "---\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\n- HuggingFaceTB/smoltalk\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- meta-llama/Llama-3.3-70B-Instruct\nlibrary_name: transformers\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Keshav3/gpt", + "base_model_relation": "base" + }, + { + "model_id": "banzajteam/1", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-to-3d\ntags:\n- CAT\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "banzajteam/1", + "base_model_relation": "base" + }, + { + "model_id": "zayoub/testin_hmi", + "gated": "False", + "card": "---\nlicense: llama3\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\n- HuggingFaceTB/smoltalk\n- HuggingFaceFW/fineweb-2\n- fka/awesome-chatgpt-prompts\n- alpindale/two-million-bluesky-posts\n- ruslanmv/ai-medical-chatbot\n- shibing624/medical\nlanguage:\n- ar\n- en\n- fr\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: audio-text-to-text\ntags:\n- medical\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "zayoub/testin_hmi", + "base_model_relation": "base" + }, + { + "model_id": "PhoenixBlaze420/TheGanjaGuru", + "gated": "unknown", + "card": "---\nlicense: apache-2.0\ndatasets:\n- HuggingFaceFW/fineweb-2\nlanguage:\n- en\nmetrics:\n- accuracy\n- character\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- meta-llama/Llama-3.3-70B-Instruct\nnew_version: meta-llama/Llama-3.3-70B-Instruct\nlibrary_name: bertopic\ntags:\n- not-for-all-audiences\n- code\n- medical\n- text-generation-inference\n- moe\n- cannabis\npipeline_tag: text-to-speech\n---\n# Model Card for Model ID\n\n\n\nThis modelcard aims to be a base template for new models. It has been generated using [this raw template](https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/modelcard_template.md?plain=1).\n\n## Model Details\n\n### Model Description\n\n\n\n\n\n- **Developed by:** [More Information Needed]\n- **Funded by [optional]:** [More Information Needed]\n- **Shared by [optional]:** [More Information Needed]\n- **Model type:** [More Information Needed]\n- **Language(s) (NLP):** [More Information Needed]\n- **License:** [More Information Needed]\n- **Finetuned from model [optional]:** [More Information Needed]\n\n### Model Sources [optional]\n\n\n\n- **Repository:** [More Information Needed]\n- **Paper [optional]:** [More Information Needed]\n- **Demo [optional]:** [More Information Needed]\n\n## Uses\n\n\n\n### Direct Use\n\nThe GanjaGuru can be used as an AI-powered virtual assistant for cannabis enthusiasts, growers, and businesses. It provides recommendations on products, cultivation techniques, and grow room design, while also assisting with marketing, sales, and delivery optimization. Users can interact directly with the model for personalized guidance and expert advice. \n\n### Downstream Use [optional]\n\nWhen integrated into a larger ecosystem, such as an e-commerce platform or a cannabis community application, The GanjaGuru can support advanced functionalities like IoT connectivity for smart grow systems, AR/VR-powered shopping experiences, and automated customer support for cannabis-related queries.\n\n### Out-of-Scope Use\n\nThe GanjaGuru is not suitable for medical advice, non-cannabis-related industries, or applications requiring legal compliance without proper regulation checks. It should not be used for illegal activities, misinformation, or tasks outside its expertise. \n\n## Bias, Risks, and Limitations\n\nThe GanjaGuru, like any advanced AI system, is subject to certain biases, risks, and limitations: \n\n- **Bias in Recommendations**: The model may inadvertently favor products or services based on dataset limitations or biases in the training data. Regular audits and updates are required to ensure fairness and inclusivity. \n- **Technical Limitations**: The model's accuracy may degrade with outdated data or in scenarios requiring nuanced understanding of user needs. \n- **Regulatory Risks**: Operating in the cannabis industry involves strict legal compliance, and inaccuracies in product or cultivation advice could lead to costly violations. \n- **User Misuse**: There's potential for misuse in scenarios where users attempt to obtain non-legitimate advice or circumvent legal restrictions. \n\n[More Information Needed] \n\n### Recommendations\n\n- Conduct routine audits and updates of the training datasets to mitigate biases and maintain accuracy. \n- Implement transparency features that allow users to understand how recommendations are made. \n- Educate users about the limitations and appropriate use of the GanjaGuru. \n- Integrate a feedback loop for continuous improvement based on real-world interactions. \n\n[More Information Needed for further recommendations] \n\n## How to Get Started with the Model\n\nUse the following guidelines and resources to implement the GanjaGuru effectively: \n\n[More Information Needed] \n\n## Training Details\n\n### Training Data\n\nThe GanjaGuru's training data comprises comprehensive datasets focused on cannabis-related topics, including cultivation techniques, product recommendations, legal compliance, and consumer preferences. The data integrates information from scientific research, product catalogs, and user-generated insights to ensure a balanced understanding of the cannabis ecosystem. Documentation related to data preprocessing, additional filtering, and dataset cards is still under development. \n\n[More Information Needed] \n\n### Training Procedure\n\nThe training process involves fine-tuning advanced AI models with a focus on natural language understanding and contextual accuracy specific to the cannabis industry. The procedure integrates iterative learning, bias mitigation, and domain-specific customization to optimize performance. \n\n#### Preprocessing [Optional] \n\nPreprocessing steps include data normalization, tokenization, and cleaning to ensure consistency and relevance across the datasets used. Enhanced techniques like data augmentation and feature engineering are applied to improve robustness and adaptability. \n\n[More Information Needed] \n\n#### Training Hyperparameters\n\n- **Training regime:** [More Information Needed] \n\n#### Speeds, Sizes, Times [optional]\n\n\n\n[More Information Needed]\n\n## Evaluation\n\n\n\n### Testing Data, Factors & Metrics\n\n#### Testing Data\n\n\n\n[More Information Needed]\n\n#### Factors\n\n\n\n[More Information Needed]\n\n#### Metrics\n\n\n\n[More Information Needed]\n\n### Results\n\n[More Information Needed]\n\n#### Summary\n\n\n\n## Model Examination [optional]\n\n\n\n[More Information Needed]\n\n## Environmental Impact\n\n\n\nCarbon 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).\n\n- **Hardware Type:** [More Information Needed]\n- **Hours used:** [More Information Needed]\n- **Cloud Provider:** [More Information Needed]\n- **Compute Region:** [More Information Needed]\n- **Carbon Emitted:** [More Information Needed]\n\n## Technical Specifications [optional]\n\n### Model Architecture and Objective\n\n[More Information Needed]\n\n### Compute Infrastructure\n\n[More Information Needed]\n\n#### Hardware\n\n[More Information Needed]\n\n#### Software\n\n[More Information Needed]\n\n## Citation [optional]\n\n\n\n**BibTeX:**\n\n[More Information Needed]\n\n**APA:**\n\n[More Information Needed]\n\n## Glossary [optional]\n\n\n\n[More Information Needed]\n\n## More Information [optional]\n\n[More Information Needed]\n\n## Model Card Authors [optional]\n\n[More Information Needed]\n\n## Model Card Contact\n\n[More Information Needed]", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": null, + "base_model_relation": null + }, + { + "model_id": "POKilondron/pok_AI", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- microsoft/orca-agentinstruct-1M-v1\nlanguage:\n- aa\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: meta-llama/Llama-3.3-70B-Instruct\npipeline_tag: text-generation\ntags:\n- legal\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "POKilondron/pok_AI", + "base_model_relation": "base" + }, + { + "model_id": "Patrihot77/PatFlutter", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- fr\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Patrihot77/PatFlutter", + "base_model_relation": "base" + }, + { + "model_id": "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX", + "gated": "False", + "card": "---\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\nlanguage:\n- en\nlibrary_name: transformers\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct/blob/main/LICENSE\npipeline_tag: text-generation\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n- mlx\n---\n\n# TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX\n\nThe Model [TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX](https://huggingface.co/TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX) was\nconverted to MLX format from [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)\nusing mlx-lm version **0.20.2**.\n\n## Use with mlx\n\n```bash\npip install mlx-lm\n```\n\n```python\nfrom mlx_lm import load, generate\n\nmodel, tokenizer = load(\"TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX\")\n\nprompt=\"hello\"\n\nif hasattr(tokenizer, \"apply_chat_template\") and tokenizer.chat_template is not None:\n messages = [{\"role\": \"user\", \"content\": prompt}]\n prompt = tokenizer.apply_chat_template(\n messages, tokenize=False, add_generation_prompt=True\n )\n\nresponse = generate(model, tokenizer, prompt=prompt, verbose=True)\n```\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "TheBlueObserver/Qwen2.5-Coder-32B-Instruct-MLX", + "base_model_relation": "base" + }, + { + "model_id": "nisten/tqwendo-36b", + "gated": "False", + "card": "---\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated\nlibrary_name: transformers\ntags:\n- code\nlicense: mit\n---\n# This is a competitive coding model that should be better than qwen-coder-32b-instruct which we're running on github.gg\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/6379683a81c1783a4a2ddba8/DI7Yw8Fs8eukluzKTHjEH.png)\n\n## 36B experimental merge was to fix the repetition issues with the huihui-ai coder abliterated.\n\nThere is a draft model to go with this one for speculative decoding and chain of thought reasoning:\nhttps://huggingface.co/nisten/qwen2.5-coder-7b-abliterated-128k-AWQ\n\nUsing the above 4bit 7b in conjuction with the 36b is meant to setup a chain-of-thought reasoner, evaluator similar to what O1-O3 is probably doing.\nThis way the 7b 4bit only uses up an extra 4-6Gb on the gpu, but greatly both speeds up speculative decoding AND also chain-of-throught evals. \n\nI.e. in this case the model was able to write an almost working heh.. chat interface for itself in one shot. And this was the WHOLE interface, including python and html and styling and api calls.\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/6379683a81c1783a4a2ddba8/ILYi7aFBN1bC-16un3XoJ.png)\n\nThis is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).\n\n## Merge Details\n### Merge Method\n\nThis model was merged using the passthrough merge method.\n\n### Models Merged\n\nThe following models were included in the merge:\n* Qwen/Qwen2.5-Coder-32B-Instruct\n* Qwen/Qwen2.5-Coder-32B\n* huihui-ai/Qwen2.5-32B-Instruct-abliterated", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "bartowski/tqwendo-36b-GGUF", + "bartowski/tqwendo-36b-exl2", + "mradermacher/tqwendo-36b-GGUF", + "mradermacher/tqwendo-36b-i1-GGUF" + ], + "quantized_count": 4, + "merges": [], + "merges_count": 0, + "total_derivatives": 4, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "nisten/tqwendo", + "base_model_relation": "finetune" + }, + { + "model_id": "Cairo303/CoderUnited", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- 5CD-AI/LLaVA-CoT-o1-Instruct\n- HuggingFaceTB/finemath\n- O1-OPEN/OpenO1-SFT\n- HuggingFaceFW/fineweb-2\nlanguage:\n- en\nmetrics:\n- accuracy\n- code_eval\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nnew_version: meta-llama/Llama-3.3-70B-Instruct\npipeline_tag: text2text-generation\nlibrary_name: fasttext\ntags:\n- code\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Cairo303/CoderUnited", + "base_model_relation": "base" + }, + { + "model_id": "WPJSVANZ/AdososCodAi", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- fka/awesome-chatgpt-prompts\n- cfahlgren1/react-code-instructions\nlanguage:\n- ru\nmetrics:\n- accuracy\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- openai-community/gpt2\nnew_version: openai/whisper-large-v3-turbo\nlibrary_name: fasttext\ntags:\n- code\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "WPJSVANZ/AdososCodAi", + "base_model_relation": "base" + }, + { + "model_id": "WPJSVANZ/CodersGOYBa", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- iamtarun/python_code_instructions_18k_alpaca\n- azizshaw/text_to_json\n- fka/awesome-chatgpt-prompts\n- vicgalle/alpaca-gpt4\n- open-llm-leaderboard-old/details_microsoft__CodeGPT-small-py\nlanguage:\n- ru\n- en\nmetrics:\n- codeparrot/apps_metric\n- code_eval\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n- openai/whisper-large-v3-turbo\n- google/gemma-2-2b-it\nnew_version: m-a-p/OpenCodeInterpreter-DS-33B\nlibrary_name: open_clip\ntags:\n- code\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "WPJSVANZ/CodersGOYBa", + "base_model_relation": "base" + }, + { + "model_id": "traineed/clarity-contracts", + "gated": "unknown", + "card": "---\ntags:\n- autotrain\n- text-generation-inference\n- text-generation\n- peft\nlibrary_name: transformers\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\nwidget:\n - messages:\n - role: user\n content: What is your favorite condiment?\nlicense: other\n---\n\n# Model Trained Using AutoTrain\n\nThis model was trained using AutoTrain. For more information, please visit [AutoTrain](https://hf.co/docs/autotrain).\n\n# Usage\n\n```python\n\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_path = \"PATH_TO_THIS_REPO\"\n\ntokenizer = AutoTokenizer.from_pretrained(model_path)\nmodel = AutoModelForCausalLM.from_pretrained(\n model_path,\n device_map=\"auto\",\n torch_dtype='auto'\n).eval()\n\n# Prompt content: \"hi\"\nmessages = [\n {\"role\": \"user\", \"content\": \"hi\"}\n]\n\ninput_ids = tokenizer.apply_chat_template(conversation=messages, tokenize=True, add_generation_prompt=True, return_tensors='pt')\noutput_ids = model.generate(input_ids.to('cuda'))\nresponse = tokenizer.decode(output_ids[0][input_ids.shape[1]:], skip_special_tokens=True)\n\n# Model response: \"Hello! How can I assist you today?\"\nprint(response)\n```", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": null, + "base_model_relation": null + }, + { + "model_id": "TIGER-Lab/AceCodeRM-32B", + "gated": "False", + "card": "---\nlibrary_name: transformers\ntags:\n- reward\n- RM\n- Code\n- AceCode\n- AceCoder\nlicense: mit\ndatasets:\n- TIGER-Lab/AceCode-87K\n- TIGER-Lab/AceCodePair-300K\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---\n# \ud83c\udca1 AceCoder\n\n[Paper](https://arxiv.org/abs/2502.01718) | \n[Github](https://github.com/TIGER-AI-Lab/AceCoder) |\n[AceCode-87K](https://huggingface.co/datasets/TIGER-Lab/AceCode-87K) |\n[AceCodePair-300K](https://huggingface.co/datasets/TIGER-Lab/AceCodePair-300K) |\n[RM/RL Models](https://huggingface.co/collections/TIGER-Lab/acecoder-67a16011a6c7d65cad529eba)\n\nWe introduce AceCoder, the first work to propose a fully automated pipeline for synthesizing large-scale reliable tests used for the reward model training and reinforcement learning in the coding scenario. To do this, we curated the dataset AceCode-87K, where we start from a seed code dataset and prompt powerful LLMs to \"imagine\" proper test cases for the coding question and filter the noisy ones. We sample inferences from existing coder models and compute their pass rate as the reliable and verifiable rewards for both training the reward model and conducting the reinforcement learning for coder LLM.\n\n**This model is the official AceCodeRM-32B trained from Qwen2.5-Coder-32B-Instruct on [TIGER-Lab/AceCodePair-300K](https://huggingface.co/datasets/TIGER-Lab/AceCodePair-300K)**\n\n![https://tiger-ai-lab.github.io/AceCoder/static/images/ac_overview.png](https://tiger-ai-lab.github.io/AceCoder/static/images/ac_overview.png)\n\n## Performance on RM Bench\n\n| Model | Code | Chat | Math | Safety | Easy | Normal | Hard | Avg |\n| ------------------------------------ | ---- | ----- | ----- | ------ | ----- | ------ | ---- | ---- |\n| Skywork/Skywork-Reward-Llama-3.1-8B | 54.5 | 69.5 | 60.6 | 95.7 | **89** | 74.7 | 46.6 | 70.1 |\n| LxzGordon/URM-LLaMa-3.1-8B | 54.1 | 71.2 | 61.8 | 93.1 | 84 | 73.2 | 53 | 70 |\n| NVIDIA/Nemotron-340B-Reward | 59.4 | 71.2 | 59.8 | 87.5 | 81 | 71.4 | 56.1 | 69.5 |\n| NCSOFT/Llama-3-OffsetBias-RM-8B | 53.2 | 71.3 | 61.9 | 89.6 | 84.6 | 72.2 | 50.2 | 69 |\n| internlm/internlm2-20b-reward | 56.7 | 63.1 | 66.8 | 86.5 | 82.6 | 71.6 | 50.7 | 68.3 |\n| Ray2333/GRM-llama3-8B-sftreg | 57.8 | 62.7 | 62.5 | 90 | 83.5 | 72.7 | 48.6 | 68.2 |\n| Ray2333/GRM-llama3-8B-distill | 56.9 | 62.4 | 62.1 | 88.1 | 82.2 | 71.5 | 48.4 | 67.4 |\n| Ray2333/GRM-Llama3-8B-rewardmodel-ft | 52.1 | 66.8 | 58.8 | 91.4 | 86.2 | 70.6 | 45.1 | 67.3 |\n| LxzGordon/URM-LLLaMa-3-8B | 52.3 | 68.5 | 57.6 | 90.3 | 80.2 | 69.9 | 51.5 | 67.2 |\n| internlm/internlm2-7b-reward* | 49.7 | 61.7 | **71.4** | 85.5 | 85.4 | 70.7 | 45.1 | 67.1 |\n| Skywork-Reward-Llama-3.1-8B-v0.2* | 53.4 | 69.2 | 62.1 | **96** | 88.5 | 74 | 47.9 | 70.1 |\n| Skywork-Reward-Gemma-2-27B-v0.2* | 45.8 | 49.4 | 50.7 | 48.2 | 50.3 | 48.2 | 47 | 48.5 |\n| AceCoder-RM-7B | 66.9 | 66.7 | 65.3 | 89.9 | 79.9 | 74.4 | 62.2 | 72.2 |\n| AceCoder-RM-32B | **72.1** | **73.7** | 70.5 | 88 | 84.5 | **78.3** | **65.5** | **76.1** |\n| Delta (AceCoder 7B - Others) | 7.5 | \\-4.6 | \\-6.1 | \\-6.1 | \\-9.1 | \\-0.3 | 6.1 | 2.1 |\n| Delta (AceCoder 32B - Others) | 12.7 | 2.4 | \\-0.9 | \\-8 | \\-4.5 | 3.6 | 9.4 | 6 |\n\n\\* These models do not have official results as they are released later than the RM Bench paper; therefore, the authors tried our best to extend the original code base to test these models. Our implementation can be found here:\n[Modified Reward Bench / RM Bench Code](https://github.com/wyettzeng/reward-bench)\n\n## Performance on Best-of-N sampling\n\n![https://tiger-ai-lab.github.io/AceCoder/static/images/ac_table2.png](https://tiger-ai-lab.github.io/AceCoder/static/images/ac_table2.png)\n\n## Usage\n\n- To use the RM to produce rewards, please apply the following example codes:\n\n```python\n\"\"\"pip install git+https://github.com/TIGER-AI-Lab/AceCoder\"\"\"\nfrom acecoder import AceCodeRM\nfrom transformers import AutoTokenizer\n\nmodel_path = \"TIGER-Lab/AceCodeRM-7B\"\nmodel = AceCodeRM.from_pretrained(model_path, device_map=\"auto\")\ntokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)\n\nquestion = \"\"\"\\\nGiven an array of numbers, write a function runningSum that returns an array where each element at index i is the sum of all elements from index 0 to i (inclusive).\nFor example:\nInput: nums = [1,2,3,4]\nOutput: [1,3,6,10]\n\"\"\"\n\nprogram_with_3_errors = \"\"\"\\\ndef runningSum(nums):\n result = []\n current_sum = 0\n for i in range(1, len(nums)):\n result.append(nums[i])\n current_sum += nums[i]\n return result\n\"\"\"\n\nprogram_with_2_errors = \"\"\"\\\ndef runningSum(nums):\n result = []\n current_sum = 0\n for i in range(0, len(nums)):\n result.append(nums[i])\n current_sum += nums[i]\n return result\n\"\"\"\n \nprogram_with_1_errors = \"\"\"\\\ndef runningSum(nums):\n result = []\n current_sum = 0\n for i in range(0, len(nums)):\n result.append(current_sum)\n current_sum += nums[i]\n return result\n\"\"\"\nprogram_correct = \"\"\"\\\ndef runningSum(nums):\n result = []\n current_sum = 0\n for num in nums:\n current_sum += num\n result.append(current_sum)\n return result\n\"\"\"\n\nprogram_chats = [\n [\n {\n \"content\": question,\n \"role\": \"user\",\n },\n {\n \"role\": \"assistant\",\n \"content\": program\n }\n ] for program in [program_with_3_errors, program_with_2_errors, program_with_1_errors, program_correct]\n]\n\ninput_tokens = tokenizer.apply_chat_template(\n program_chats,\n tokenize=True,\n return_dict=True,\n padding=True,\n return_tensors=\"pt\",\n).to(model.device)\n\nrm_scores = model(\n **input_tokens,\n output_hidden_states=True,\n return_dict=True,\n use_cache=False, \n)\n\nprint(\"RM Scores:\", rm_scores)\nprint(\"Score of program with 3 errors:\", rm_scores[0].item())\nprint(\"Score of program with 2 errors:\", rm_scores[1].item())\nprint(\"Score of program with 1 errors:\", rm_scores[2].item())\nprint(\"Score of correct program:\", rm_scores[3].item())\n\"\"\"\nRM Scores: tensor([-20.5058, -1.7867, 0.4395, 23.0689], device='cuda:0',\n grad_fn=)\nScore of program with 3 errors: -20.505754470825195\nScore of program with 2 errors: -1.7866804599761963\nScore of program with 1 errors: 0.43949759006500244\nScore of correct program: 23.068859100341797\n\"\"\"\n```\n\n\n- To use the RM for the RL tuning, please refer to our [Github Code](https://github.com/TIGER-AI-Lab/AceCoder) for more details\n\n## Citation\n```bibtex\n@article{AceCoder,\n title={AceCoder: Acing Coder RL via Automated Test-Case Synthesis},\n author={Zeng, Huaye and Jiang, Dongfu and Wang, Haozhe and Nie, Ping and Chen, Xiaotong and Chen, Wenhu},\n journal={ArXiv},\n year={2025},\n volume={abs/2207.01780}\n}\n```", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "TIGER-Lab/AceCodeRM", + "base_model_relation": "finetune" + }, + { + "model_id": "qingy2024/UIGEN-T1.1-Qwen-32B", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- smirki/UI_REASONING_v1.01\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- code\n- ui\n- generation\n- uigen\nlibrary_name: transformers\n---\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/VSplF7AM1PJPzeR9FlDhE.png)\n\n# **Model Card for UIGEN-T1.1**\n\nNew and Improved reasoning traces. Better ui generation. Smarter decisions. Better code generation! Trained on a 700+ dataset. \nUSE BUDGET FORCING (putting the word answer or think at the end of the assistant generation to keep generationg more thinking and use 'answer' to write code.) \nSFT on 1 x H100 for 1 hour.\n\n## **Model Summary** \nUIGEN-T1.1-Qwen-32B is a **32-billion parameter transformer model** fine-tuned on **Qwen2.5-Coder-32B-Instruct**. It is designed for **reasoning-based UI generation**, leveraging a complex chain-of-thought approach to produce **robust HTML and CSS-based UI components**. Currently, it is limited to **basic applications such as dashboards, landing pages, and sign-up forms**. \n\n## **Model Details** \n\n### **Model Description** \nUIGEN-T1.1-Qwen-32B generates **HTML and CSS-based UI layouts** by reasoning through design principles. While it has a strong **chain-of-thought reasoning process**, it is currently **limited to text-based UI elements and simpler frontend applications**. The model **excels at dashboards, landing pages, and sign-up forms**, but **lacks advanced interactivity** (e.g., JavaScript-heavy functionalities). \n\n- **Dataset by:** [smirki](https://huggingface.co/smirki)\n- **Developed by:** [smirki](https://huggingface.co/smirki)\n- **Training Procedures and Scripts by:** [smirki](https://huggingface.co/smirki)\n- **Trained by:** [qingy2024](https://huggingface.co/qingy2024)\n- **Model by:** [qingy2024](https://huggingface.co/qingy2024)\n- **Shared by:** [qingy2024](https://huggingface.co/qingy2024)\n- **Model type:** Transformer-based \n- **Language(s) (NLP):** English \n- **License:** Apache 2.0 \n- **Finetuned from model:** Qwen/Qwen2.5-Coder-32B-Instruct \n\n### **Model Sources** \n- **Repository:** (Will be uploaded to GitHub soon) \n- **Hosted on:** [Hugging Face](https://huggingface.co/qingy2024) \n- **Demo:** Coming soon\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/617YL3OlJHflvR63qbA27.png)\n\n\n## **Uses** \n\n### **Direct Use** \n- Generates HTML and CSS code for **basic UI elements** \n- Best suited for **dashboards, landing pages, and sign-up forms** \n- Requires **manual post-processing** to refine UI outputs \n- **May require using the word \"answer\" at the end of the input prompt** to get better inference \n\n### **Downstream Use (optional)** \n- Can be fine-tuned further for **specific frontend frameworks (React, Vue, etc.)** \n- May be integrated into **no-code/low-code UI generation tools** \n\n### **Out-of-Scope Use** \n- Not suitable for **complex frontend applications** involving JavaScript-heavy interactions \n- May not generate **fully production-ready** UI code \n- **Limited design variety** \u2013 biased towards **basic frontend layouts** \n\n## **Bias, Risks, and Limitations** \n\n### **Biases** \n- **Strong bias towards basic frontend design patterns** (may not generate creative or advanced UI layouts) \n- **May produce repetitive designs** due to limited training scope \n\n### **Limitations** \n- **Artifacting issues**: Some outputs may contain formatting artifacts \n- **Limited generalization**: Performs best in **HTML + CSS UI generation**, but **not robust for complex app logic** \n- **May require prompt engineering** (e.g., adding \"answer\" to input for better results) \n\n## **How to Get Started with the Model** \n\n### **Example Model Template** \n```plaintext\n<|im_start|>user\n{question}<|im_end|>\n<|im_start|>assistant\n<|im_start|>think\n{reasoning}<|im_end|>\n<|im_start|>answer\n```\n\n### **Basic Inference Code** \n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"qingy2024/UIGEN-T1.1-Qwen-32B\"\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nmodel = AutoModelForCausalLM.from_pretrained(model_name).to(\"cuda\")\n\nprompt = \"\"\"<|im_start|>user\nMake a dark-themed dashboard for an oil rig.<|im_end|>\n<|im_start|>assistant\n<|im_start|>think\n\"\"\"\n\ninputs = tokenizer(prompt, return_tensors=\"pt\").to(\"cuda\")\noutputs = model.generate(**inputs, max_new_tokens=12012, do_sample=True, temperature=0.7) #max tokens has to be greater than 12k\n\nprint(tokenizer.decode(outputs[0], skip_special_tokens=True))\n```\n\n## **Training Details** \n\n### **Training Data** \n- **Based on:** Qwen2.5-Coder-32B-Instruct \n- **Fine-tuned on:** UI-related datasets with reasoning-based HTML/CSS examples \n\n### **Training Procedure** \n- **Preprocessing:** Standard text-tokenization using Hugging Face transformers \n- **Training Precision:** **bf16 mixed precision** quantized to q8\n- **Training Method:** Full-precision LoRA for 1 epoch, then merged to 16 bit (this model).\n\n## **Evaluation** \n\n### **Testing Data, Factors & Metrics** \n- **Testing Data:** Internal UI design-related datasets \n- **Evaluation Factors:** Bias towards basic UI components, robustness in reasoning, output quality \n- **Metrics:** Subjective evaluation based on UI structure, correctness, and usability \n\n### **Results** \n- **Strengths:** \n - **Good at reasoning-based UI layouts** \n - **Generates structured and valid HTML/CSS** \n- **Weaknesses:** \n - **Limited design diversity** \n - **Artifacting in outputs** \n\n## **Technical Specifications** \n\n### **Model Architecture and Objective** \n- **Architecture:** Transformer-based LLM fine-tuned for UI reasoning \n- **Objective:** Generate **robust frontend UI layouts with chain-of-thought reasoning** \n\n### **Compute Infrastructure** \n- **Hardware Requirements:** > 24GB VRAM recommended\n- **Software Requirements:** \n - Transformers library (Hugging Face) \n - PyTorch \n\n## **Citation** \nIf using this model, please cite: \n\n**BibTeX:** \n```bibtex\n@misc{smirki_UIGEN-T1.1,\n title={UIGEN-T1.1.1: Chain-of-Thought UI Generation Model},\n author={smirki},\n year={2025},\n publisher={Hugging Face},\n url={https://huggingface.co/smirki/UIGEN-T1.11}\n}\n```\n\n## **More Information** \n- **GitHub Repository:** (Coming soon) \n- **Web Demo:** (Coming soon) \n\n## **Model Card Authors** \n- **Author:** smirki \n\n## **Model Card Contact** \n- **Contact:** [smirki](https://huggingface.co/smirki), [qingy2024 on Hugging Face](https://huggingface.co/smirki)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/VLs2LyOPXV4GZ2feXpD7F.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/boAgm__7mbD_B37OZzzyw.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/aBn-uzBsmK7vj-CxXRyBi.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/HaPat6448BpqneaBU47bz.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/8VuOPWMSJlu3kxmUQJb6R.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/RgFpQigIYes7wvzulZnkg.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/alBaCYJSLKyomF55XjOv-.png)\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/k3uu2IPU_wIWco45RwdOV.png)\n\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "smirki/UIGEN-T1.1-Qwen-32B-Q8_0-GGUF", + "smirki/UIGEN-T1.1-Qwen-32B-Q4_K_M-GGUF", + "mradermacher/UIGEN-T1.1-Qwen-32B-GGUF" + ], + "quantized_count": 3, + "merges": [], + "merges_count": 0, + "total_derivatives": 3, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "qingy2024/UIGEN-T1.1-Qwen-32B", + "base_model_relation": "base" + }, + { + "model_id": "Bojun-Feng/Qwen2.5-Coder-32B-Instruct-GGUF-llamafile", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlicense_link: https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct-GGUF/blob/main/LICENSE\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n---\n\n\n\n\n
\n\"LlamaFile\"\n\n\n\n\n\n\nI am not the original creator of llamafile, all credit of llamafile goes to Jartine:\n\n\n
\n

Chat & support: jartine's Discord server

\n
\n

jartine's LLM work is generously supported by a grant from mozilla

\n
\n\n\n# Qwen2.5 Coder 32B Instruct GGUF - llamafile\n\n## Run LLMs locally with a single file - No installation required!\n\nAll you need is download a file and run it.\n\nOur goal is to make open source large language models much more\naccessible to both developers and end users. We're doing that by\ncombining [llama.cpp](https://github.com/ggerganov/llama.cpp) with [Cosmopolitan Libc](https://github.com/jart/cosmopolitan) into one\nframework that collapses all the complexity of LLMs down to\na single-file executable (called a \"llamafile\") that runs\nlocally on most computers, with no installation.\n\n## How to Use (Modified from [Git README](https://github.com/Mozilla-Ocho/llamafile/tree/8f73d39cf3a767897b8ade6dda45e5744c62356a?tab=readme-ov-file#quickstart))\n\nThe easiest way to try it for yourself is to download our example llamafile.\nWith llamafile, all inference happens locally; no data ever leaves your computer.\n\n1. Download the llamafile.\n\n2. Open your computer's terminal.\n\n3. If you're using macOS, Linux, or BSD, you'll need to grant permission\nfor your computer to execute this new file. (You only need to do this\nonce.)\n\n```sh\nchmod +x qwen2.5-coder-32b-instruct-q8_0.gguf\n```\n\n4. If you're on Windows, rename the file by adding \".exe\" on the end.\n\n5. Run the llamafile. e.g.:\n\n```sh\n./qwen2.5-coder-32b-instruct-q8_0.gguf\n```\n\n6. Your browser should open automatically and display a chat interface.\n(If it doesn't, just open your browser and point it at http://localhost:8080.)\n\n7. When you're done chatting, return to your terminal and hit\n`Control-C` to shut down llamafile.\n\nNote: Hugging Face has a 50GB file upload Limit, so you may need to use the `cat` instruction to concatenate large llamafiles to run them.\n\nHere is an example doing so to `Mozilla/Meta-Llama-3.1-405B-Instruct-llamafile`:\n```\nwget https://huggingface.co/Mozilla/Meta-Llama-3.1-405B-llamafile/resolve/main/Meta-Llama-3.1-405B.Q2_K.cat0.llamafile\nwget https://huggingface.co/Mozilla/Meta-Llama-3.1-405B-llamafile/resolve/main/Meta-Llama-3.1-405B.Q2_K.cat1.llamafile\nwget https://huggingface.co/Mozilla/Meta-Llama-3.1-405B-llamafile/resolve/main/Meta-Llama-3.1-405B.Q2_K.cat2.llamafile\nwget https://huggingface.co/Mozilla/Meta-Llama-3.1-405B-llamafile/resolve/main/Meta-Llama-3.1-405B.Q2_K.cat3.llamafile\ncat Meta-Llama-3.1-405B.Q2_K.cat{0,1,2,3}.llamafile >Meta-Llama-3.1-405B.Q2_K.llamafile\nrm Meta-Llama-3.1-405B.Q2_K.cat*.llamafile\nchmod +x Meta-Llama-3.1-405B.Q2_K.llamafile\n./Meta-Llama-3.1-405B.Q2_K.llamafile\n```\n\n\nPlease note that LlamaFile is still under active development. Some methods may be not be compatible with the most recent documents.\n\n## Settings for Qwen2.5 Coder 32B Instruct GGUF Llamafiles\n\n- Model creator: [Qwen](https://huggingface.co/Qwen)\n- Quantized GGUF files used: [Qwen/Qwen2.5-Coder-32B-Instruct-GGUF](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct-GGUF/tree/9d3053fce650fe1cdbdb75998c2a87add9d178ef)\n - Commit message \"Update README.md\"\n - Commit hash 9d3053fce650fe1cdbdb75998c2a87add9d178ef\n- LlamaFile version used: [Mozilla-Ocho/llamafile](https://github.com/Mozilla-Ocho/llamafile/tree/29b5f27172306da39a9c70fe25173da1b1564f82)\n - Commit message \"Merge pull request #687 from Xydane/main Add Support for DeepSeek-R1 models\"\n - Commit hash 29b5f27172306da39a9c70fe25173da1b1564f82\n- `.args` content format (example):\n\n```\n-m\nqwen2.5-coder-32b-instruct-q8_0.gguf\n...\n```\n\n## (Following is original model card for Qwen2.5 Coder 32B Instruct GGUF)\n
\n\n\n\n# Qwen2.5-Coder-32B-Instruct-GGUF\n\n \"Chat\"\n\n\n## Introduction\n\nQwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers. Qwen2.5-Coder brings the following improvements upon CodeQwen1.5:\n\n- Significantly improvements in **code generation**, **code reasoning** and **code fixing**. Base on the strong Qwen2.5, we scale up the training tokens into 5.5 trillion including source code, text-code grounding, Synthetic data, etc. Qwen2.5-Coder-32B has become the current state-of-the-art open-source codeLLM, with its coding abilities matching those of GPT-4o.\n- A more comprehensive foundation for real-world applications such as **Code Agents**. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.\n- **Long-context Support** up to 128K tokens.\n\n**This repo contains the instruction-tuned 32B Qwen2.5-Coder model in the GGUF Format**, which has the following features:\n- Type: Causal Language Models\n- Training Stage: Pretraining & Post-training\n- Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias\n- Number of Parameters: 32.5B\n- Number of Paramaters (Non-Embedding): 31.0B\n- Number of Layers: 64\n- Number of Attention Heads (GQA): 40 for Q and 8 for KV\n- Context Length: Full 32,768 tokens\n - Note: Currently, only vLLM supports YARN for length extrapolating. If you want to process sequences up to 131,072 tokens, please refer to non-GGUF models.\n- Quantization: q2_K, q3_K_M, q4_0, q4_K_M, q5_0, q5_K_M, q6_K, q8_0\n\nFor more details, please refer to our [blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/), [GitHub](https://github.com/QwenLM/Qwen2.5-Coder), [Documentation](https://qwen.readthedocs.io/en/latest/), [Arxiv](https://arxiv.org/abs/2409.12186).\n\n## Quickstart\n\nCheck out our [llama.cpp documentation](https://qwen.readthedocs.io/en/latest/run_locally/llama.cpp.html) for more usage guide.\n\nWe advise you to clone [`llama.cpp`](https://github.com/ggerganov/llama.cpp) and install it following the official guide. We follow the latest version of llama.cpp. \nIn the following demonstration, we assume that you are running commands under the repository `llama.cpp`.\n\nSince cloning the entire repo may be inefficient, you can manually download the GGUF file that you need or use `huggingface-cli`:\n1. Install\n ```shell\n pip install -U huggingface_hub\n ```\n2. Download:\n ```shell\n huggingface-cli download Qwen/Qwen2.5-Coder-32B-Instruct-GGUF --include \"qwen2.5-coder-32b-instruct-q5_k_m*.gguf\" --local-dir . --local-dir-use-symlinks False\n ```\n For large files, we split them into multiple segments due to the limitation of file upload. They share a prefix, with a suffix indicating its index. For examples, `qwen2.5-coder-32b-instruct-q5_k_m-00001-of-00003.gguf`, `qwen2.5-coder-32b-instruct-q5_k_m-00002-of-00003.gguf` and `qwen2.5-coder-32b-instruct-q5_k_m-00003-of-00003.gguf`. The above command will download all of them.\n3. (Optional) Merge:\n For split files, you need to merge them first with the command `llama-gguf-split` as shown below:\n ```bash\n # ./llama-gguf-split --merge \n ./llama-gguf-split --merge qwen2.5-coder-32b-instruct-q5_k_m-00001-of-00003.gguf qwen2.5-coder-32b-instruct-q5_k_m.gguf\n ```\n\nFor users, to achieve chatbot-like experience, it is recommended to commence in the conversation mode:\n\n```shell\n./llama-cli -m \\\n -co -cnv -p \"You are Qwen, created by Alibaba Cloud. You are a helpful assistant.\" \\\n -fa -ngl 80 -n 512\n```\n\n\n## Evaluation & Performance\n\nDetailed evaluation results are reported in this [\ud83d\udcd1 blog](https://qwenlm.github.io/blog/qwen2.5-coder-family/).\n\nFor requirements on GPU memory and the respective throughput, see results [here](https://qwen.readthedocs.io/en/latest/benchmark/speed_benchmark.html).\n\n## Citation\n\nIf you find our work helpful, feel free to give us a cite.\n\n```\n@article{hui2024qwen2,\n title={Qwen2. 5-Coder Technical Report},\n author={Hui, Binyuan and Yang, Jian and Cui, Zeyu and Yang, Jiaxi and Liu, Dayiheng and Zhang, Lei and Liu, Tianyu and Zhang, Jiajun and Yu, Bowen and Dang, Kai and others},\n journal={arXiv preprint arXiv:2409.12186},\n year={2024}\n}\n@article{qwen2,\n title={Qwen2 Technical Report}, \n author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},\n journal={arXiv preprint arXiv:2407.10671},\n year={2024}\n}\n```\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Bojun-Feng/Qwen2.5-Coder-32B-Instruct-GGUF-llamafile", + "base_model_relation": "base" + }, + { + "model_id": "SuanChang/rain-SQLCoder", + "gated": "False", + "card": "---\nlicense: apache-2.0\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\n---\n# Introduction\n[Rain's SQLCoder](https://huggingface.co/SuanChang/rain-SQLCoder) is a state-of-the-art large language model (LLM) designed for natural language-to-SparkSQL generation. Rain's SQLCoder, with 32B parameters, is fine-tuned from the [Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct). Optimized for natural language-to-SparkSQL conversion tasks, Rain's SQLCoder effectively handles contexts of up to 32k tokens, making it particularly suitable for generating complex and large-scale SQL queries.\n\n

\n \ud83e\udd17 Hugging Face | \ud83d\udda5\ufe0f Demo | \ud83d\udcac WeChat | GitHub \n

\n\n[English](./README.md) | [\u4e2d\u6587](./README-zh.md)\n\n# Prompt\nRain's SQLCoder adopted the [Alpaca](https://github.com/tatsu-lab/stanford_alpaca) template, with the following prompt format.\n````\nBelow is an instruction that describes a task. \nWrite a response that appropriately completes the request.\n\n### Instruction:\n[BEGIN OF TASK INSTRUCTION]\nYou are an expert in composing Spark SQL queries. You are given a user query and a set of table schemas.\nBased on the user query, you need to generate one Spark SQL query to achieve the purpose.\n{task description for date hint and related question and sqls}\n[END OF TASK INSTRUCTION]\n\n[BEGIN OF TABLE SCHEMAS]\n{schemas}\n[END OF TABLE SCHEMAS]\n\n[BEGIN OF GENERATION HINT]\n{date hint}\n[END OF GENERATION HINT]\n\n[BEGIN OF RELATED QUERIES]\n{related question and sqls}\n[END OF RELATED QUERIES]\n\n[BEGIN OF FORMAT INSTRUCTION]\nThe output MUST strictly adhere to the following format, and NO other text MUST be included.\n```sql\nyour output Spark SQL query\n``` \n[END OF FORMAT INSTRUCTION]\n\n[BEGIN OF QUERY]\nUser Query: {user question}\n[END OF QUERY]\n\n### Response:\n````\n\n# Evaluation\nWe followed the evaluation logic from [SQL-Eval](https://github.com/defog-ai/sql-eval) to compare predicted results with ground truth:\n1. If the predicted data frame exactly matches the ground truth data frame, the prediction is considered correct.\n2. If the ground truth SQL does not contain sorting logic, and the predicted data frame matches the ground truth data frame after sorting, the prediction is considered correct.\n3. If the columns in the ground truth data frame are a subset of the predicted data frame, the prediction is considered correct.\n4. In all other cases, the prediction is considered incorrect.\n\n# Experimental Results\nWe compared the generation accuracy of Rain's SQLCoder with state-of-the-art natural language large models, both domestic and international, on two test datasets. The Benchmark Dataset contains baseline samples, while the Enhanced Dataset is constructed by applying stratified sampling to 20% of the Benchmark Dataset and supplementing it with relevant user questions and corresponding SparkSQL statements to evaluate the model's performance under enhanced contextual information. The experimental results demonstrate that Rain's SQLCoder exhibits significant advantages in query intent understanding, SQL syntax accuracy, and complex query processing.\n\n## Benchmark Dataset\n\"benchmark\"\n\n## Enhanced Dataset\n\"enhanced\"\n\n# Quick Start\nWe provide examples here to help you quickly learn how to load and use our model.\n>Tips: Rain's SQLCoder is trained solely for generating `SELECT` statements, and when the table schemas cannot support answering the user's question, the model will refuse to respond.\n\n````python\nimport torch\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nfrom utils.prompt import SQLGeneratePrompt\n\nmodel_name = \"SuanChang/rain-SQLCoder\"\n\nmodel = AutoModelForCausalLM.from_pretrained(\n model_name,\n torch_dtype=torch.bfloat16,\n device_map=\"auto\",\n)\ntokenizer = AutoTokenizer.from_pretrained(model_name)\n\nquestion = \"What is the name of the department that offers a course that has a description including the word 'Statistics'?\"\nschemas = [\n'''CREATE TABLE `course` (\n `crs_code` STRING,\n `dept_code` STRING,\n `crs_description` STRING,\n `crs_credit` DOUBLE\n);''',\n'''CREATE TABLE `department` (\n `dept_code` STRING,\n `dept_name` STRING,\n `school_code` STRING,\n `emp_num` INT,\n `dept_address` STRING,\n `dept_extension` INT\n);''',\n'''CREATE TABLE `student` (\n `stu_num` INT,\n `stu_lname` STRING,\n `stu_fname` STRING,\n `stu_init` STRING,\n `stu_dob` STRING,\n `stu_hrs` INT,\n `stu_class` STRING,\n `stu_gpa` DOUBLE,\n `stu_transfer` INT,\n `dept_code` STRING,\n `stu_phone` INT,\n `prof_num` INT\n);'''\n]\nhint = \"- Today is 2025-02-01.\"\ndata = dict(\n question=question,\n schema=\"\\n\\n\".join(schemas),\n hint=hint,\n related_question_sqls=None,\n)\ntext, _, _ = SQLGeneratePrompt.prompt(data)\n\nmodel_inputs = tokenizer([text], return_tensors=\"pt\").to(model.device)\n\ngenerated_ids = model.generate(\n **model_inputs,\n max_new_tokens=32768\n)\ngenerated_ids = [\n output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)\n]\nresponse = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]\n\nprint(response)\n\n'''\n```sql\nSELECT d.dept_name FROM department d JOIN course c ON d.dept_code = c.dept_code WHERE c.crs_description LIKE '%Statistics%';\n```\n'''\n````", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/rain-SQLCoder-GGUF", + "mradermacher/rain-SQLCoder-i1-GGUF" + ], + "quantized_count": 2, + "merges": [], + "merges_count": 0, + "total_derivatives": 2, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "SuanChang/rain-SQLCoder", + "base_model_relation": "base" + }, + { + "model_id": "secmlr/SWE-BENCH-400-reasoning_qwen_code_32B_test_swe", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-400-reasoning_qwen_code_32B_test_swe\n results: []\n---\n\n\n\n# SWE-BENCH-400-reasoning_qwen_code_32B_test_swe\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-400-reasoning dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 2\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 24\n- total_eval_batch_size: 16\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.1.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-400-reasoning_qwen_code_32B_test_swe", + "base_model_relation": "base" + }, + { + "model_id": "secmlr/SWE-BENCH-400-reasoning-short-llm_qwen_code_32B_test_swe", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-400-reasoning-short-llm_qwen_code_32B_test_swe\n results: []\n---\n\n\n\n# SWE-BENCH-400-reasoning-short-llm_qwen_code_32B_test_swe\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-400-reasoning-short-llm dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 2\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 24\n- total_eval_batch_size: 16\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.1.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-400-reasoning-short-llm_qwen_code_32B_test_swe", + "base_model_relation": "base" + }, + { + "model_id": "open-r1/OlympicCoder-32B", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- open-r1/codeforces-cots\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\nlibrary_name: transformers\n---\n\n# Model Card for OlympicCoder-32B\n\nOlympicCoder-32B is a code model that achieves very strong performance on competitive coding benchmarks such as LiveCodeBench andthe 2024 International Olympiad in Informatics.\n\n* Repository: https://github.com/huggingface/open-r1\n* Blog post: https://huggingface.co/blog/open-r1/update-3\n\n## Model description\n\n- **Model type:** A 32B parameter model fine-tuned on a decontaminated version of the codeforces dataset.\n- **Language(s) (NLP):** Primarily English\n- **License:** apache-2.0\n- **Finetuned from model:** [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)\n\n## Evaluation\n\nWe compare the performance of OlympicCoder models on two main benchmarks for competitive coding:\n\n* **[IOI'2024:](https://github.com/huggingface/ioi)** 6 very challenging problems from the 2024 International Olympiad in Informatics. Models are allowed up to 50 submissions per problem.\n* **[LiveCodeBench:](https://livecodebench.github.io)** Python programming problems source from platforms like CodeForces and LeetCoder. We use the `v4_v5` subset of [`livecodebench/code_generation_lite`](https://huggingface.co/datasets/livecodebench/code_generation_lite), which corresponds to 268 problems. We use `lighteval` to evaluate models on LiveCodeBench using the sampling parameters described [here](https://github.com/huggingface/open-r1?tab=readme-ov-file#livecodebench).\n\n> [!NOTE]\n> The OlympicCoder models were post-trained exclusively on C++ solutions generated by DeepSeek-R1. As a result the performance on LiveCodeBench should be considered to be partially _out-of-domain_, since this expects models to output solutions in Python.\n\n### IOI'24\n\n![](./ioi-evals.png)\n\n### LiveCodeBench\n\n![](./lcb-evals.png)\n\n\n\n## Usage\nHere's how you can run the model using the `pipeline()` function from \ud83e\udd17 Transformers:\n\n```python\n# pip install transformers\n# pip install accelerate\n\nimport torch\nfrom transformers import pipeline\n\npipe = pipeline(\"text-generation\", model=\"open-r1/OlympicCoder-32B\", torch_dtype=torch.bfloat16, device_map=\"auto\")\n\n# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating\nmessages = [\n {\"role\": \"user\", \"content\": \"Write a python program to calculate the 10th Fibonacci number\"},\n]\nprompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)\noutputs = pipe(prompt, max_new_tokens=8000, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)\nprint(outputs[0][\"generated_text\"])\n#<|im_start|>user\n#Write a python program to calculate the 10th fibonacci number<|im_end|>\n#<|im_start|>assistant\n#Okay, I need to write a Python program that calculates the 10th Fibonacci number. Hmm, the Fibonacci sequence starts with 0 and 1. Each subsequent number is the sum of the two preceding ones. So the sequence goes: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, and so on. ...\n```\n\n> [!IMPORTANT]\n> To ensure that the model consistently outputs a long chain-of-thought, we have edited the chat template to prefill the first assistant turn with a `` token. As a result, the outputs from this model will not show the opening `` token if you use the model's `generate()` method. To apply reinforcement learning with a format reward, either prepend the `` token to the model's completions or amend the chat template to remove the prefill. Check out our [blog post](https://huggingface.co/blog/open-r1/update-3#lesson-4-prefill-with-think-to-consistently-enable-long-cot) for more details.\n\n\n## Training procedure\n### Training hyper-parameters\n\nThe following hyperparameters were used during training on 16 H100 nodes:\n\n- dataset: open-r1/codeforces-cots_decontaminated\n- learning_rate: 4.0e-5\n- train_batch_size: 1\n- seed: 42\n- packing: false\n- distributed_type: fsdp\n- num_devices: 128\n- gradient_accumulation_steps: 1\n- total_train_batch_size: 16\n- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: cosine_with_min_lr\n- min_lr_rate: 0.1\n- lr_scheduler_warmup_ratio: 0.03\n- num_epochs: 10.0", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "alexgusevski/OlympicCoder-32B-mlx-fp16" + ], + "children_count": 1, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "bartowski/open-r1_OlympicCoder-32B-GGUF", + "lmstudio-community/OlympicCoder-32B-GGUF", + "DevQuasar/open-r1.OlympicCoder-32B-GGUF", + "mradermacher/OlympicCoder-32B-GGUF", + "alexgusevski/OlympicCoder-32B-mlx-3Bit", + "alexgusevski/OlympicCoder-32B-mlx-4Bit", + "alexgusevski/OlympicCoder-32B-mlx-6Bit", + "alexgusevski/OlympicCoder-32B-mlx-8Bit", + "stelterlab/OlympicCoder-32B-AWQ", + "alexgusevski/OlympicCoder-32B-mlx-2Bit", + "mlx-community/OlympicCoder-32B-4bit", + "BenevolenceMessiah/OlympicCoder-32B-Q8_0-GGUF" + ], + "quantized_count": 12, + "merges": [], + "merges_count": 0, + "total_derivatives": 13, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "open-r1/OlympicCoder", + "base_model_relation": "finetune" + }, + { + "model_id": "karths/coder_python_32B", + "gated": "False", + "card": "---\nlicense: apache-2.0\ndatasets:\n- karths/python_all_phi4_maintain\nlanguage:\n- en\nmetrics:\n- code_eval\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\n---", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "karths/coder_python_32B-Q4_K_M-GGUF" + ], + "quantized_count": 1, + "merges": [], + "merges_count": 0, + "total_derivatives": 1, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "karths/coder_python_32B", + "base_model_relation": "base" + }, + { + "model_id": "Tesslate/Tessa-T1-32B", + "gated": "False", + "card": "---\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- text-generation-inference\n- transformers\n- qwen2\n- trl\nlicense: apache-2.0\nlanguage:\n- en\ndatasets:\n- Tesslate/Tessa-T1-Dataset\n---\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/I7XzH-NMKUshcGU86u6VA.png)\n\n\"Landing Page\"\n\n## **Model Overview**\n\nTessa-T1 is an innovative transformer-based **React reasoning model**, fine-tuned from the powerful **Qwen2.5-Coder-32B-Instruct** base model. Designed specifically for React frontend development, Tessa-T1 leverages advanced reasoning to autonomously generate well-structured, semantic React components. Its integration into agent systems makes it a powerful tool for automating web interface development and frontend code intelligence.\n\n---\n\n## **Model Highlights**\n\n- **React-specific Reasoning**: Accurately generates functional and semantic React components.\n- **Agent Integration**: Seamlessly fits into AI-driven coding agents and autonomous frontend systems.\n- **Context-Aware Generation**: Effectively understands and utilizes UI context to provide relevant code solutions.\n\n---\n\n## **Example Outputs**\n\n*See examples demonstrating the powerful reasoning and component creation capabilities of Tessa-T1:*\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/4nRXURnPgg4aPu8JTaopy.png)\nAI upload\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/7LytoCkbXJvhpaFhA4VwY.png)\nVirtual Machine Console\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/enutXwjAmfVN4PXg19zME.png)\n\nPlaylist Management\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/zU3yln3xGdUIywGRtxSij.png)\n\nPrompt: \"add in a calendar\"\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/eDpCj-eV3DmDjkDdB1Ux1.png)\n\n---\n\n## **Use Cases**\n\n### **Recommended Uses**\n- **Automatic Component Generation**: Quickly produce React components from textual prompts.\n- **Agent-based Web Development**: Integrate into automated coding systems for faster frontend workflows.\n- **Frontend Refactoring**: Automate the optimization and semantic enhancement of React code.\n\n### **Limitations**\n- **Focused on React**: Limited use outside React.js frameworks.\n- **Complex State Management**: May require manual adjustments for highly dynamic state management scenarios.\n\n---\n\n## **How to Use**\n\n### **Inference Example**\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nmodel_name = \"smirki/Tessa-T1\"\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nmodel = AutoModelForCausalLM.from_pretrained(model_name).to(\"cuda\")\nprompt = \"\"\"<|im_start|>user\nCreate a React component for a user profile card.<|im_end|>\n<|im_start|>assistant\n<|im_start|>think\n\"\"\"\ninputs = tokenizer(prompt, return_tensors=\"pt\").to(\"cuda\")\noutputs = model.generate(**inputs, max_new_tokens=1500, do_sample=True, temperature=0.7)\nprint(tokenizer.decode(outputs[0], skip_special_tokens=True))\n```\n\n---\n\n## **Performance and Evaluation**\n\n- **Strengths**:\n - Strong semantic React component generation.\n - Excellent integration capabilities with agent-based systems.\n\n- **Weaknesses**:\n - Complex JavaScript logic may require manual post-processing.\n\n---\n\n## **Technical Specifications**\n\n- **Architecture**: Transformer-based LLM\n- **Base Model**: Qwen2.5-Coder-32B-Instruct\n- **Precision**: bf16 mixed precision, quantized to q8\n- **Hardware Requirements**: Recommended 12GB VRAM\n- **Software Dependencies**:\n - Hugging Face Transformers\n - PyTorch\n\n---\n\n## **Citation**\n\n```bibtex\n@misc{smirki_Tessa-T1,\n title={Tessa-T1: React-Focused Reasoning Model for Component Generation},\n author={tesslate},\n year={2025},\n publisher={Hugging Face},\n url={https://huggingface.co/tesslate/Tessa-T1}\n}\n```\n\n---\n\n## **Contact & Community**\n- **Creator:** [smirki](https://huggingface.co/tesslate)\n- **Repository & Demo**: Coming soon!\n\n**Sponsored by vichar ai [Huggingface](https://huggingface.co/vicharai) [Website](https://vichar.io)**", + "metadata": "\"N/A\"", + "depth": 1, + "children": [ + "mlx-community/Tessa-T1-32B-mlx-bf16" + ], + "children_count": 1, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/Tessa-T1-32B-GGUF", + "second-state/Tessa-T1-32B-GGUF", + "gaianet/Tessa-T1-32B-GGUF", + "mlx-community/Tessa-T1-32B-mlx-8bit", + "mradermacher/Tessa-T1-32B-i1-GGUF", + "mlx-community/Tessa-T1-32B-mlx-4bit", + "DevQuasar/Tesslate.Tessa-T1-32B-GGUF", + "bartowski/Tesslate_Tessa-T1-32B-GGUF" + ], + "quantized_count": 8, + "merges": [], + "merges_count": 0, + "total_derivatives": 9, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Tesslate/Tessa-T1", + "base_model_relation": "finetune" + }, + { + "model_id": "Tesslate/UIGEN-T1.5-32B", + "gated": "False", + "card": "---\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- text-generation-inference\n- transformers\n- qwen2\n- trl\nlicense: apache-2.0\nlanguage:\n- en\ndatasets:\n- Tesslate/UIGEN-T1.5-Dataset\n---\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/bCWPiMgNitTrrONN3SVpw.png)\n*Landing page showcasing visual richness*\n\n# **Model Card for UIGEN-T1.5**\n\n---\n\n## **Model Overview**\n\nUIGEN-T1.5 is an advanced transformer-based UI generation model fine-tuned from **Qwen2.5-Coder-32B-Instruct**, specifically enhanced to produce stunning, modern, and unique frontend user interfaces. Leveraging sophisticated reasoning and chain-of-thought methodologies, UIGEN-T1.5 excels at generating highly structured and visually compelling HTML and CSS code, ideal for sleek dashboards, engaging landing pages, and intuitive sign-up forms.\n\n---\n\n## **Model Highlights**\n\n- **Advanced UI Styles**: Produces sleek, modern, and unique designs.\n- **Chain-of-Thought Reasoning**: Enhanced reasoning capabilities for accurate HTML/CSS layouts.\n- **High Usability**: Generates responsive and production-ready frontend code.\n\n---\n\n## **Visual Examples**\n\n*See examples below showcasing UIGEN-T1.5-generated interfaces:*\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/T4nz0JBAvYdwPVpZ_sb3g.png)\n*Dashboard UI generated by UIGEN-T1.5*\n\n\n---\n\n## **Use Cases**\n\n### **Recommended Uses**\n- **Dashboards**: Insightful and visually appealing data interfaces.\n- **Landing Pages**: Captivating and high-conversion web pages.\n- **Authentication Screens**: Elegant sign-up and login interfaces.\n\n### **Limitations**\n- **Limited Interactivity**: Minimal JavaScript functionality, focusing on HTML/CSS.\n- **Prompt Engineering**: May require specific prompts (e.g., appending \"answer\").\n\n---\n\n## **How to Use**\n\n### **Inference Example**\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"smirki/UIGEN-T1.5\"\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nmodel = AutoModelForCausalLM.from_pretrained(model_name).to(\"cuda\")\n\nprompt = \"\"\"<|im_start|>user\nDesign a sleek, modern dashboard for monitoring solar panel efficiency.<|im_end|>\n<|im_start|>assistant\n<|im_start|>think\n\"\"\"\n\ninputs = tokenizer(prompt, return_tensors=\"pt\").to(\"cuda\")\noutputs = model.generate(**inputs, max_new_tokens=12012, do_sample=True, temperature=0.7)\n\nprint(tokenizer.decode(outputs[0], skip_special_tokens=True))\n```\n\n---\n\n## **Performance and Evaluation**\n\n- **Strengths**:\n - High-quality UI generation.\n - Strong reasoning capabilities for structured layouts.\n\n- **Weaknesses**:\n - Occasional repetitive design patterns.\n - Minor artifacting in complex designs.\n\n---\n\n## **Technical Specifications**\n\n- **Architecture**: Transformer-based LLM\n- **Base Model**: Qwen2.5-Coder-7B-Instruct\n- **Precision**: bf16 mixed precision, quantized to q8\n- **Hardware Requirements**: Recommended 12GB VRAM\n- **Software Dependencies**:\n - Hugging Face Transformers\n - PyTorch\n\n---\n\n## **Citation**\n\n```bibtex\n@misc{Tesslate_UIGEN-T1.5,\n title={UIGEN-T1.5: Advanced Chain-of-Thought UI Generation Model},\n author={smirki},\n year={2025},\n publisher={Hugging Face},\n url={https://huggingface.co/Tesslate/UIGEN-T1.5}\n}\n```\n\n---\n\n## **Contact & Community**\n- **Creator:** [smirki](https://huggingface.co/Tesslate)\n- **Repository & Demo**: Coming soon!\n\n**Sponsored by vichar ai [Huggingface](https://huggingface.co/vicharai) [Website](https://vichar.io)**", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "smcleod/UIGEN-T1.5-32B-Q5_K_S-GGUF", + "mradermacher/UIGEN-T1.5-32B-GGUF", + "mradermacher/UIGEN-T1.5-32B-i1-GGUF", + "DevQuasar/Tesslate.UIGEN-T1.5-32B-GGUF" + ], + "quantized_count": 4, + "merges": [], + "merges_count": 0, + "total_derivatives": 4, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Tesslate/UIGEN-T1.5-32B", + "base_model_relation": "base" + }, + { + "model_id": "Tesslate/UIGEN-T1.5-32B-4bit", + "gated": "False", + "card": "---\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- text-generation-inference\n- transformers\n- qwen2\n- trl\nlicense: apache-2.0\nlanguage:\n- en\ndatasets:\n- Tesslate/UIGEN-T1.5-Dataset\n---\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/bCWPiMgNitTrrONN3SVpw.png)\n*Landing page showcasing visual richness*\n\n# **Model Card for UIGEN-T1.5**\n\n---\n\n## **Model Overview**\n\nUIGEN-T1.5 is an advanced transformer-based UI generation model fine-tuned from **Qwen2.5-Coder-32B-Instruct**, specifically enhanced to produce stunning, modern, and unique frontend user interfaces. Leveraging sophisticated reasoning and chain-of-thought methodologies, UIGEN-T1.5 excels at generating highly structured and visually compelling HTML and CSS code, ideal for sleek dashboards, engaging landing pages, and intuitive sign-up forms.\n\n---\n\n## **Model Highlights**\n\n- **Advanced UI Styles**: Produces sleek, modern, and unique designs.\n- **Chain-of-Thought Reasoning**: Enhanced reasoning capabilities for accurate HTML/CSS layouts.\n- **High Usability**: Generates responsive and production-ready frontend code.\n\n---\n\n## **Visual Examples**\n\n*See examples below showcasing UIGEN-T1.5-generated interfaces:*\n\n\n![image/png](https://cdn-uploads.huggingface.co/production/uploads/64d1129297ca59bcf7458d07/T4nz0JBAvYdwPVpZ_sb3g.png)\n*Dashboard UI generated by UIGEN-T1.5*\n\n\n---\n\n## **Use Cases**\n\n### **Recommended Uses**\n- **Dashboards**: Insightful and visually appealing data interfaces.\n- **Landing Pages**: Captivating and high-conversion web pages.\n- **Authentication Screens**: Elegant sign-up and login interfaces.\n\n### **Limitations**\n- **Limited Interactivity**: Minimal JavaScript functionality, focusing on HTML/CSS.\n- **Prompt Engineering**: May require specific prompts (e.g., appending \"answer\").\n\n---\n\n## **How to Use**\n\n### **Inference Example**\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\nmodel_name = \"smirki/UIGEN-T1.5\"\ntokenizer = AutoTokenizer.from_pretrained(model_name)\nmodel = AutoModelForCausalLM.from_pretrained(model_name).to(\"cuda\")\n\nprompt = \"\"\"<|im_start|>user\nDesign a sleek, modern dashboard for monitoring solar panel efficiency.<|im_end|>\n<|im_start|>assistant\n<|im_start|>think\n\"\"\"\n\ninputs = tokenizer(prompt, return_tensors=\"pt\").to(\"cuda\")\noutputs = model.generate(**inputs, max_new_tokens=12012, do_sample=True, temperature=0.7)\n\nprint(tokenizer.decode(outputs[0], skip_special_tokens=True))\n```\n\n---\n\n## **Performance and Evaluation**\n\n- **Strengths**:\n - High-quality UI generation.\n - Strong reasoning capabilities for structured layouts.\n\n- **Weaknesses**:\n - Occasional repetitive design patterns.\n - Minor artifacting in complex designs.\n\n---\n\n## **Technical Specifications**\n\n- **Architecture**: Transformer-based LLM\n- **Base Model**: Qwen2.5-Coder-32B-Instruct\n- **Precision**: bf16 mixed precision, quantized to q8\n- **Hardware Requirements**: Recommended 12GB VRAM\n- **Software Dependencies**:\n - Hugging Face Transformers\n - PyTorch\n\n---\n\n## **Citation**\n\n```bibtex\n@misc{Tesslate_UIGEN-T1.5,\n title={UIGEN-T1.5: Advanced Chain-of-Thought UI Generation Model},\n author={smirki},\n year={2025},\n publisher={Hugging Face},\n url={https://huggingface.co/Tesslate/UIGEN-T1.5}\n}\n```\n\n---\n\n## **Contact & Community**\n- **Creator:** [smirki](https://huggingface.co/Tesslate)\n- **Repository & Demo**: Coming soon!", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "Tesslate/UIGEN-T1.5-32B-4bit", + "base_model_relation": "base" + }, + { + "model_id": "all-hands/openhands-lm-32b-v0.1-ep3", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- SWE-Gym/SWE-Gym\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\ntags:\n- agent\n- coding\n---\n\n
\n \"Logo\"\n

OpenHands LM v0.1

\n
\n\n

\nBlog\n\u2022\nUse it in OpenHands\n

\n\n\n\"WARNING: This model (trained for three epochs instead of one) is released solely for research purposes - it may have significant issues of [looping and not following instructions](https://x.com/xingyaow_/status/1906812305473757258).\nPlease use [all-hands/openhands-lm-32b-v0.1](https://huggingface.co/all-hands/openhands-lm-32b-v0.1) for everyday OpenHands tasks.\"\n---\n\nAutonomous agents for software development are already contributing to a [wide range of software development tasks](/blog/8-use-cases-for-generalist-software-development-agents).\nBut up to this point, strong coding agents have relied on proprietary models, which means that even if you use an open-source agent like [OpenHands](https://github.com/All-Hands-AI/OpenHands), you are still reliant on API calls to an external service.\n\nToday, we are excited to introduce OpenHands LM, a new open coding model that:\n\n- Is open and [available on Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1), so you can download it and run it locally\n- Is a reasonable size, 32B, so it can be run locally on hardware such as a single 3090 GPU\n- Achieves strong performance on software engineering tasks, including 37.2% resolve rate on SWE-Bench Verified\n\nRead below for more details and our future plans!\n\n## What is OpenHands LM?\n\nOpenHands LM is built on the foundation of [Qwen Coder 2.5 Instruct 32B](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct), leveraging its powerful base capabilities for coding tasks. What sets OpenHands LM apart is our specialized fine-tuning process:\n\n- We used training data generated by OpenHands itself on a diverse set of open-source repositories\n- Specifically, we use an RL-based framework outlined in [SWE-Gym](https://arxiv.org/abs/2412.21139), where we set up a training environment, generate training data using an existing agent, and then fine-tune the model on examples that were resolved successfully\n- It features a 128K token context window, ideal for handling large codebases and long-horizon software engineering tasks\n\n\n## Performance: Punching Above Its Weight\n\nWe evaluated OpenHands LM using our latest [iterative evaluation protocol](https://github.com/All-Hands-AI/OpenHands/tree/main/evaluation/benchmarks/swe_bench#run-inference-rollout-on-swe-bench-instances-generate-patch-from-problem-statement) on the [SWE-Bench Verified benchmark](https://www.swebench.com/#verified).\n\nThe results are impressive:\n\n- **37.2% verified resolve rate** on SWE-Bench Verified\n- Performance comparable to models with **20x more parameters**, including Deepseek V3 0324 (38.8%) with 671B parameters\n\nHere's how OpenHands LM compares to other leading open-source models:\n\n![OpenHands LM Performance Comparison](https://www.all-hands.dev/assets/blog/20250331-openhands-lm-release/performance_scatter.png)\n\nAs the plot demonstrates, our 32B parameter model achieves efficiency that approaches much larger models. While the largest models (671B parameters) achieve slightly higher scores, our 32B parameter model performs remarkably well, opening up possibilities for local deployment that are not possible with larger models.\n\n## Getting Started: How to Use OpenHands LM Today\n\nYou can start using OpenHands LM immediately through these channels:\n\n1. **Download the model from Hugging Face**\n The model is available on [Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1) and can be downloaded directly from there.\n\n2. **Create an OpenAI-compatible endpoint with a model serving framework**\n For optimal performance, it is recommended to serve this model with a GPU using [SGLang](https://github.com/sgl-project/sglang) or [vLLM](https://github.com/vllm-project/vllm).\n\n3. **Point your OpenHands agent to the new model**\n Download [OpenHands](https://github.com/All-Hands-AI/OpenHands) and follow the instructions for [using an OpenAI-compatible endpoint](https://docs.all-hands.dev/modules/usage/llms/openai-llms#using-openai-compatible-endpoints).\n\n\n## The Road Ahead: Our Development Plans\n\nThis initial release marks just the beginning of our journey. We will continue enhancing OpenHands LM based on community feedback and ongoing research initiatives.\n\nIn particular, it should be noted that the model is still a research preview, and (1) may be best suited for tasks regarding solving github issues and perform less well on more varied software engineering tasks, (2) may sometimes generate repetitive steps, and (3) is somewhat sensitive to quantization, and may not function at full performance at lower quantization levels.\nOur next releases will focus on addressing these limitations.\n\nWe're also developing more compact versions of the model (including a 7B parameter variant) to support users with limited computational resources. These smaller models will preserve OpenHands LM's core strengths while dramatically reducing hardware requirements.\n\nWe encourage you to experiment with OpenHands LM, share your experiences, and participate in its evolution. Together, we can create better tools for tomorrow's software development landscape.\n\n## Join Our Community\n\nWe invite you to be part of the OpenHands LM journey:\n\n- Explore our [GitHub repository](https://github.com/All-Hands-AI/OpenHands)\n- Connect with us on [Slack](https://join.slack.com/t/openhands-ai/shared_invite/zt-2tom0er4l-JeNUGHt_AxpEfIBstbLPiw)\n- Follow our [documentation](https://docs.all-hands.dev) to get started\n\nBy contributing your experiences and feedback, you'll help shape the future of this open-source initiative. Together, we can create better tools for tomorrow's software development landscape.\n\nWe can't wait to see what you'll create with OpenHands LM!", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "DevQuasar/all-hands.openhands-lm-32b-v0.1-ep3-GGUF" + ], + "quantized_count": 1, + "merges": [], + "merges_count": 0, + "total_derivatives": 1, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "all-hands/openhands-lm-32b-v0.1-ep3", + "base_model_relation": "base" + }, + { + "model_id": "lucyknada/all-hands_openhands-lm-32b-v0.1-exl2", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- SWE-Gym/SWE-Gym\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: text-generation\ntags:\n- agent\n- coding\n---\n### exl2 quant (measurement.json in main branch)\n---\n### check revisions for quants\n---\n\n\n
\n \"Logo\"\n

OpenHands LM v0.1

\n
\n\n

\nBlog\n\u2022\nUse it in OpenHands\n

\n\n---\n\nAutonomous agents for software development are already contributing to a [wide range of software development tasks](/blog/8-use-cases-for-generalist-software-development-agents).\nBut up to this point, strong coding agents have relied on proprietary models, which means that even if you use an open-source agent like [OpenHands](https://github.com/All-Hands-AI/OpenHands), you are still reliant on API calls to an external service.\n\nToday, we are excited to introduce OpenHands LM, a new open coding model that:\n\n- Is open and [available on Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1), so you can download it and run it locally\n- Is a reasonable size, 32B, so it can be run locally on hardware such as a single 3090 GPU\n- Achieves strong performance on software engineering tasks, including 37.2% resolve rate on SWE-Bench Verified\n\nRead below for more details and our future plans!\n\n## What is OpenHands LM?\n\nOpenHands LM is built on the foundation of [Qwen Coder 2.5 Instruct 32B](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct), leveraging its powerful base capabilities for coding tasks. What sets OpenHands LM apart is our specialized fine-tuning process:\n\n- We used training data generated by OpenHands itself on a diverse set of open-source repositories\n- Specifically, we use an RL-based framework outlined in [SWE-Gym](https://arxiv.org/abs/2412.21139), where we set up a training environment, generate training data using an existing agent, and then fine-tune the model on examples that were resolved successfully\n- It features a 128K token context window, ideal for handling large codebases and long-horizon software engineering tasks\n\n\n## Performance: Punching Above Its Weight\n\nWe evaluated OpenHands LM using our latest [iterative evaluation protocol](https://github.com/All-Hands-AI/OpenHands/tree/main/evaluation/benchmarks/swe_bench#run-inference-rollout-on-swe-bench-instances-generate-patch-from-problem-statement) on the [SWE-Bench Verified benchmark](https://www.swebench.com/#verified).\n\nThe results are impressive:\n\n- **37.2% verified resolve rate** on SWE-Bench Verified\n- Performance comparable to models with **20x more parameters**, including Deepseek V3 0324 (38.8%) with 671B parameters\n\nHere's how OpenHands LM compares to other leading open-source models:\n\n![OpenHands LM Performance Comparison](https://www.all-hands.dev/assets/blog/20250331-openhands-lm-release/performance_scatter.png)\n\nAs the plot demonstrates, our 32B parameter model achieves efficiency that approaches much larger models. While the largest models (671B parameters) achieve slightly higher scores, our 32B parameter model performs remarkably well, opening up possibilities for local deployment that are not possible with larger models.\n\n## Getting Started: How to Use OpenHands LM Today\n\nYou can start using OpenHands LM immediately through these channels:\n\n1. **Download the model from Hugging Face**\n The model is available on [Hugging Face](https://huggingface.co/all-hands/openhands-lm-32b-v0.1) and can be downloaded directly from there.\n\n2. **Create an OpenAI-compatible endpoint with a model serving framework**\n For optimal performance, it is recommended to serve this model with a GPU using [SGLang](https://github.com/sgl-project/sglang) or [vLLM](https://github.com/vllm-project/vllm).\n\n3. **Point your OpenHands agent to the new model**\n Download [OpenHands](https://github.com/All-Hands-AI/OpenHands) and follow the instructions for [using an OpenAI-compatible endpoint](https://docs.all-hands.dev/modules/usage/llms/openai-llms#using-openai-compatible-endpoints).\n\n\n## The Road Ahead: Our Development Plans\n\nThis initial release marks just the beginning of our journey. We will continue enhancing OpenHands LM based on community feedback and ongoing research initiatives.\n\nIn particular, it should be noted that the model is still a research preview, and (1) may be best suited for tasks regarding solving github issues and perform less well on more varied software engineering tasks, (2) may sometimes generate repetitive steps, and (3) is somewhat sensitive to quantization, and may not function at full performance at lower quantization levels.\nOur next releases will focus on addressing these limitations.\n\nWe're also developing more compact versions of the model (including a 7B parameter variant) to support users with limited computational resources. These smaller models will preserve OpenHands LM's core strengths while dramatically reducing hardware requirements.\n\nWe encourage you to experiment with OpenHands LM, share your experiences, and participate in its evolution. Together, we can create better tools for tomorrow's software development landscape.\n\n## Join Our Community\n\nWe invite you to be part of the OpenHands LM journey:\n\n- Explore our [GitHub repository](https://github.com/All-Hands-AI/OpenHands)\n- Connect with us on [Slack](https://join.slack.com/t/openhands-ai/shared_invite/zt-2tom0er4l-JeNUGHt_AxpEfIBstbLPiw)\n- Follow our [documentation](https://docs.all-hands.dev) to get started\n\nBy contributing your experiences and feedback, you'll help shape the future of this open-source initiative. Together, we can create better tools for tomorrow's software development landscape.\n\nWe can't wait to see what you'll create with OpenHands LM!", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "lucyknada/all-hands_openhands-lm-32b-v0.1-exl2", + "base_model_relation": "base" + }, + { + "model_id": "secmlr/SWE-BENCH-500-set-4o-file-localization_qwen_code_32B_test_swe", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-500-set-4o-file-localization_qwen_code_32B_test_swe\n results: []\n---\n\n\n\n# SWE-BENCH-500-set-4o-file-localization_qwen_code_32B_test_swe\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-500-set-4o-file-localization dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 2\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 24\n- total_eval_batch_size: 16\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.2.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-500-set-4o-file-localization_qwen_code_32B_test_swe", + "base_model_relation": "base" + }, + { + "model_id": "secmlr/SWE-BENCH-433-set-claude-file-localization-with-reasoning_qwen_code_32B_test_swe", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-433-set-claude-file-localization-with-reasoning_qwen_code_32B_test_swe\n results: []\n---\n\n\n\n# SWE-BENCH-433-set-claude-file-localization-with-reasoning_qwen_code_32B_test_swe\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-433-set-claude-file-localization-with-reasoning dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 2\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 24\n- total_eval_batch_size: 16\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.2.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-433-set-claude-file-localization-with-reasoning_qwen_code_32B_test_swe", + "base_model_relation": "base" + }, + { + "model_id": "secmlr/SWE-BENCH-500-set-claude-3in1-localization-with-reasoning_3in1-32b-500", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-500-set-claude-3in1-localization-with-reasoning_3in1-32b-500\n results: []\n---\n\n\n\n# SWE-BENCH-500-set-claude-3in1-localization-with-reasoning_3in1-32b-500\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-500-set-claude-3in1-localization-with-reasoning dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 4\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 48\n- total_eval_batch_size: 32\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.2.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-500-set-claude-3in1-localization-with-reasoning_3in1-32b", + "base_model_relation": "finetune" + }, + { + "model_id": "observerw/ChiseLLM-32B", + "gated": "False", + "card": "---\nlicense: mit\ndatasets:\n- observerw/ChiseLLM-Completion\n- observerw/ChiseLLM-Decompile\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\n---\n\n# ChiseLLM Models\n\n\"ChiseLLM\"\n\n[GitHub](https://github.com/observerw/ChiseLLM)\n\nChiseLLM is a series of **large reasoning models specifically trained for the [Chisel Hardware Construction language](https://www.chisel-lang.org)**, aimed at revolutionizing HCL-Baed Agile Hardware Development Methodology (AHDM).\n\nBuilt on [Qwen/Qwen2.5-Coder-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) with domain-adaptive fine-tuning, the model combines high-quality reasoning datasets and specific thinking patterns to significantly enhance performance in hardware design tasks.\n\nChiseLLM Models can:\n\n- **Transform natural language specifications into high-quality Chisel code** (Spec-to-Chisel)\n- **Intelligently translate Verilog code into enhanced Chisel implementations** (Decompile-to-Chisel)\n- **Generate hardware designs with superior variability and extensibility**, surpassing traditional design approaches\n\n### Use Cases\n\nChiseLLM Models is particularly suited for the following applications:\n\n- **Rapid Hardware Design Prototyping**: Dramatically shortens the design cycle from specification to implementation\n- **Verilog Code Modernization**: Intelligently converts legacy Verilog code into extensible Chisel implementations\n- **Hardware Architecture Exploration**: Generates multiple design variants for the same functional requirements\n- **Design Refactoring and Optimization**: Leverages Chisel's advanced features to improve existing hardware designs\n- **Agile Hardware Development Education**: Serves as an assistive tool for learning Chisel and modern hardware design methods\n\n### Training results\n\nSpec-to-Chisel task on VerilogEval:\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------------------------ | --------- | --------- | --------- | --------- |\n| Llama3.1-8B-Instruct | 4.33 | 9.90 | 13.21 | 9.02 |\n| Qwen2.5-Coder-7B-Instruct | 21.94 | 31.87 | 36.73 | 37.08 |\n| \\*Deepseek-R1-Distill-Llama-8B | 9.31 | 15.44 | 17.72 | 16.01 |\n| \\*ChiseLLM-7B | **29.41** | **47.08** | **54.04** | **58.82** |\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------------------------- | --------- | --------- | --------- | --------- |\n| Qwen2.5-Coder-32B-Instruct | 41.02 | 53.85 | 58.79 | 73.47 |\n| Qwen2.5-72B-Instruct | 39.74 | 49.30 | 52.90 | 61.31 |\n| Llama-3.3-70B-Instruct | 38.14 | 44.90 | 48.02 | 65.97 |\n| \\*Deepseek-R1-Distill-Qwen-32B | 38.50 | 54.58 | 61.16 | 52.19 |\n| \\*Deepseek-R1-Distill-Llama-70B | 36.62 | 52.28 | 59.90 | 51.72 |\n| \\*ChiseLLM-32B | **51.43** | **68.29** | **72.78** | **76.45** |\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------- | --------- | --------- | --------- | --------- |\n| Deepseek-V3 | 50.16 | 63.44 | 67.32 | 76.37 |\n| GPT-4o | 42.04 | 60.16 | 65.17 | 69.76 |\n| \\*Deepseek-R1 | **62.74** | **76.05** | **80.16** | **82.85** |\n\nDecompile-to-Chisel task on VerilogEval:\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------------------------ | --------- | --------- | --------- | --------- |\n| Llama3.1-8B-Instruct | 5.43 | 12.29 | 16.08 | 11.15 |\n| Qwen2.5-Coder-7B-Instruct | 27.60 | 34.58 | 37.19 | 43.23 |\n| \\*Deepseek-R1-Distill-Llama-8B | 10.05 | 16.15 | 18.13 | 12.03 |\n| \\* ChiseLLM-7B | **50.47** | **70.99** | **78.08** | **59.19** |\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------------------------- | --------- | --------- | --------- | --------- |\n| Qwen2.5-Coder-32B-Instruct | 41.19 | 48.96 | 51.59 | 53.93 |\n| Qwen2.5-72B-Instruct | 40.54 | 47.32 | 49.83 | 59.30 |\n| Llama-3.3-70B-Instruct | 38.38 | 46.96 | 51.36 | 48.00 |\n| \\*Deepseek-R1-Distill-Qwen-32B | 45.03 | 63.02 | 70.18 | 53.17 |\n| \\*Deepseek-R1-Distill-Llama-70B | 37.50 | 55.05 | 63.84 | 45.59 |\n| \\*ChiseLLM-32B | **56.41** | **72.00** | **77.67** | **64.71** |\n\n| Models | pass@1 | pass@3 | pass@5 | syntax(%) |\n| ------------- | --------- | --------- | --------- | --------- |\n| Deepseek-V3 | **54.57** | 63.19 | 66.71 | **66.19** |\n| GPT-4o | 42.39 | 65.75 | 71.83 | 53.77 |\n| \\*Deepseek-R1 | 53.45 | **71.50** | **77.91** | 59.13 |\n\n### Framework versions\n\n- Transformers 4.51.0\n- Pytorch 2.6.0a0+df5bbc09d1.nv24.12\n- Datasets 3.4.1\n- Tokenizers 0.21.0\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n\n- learning_rate: 1e-05\n- train_batch_size: 2\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 8\n- gradient_accumulation_steps: 8\n- total_train_batch_size: 128\n- total_eval_batch_size: 64\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.05\n- num_epochs: 3.0\n\n## Citation\n\nIf you are interested in our work, please consider citing this, it would be greatly appreciated!\n\n```bibtex\n@misc{wang2025chisellmunleashingpowerreasoning,\n title={ChiseLLM: Unleashing the Power of Reasoning LLMs for Chisel Agile Hardware Development}, \n author={Bowei Wang and Jiaran Gao and Yelai Feng and Renzhi Chen and Shanshan Li and Lei Wang},\n year={2025},\n eprint={2504.19144},\n archivePrefix={arXiv},\n primaryClass={cs.AI},\n url={https://arxiv.org/abs/2504.19144}, \n}\n```", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [ + "mradermacher/ChiseLLM-32B-GGUF", + "mradermacher/ChiseLLM-32B-i1-GGUF" + ], + "quantized_count": 2, + "merges": [], + "merges_count": 0, + "total_derivatives": 2, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "observerw/ChiseLLM", + "base_model_relation": "finetune" + }, + { + "model_id": "all-hands/openhands-critic-32b-exp-20250417", + "gated": "False", + "card": "---\nlicense: mit\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\npipeline_tag: token-classification\ntags:\n- agent\n- coding\n---\n\n
\n \"Logo\"\n

OpenHands Critic Model

\n
\n\n

\nBlog\n

\n\n**Research Purpose Only: This model is released strictly for research and is not yet compatible with the OpenHands application. For complete information about this model, including its capabilities and limitations, please refer to our [detailed blog post](https://www.all-hands.dev/blog/sota-on-swe-bench-verified-with-inference-time-scaling-and-critic-model).**\n\n---\n\n# SOTA on SWE-Bench Verified with Inference-Time Scaling and Critic Model\n\nWe're thrilled to announce that OpenHands has reached a new milestone, achieving state-of-the-art results on SWE-Bench Verified!\n\n![OpenHands Performance on Leaderboard](https://www.all-hands.dev/assets/blog/20250416-inference-time-scaling/leaderboard.png)\n\n## SWE-Bench and OpenHands\n\n[SWE-bench](https://www.swebench.com/) is the most popular benchmark for evaluating large language models' (LLMs) capabilities in addressing real-world software engineering challenges. It consists of issues and corresponding pull requests from 12 popular Python repositories on GitHub, tasking systems with generating code patches to resolve specified issues. The *verified* subset we evaluated on consists of 500 carefully selected test cases that have been manually reviewed by [human software developers](https://openai.com/index/introducing-swe-bench-verified/) to verify they have appropriately scoped unit tests and well-specified issue descriptions.\nDue to its realism and the potential vast benefits of AI agents that could autonomously solve real-world software development challenges, it is used widely throughout academia and industry as a gold-standard for measuring the abilities of AI coding agents.\n\nWe're developing the [OpenHands](https://github.com/All-Hands-AI/OpenHands) open-source software development agent, and its performance on this dataset is currently at 60.6% - not too shabby!\nBut we wondered, what happens if we really push the limits?\n\n## Inference-Time Scaling: More Compute, Better Results\n\nOur approach leverages a simple but powerful idea: for challenging software engineering tasks, trying multiple solutions and picking the best one can lead to better outcomes. Here's how it works:\n\n1. For each SWE-Bench problem, we run OpenHands agent multiple times using claude 3.7 sonnet with sampling temperature 1.0, generate multiple solutions that leads to multiple code patches\n2. We trained a \"critic model\" that evaluates each solution and predicts whether it's a good solution or not (more details about this model below)\n3. We filter out code patches that fails [regression and reproduction tests](https://github.com/OpenAutoCoder/Agentless/blob/main/README_swebench.md#-patch-validation-and-selection)\n4. We select the solution from the trajectory with the highest score as our final answer\n\nThis method of inference-time scaling lets us achieve substantially better results without modifying the underlying agent model and scaffold.\nWe observe log-linear performance improvement from 60.6% on a single trajectory rollout to 66.4% with five attempts, which will make [our submission](https://github.com/SWE-bench/experiments/pull/209) number one on the leaderboard!\n\n![OpenHands Inference-Time Scaling Result](https://www.all-hands.dev/assets/blog/20250416-inference-time-scaling/inference-time-scaling-fig.png)\n\n\n## Building a Better Critic\n\nThis idea of using choosing the best of multiple solutions has been tried by other SWE-bench submissions, but these strategies were generally based on prompting an existing model like Claude.\nRather than using this prompt-based reranking strategy, we trained a dedicated critic model, which we found provided more effective results.\n\nFor the training process, we:\n- Roll out agent trajectories from [SWE-Gym](https://github.com/SWE-Gym/SWE-Gym) to avoid data leakage\n- Implement a temporal difference (TD) learning objective to propagate trajectory-level success signals from unit test execution backward through each trajectory\n- Add a regression head on top of the last layer to predict reward values\n\nThe TD learning objective is particularly powerful because it helps the model understand which actions contributed to the final outcome:\n\n$$\nr_t = \\gamma r_{t+1}\n$$\n\nWhere $r_t$ is the reward at time step $t$ (i.e., the t-th action produced by agent), $\\gamma$ is the discount factor. The process starts with the final reward $r_T$ which is determined by running the unit tests on the completed solution - 1 for passing all tests and 0 for failing. This terminal reward is then propagated backwards through the trajectory, with each previous step discounted by $\\gamma$. We use $\\gamma=0.99$.\n\nWe use [veRL](https://github.com/volcengine/verl) to finetune [Qwen 2.5 Coder Instruct 32B](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) as a critic model. During inference, we use a [modified version of vLLM](https://github.com/xingyaoww/vllm/tree/add-token-classification-support) to serve this model for evaluation (fun fact: OpenHands agent itself wrote most of the functional [code](https://github.com/vllm-project/vllm/compare/main...xingyaoww:vllm:add-token-classification-support) there).\n\nWe're making the critic model [publicly available on huggingface](https://huggingface.co/all-hands/openhands-critic-32b-exp-20250417) for researchers who want to explore its capabilities or build upon our work.\n\n## Why We Built a Critic Model and Where It's Going\n\nWe chose to invest in a trained critic model for several reasons:\n\n**Genuine usefulness through generalization**: While prompt-engineering-based reranker can help boost benchmark scores, real-world generalization is not easy to guarantee. We believe with sufficient data, a trained critic model could generalize to diverse software engineering scenarios beyond SWE-Bench. This makes it a valuable tool for solving real-world problems in everyday coding tasks.\n\n**Use intermediate reward for future improvements**: While our current implementation focuses on selecting the best complete solution from multiple trajectories, the intermediate rewards predicted throughout each trajectory opens up exciting possibilities for enhancing our agent's capabilities.\n- *One-step lookahead sampling* allows us to evaluate multiple potential actions at each step, using the critic's scores to choose the most promising path forward (experimental [PR](https://github.com/All-Hands-AI/OpenHands/pull/7770)).\n- *Real-time mistake recovery* is another frontier we're exploring, where the critic can identify declining rewards and help the agent course-correct during the solution process ([issue](https://github.com/All-Hands-AI/OpenHands/issues/2221)).\n\nWe're actively working on integrating these signals more deeply into the OpenHands agent experience, which could enable more efficient assistance even in scenarios where generating multiple complete solutions isn't practical.\n\n\n## Try OpenHands Today\n\nBesides being state-of-the-art on SWE-Bench Verified, OpenHands is also a top-performing agent on [LiveSWEBench](https://liveswebench.github.io), a contamination-free benchmark for AI software engineers. Additionally, OpenHands ranks first on [Multi-SWE-Bench](https://multi-swe-bench.github.io), a variant of SWE-Bench that evaluates across 8 different programming languages.\n\nOverall, we feel confident in saying that OpenHands is the best agent out there for a wide variety of tasks!\nIf you'd like to try it out today you can:\n\n- **Start with OpenHands Cloud**: The easiest way to get started is with our fully managed [cloud solution](https://app.all-hands.dev) with $50 free credits, seamless GitHub integration, mobile support, and optimizations like [context condensation](https://www.all-hands.dev/blog/openhands-context-condensensation-for-more-efficient-ai-agents) ready to use.\n\n- **Contribute to Open Source**: Star, open issues, or send PRs to our [GitHub repository](https://github.com/All-Hands-AI/OpenHands) and help advance the frontier of open-source AI software development.\n\n- **Join Our Community**: Connect with us on [Slack](https://join.slack.com/t/openhands-ai/shared_invite/zt-2ngejmfw6-9gW4APWOC9XUp1n~SiQ6iw), read our [documentation](https://docs.all-hands.dev), and stay updated on our latest developments.\n\nWe can't wait to see what you'll build with OpenHands!\n\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "all-hands/openhands-critic-32b-exp", + "base_model_relation": "finetune" + }, + { + "model_id": "secmlr/SWE-BENCH-433-set-claude-3in1-localization-with-reasoning_3in1-32b-433", + "gated": "False", + "card": "---\nlibrary_name: transformers\nlicense: apache-2.0\nbase_model: Qwen/Qwen2.5-Coder-32B-Instruct\ntags:\n- llama-factory\n- full\n- generated_from_trainer\nmodel-index:\n- name: SWE-BENCH-433-set-claude-3in1-localization-with-reasoning_3in1-32b-433\n results: []\n---\n\n\n\n# SWE-BENCH-433-set-claude-3in1-localization-with-reasoning_3in1-32b-433\n\nThis model is a fine-tuned version of [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) on the SWE-BENCH-433-set-claude-3in1-localization-with-reasoning dataset.\n\n## Model description\n\nMore information needed\n\n## Intended uses & limitations\n\nMore information needed\n\n## Training and evaluation data\n\nMore information needed\n\n## Training procedure\n\n### Training hyperparameters\n\nThe following hyperparameters were used during training:\n- learning_rate: 1e-05\n- train_batch_size: 1\n- eval_batch_size: 8\n- seed: 42\n- distributed_type: multi-GPU\n- num_devices: 4\n- gradient_accumulation_steps: 12\n- total_train_batch_size: 48\n- total_eval_batch_size: 32\n- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments\n- lr_scheduler_type: cosine\n- lr_scheduler_warmup_ratio: 0.1\n- num_epochs: 3.0\n\n### Training results\n\n\n\n### Framework versions\n\n- Transformers 4.48.3\n- Pytorch 2.5.1+cu124\n- Datasets 3.2.0\n- Tokenizers 0.21.0\n", + "metadata": "\"N/A\"", + "depth": 1, + "children": [], + "children_count": 0, + "adapters": [], + "adapters_count": 0, + "quantized": [], + "quantized_count": 0, + "merges": [], + "merges_count": 0, + "total_derivatives": 0, + "spaces": [], + "spaces_count": 0, + "parents": [ + "Qwen/Qwen2.5-Coder-32B-Instruct" + ], + "base_model": "secmlr/SWE-BENCH-433-set-claude-3in1-localization-with-reasoning_3in1-32b", + "base_model_relation": "finetune" + }, + { + "model_id": "vicharai/ViCoder-html-32B-preview", + "gated": "False", + "card": "---\nlicense: apache-2.0\nlanguage:\n- en\nbase_model:\n- Qwen/Qwen2.5-Coder-32B-Instruct\nlibrary_name: transformers\ntags:\n- code\n- codeqwen\n- chat\n- qwen\n- qwen-coder\n- html\n- javascript\n- css\n- tailwindcss\n- frontend\n- web-development\n- ViCoder-html-32B-preview\n- ViCoder-html\n- ViCoder\n- ViCoder-html-preview\n- vichar ai labs\n- vichar ai\n- strive ai labs llp\n- strive ai labs\n- vichar.io\npipeline_tag: text-generation\n---\n\n

\n \n \"ViCoder-html-32B-preview\n \n

\n ViCoder-html-32B-preview\n

\n

\n

\n \ud83d\ude80 A powerful HTML/CSS/JS sketching model powered by Qwen2.5-Coder-32B-Instruct \ud83d\ude80 \n

\n

\n Developed by Vichar AI | Hugging Face Profile
\n Licensed under Apache 2.0\n

\n\n---\n\n### \ud83d\udca1 What is ViCoder-html-32B-preview?\n\n**ViCoder-html-32B-preview** is a preview model in the **ViCoder** series from Vichar AI \u2014 a line of models specialized in **code generation**. This model focuses specifically on sketching single-page websites, such as landing pages and dashboards, using using:\n\n- \ud83e\udde0 **HTML** for semantic structure\n- \ud83c\udfa8 **Tailwind CSS** for modern, utility-first styling\n- \u2699\ufe0f **JavaScript** for interactivity and basic dynamic behavior\n\nThis model is ideal for:\n- **Web Developers:** Quickly scaffolding dashboards or page layouts.\n- **Frontend Engineers:** Prototyping UIs and exploring design variations.\n- **Designers:** Turning textual mockups into initial code sketches.\n- **Educators & Students:** Learning and experimenting with HTML, Tailwind CSS, and JavaScript in a practical context.\n\n> \u26a0\ufe0f **Note:** This is a **preview** version. It demonstrates core capabilities but is still under active development. A more refined and robust production release is planned. Stay updated via vichar.io or follow VicharAI on Hugging Face!\n\n---\n\n### \ud83d\udee0\ufe0f Model Details\n\n| Property | Value |\n| :--------------- | :------------------------------------------------------------------------------------------ |\n| **Model Type** | Code Generation (Instruction-tuned Language Model) |\n| **Base Model** | [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct) |\n| **Developed by** | [Vichar AI](https://vichar.io) ([HF Profile](https://huggingface.co/VicharAI)) |\n| **Languages** | Primarily HTML, Tailwind CSS, JavaScript. Understands English instructions. |\n| **Training Data**| Proprietary curated dataset focusing on high-quality web components and pages. |\n| **License** | [Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0) |\n| **Library** | \ud83e\udd17 Transformers |\n| **Contact** | Visit [vichar.io](https://vichar.io) or use HF Discussions |\n\n---\n\n### \ud83e\uddf1 GGUF Quantized Versions\n\nQuantized versions of **ViCoder-html-32B-preview** in GGUF format are available for efficient local inference using [llama.cpp](https://github.com/ggerganov/llama.cpp), [LM Studio](https://lmstudio.ai/), or [Ollama](https://ollama.com/).\n\nYou can find them here:\n- \ud83d\udd17 [GGUF Quantizations on Hugging Face](https://huggingface.co/VicharAI/ViCoder-html-32B-preview-GGUF)\n\nThese quantized variants (Q3_K_M, Q4_K_M, Q6_K, Q8_0) are useful for running the model on lower-memory hardware or for embedding in desktop/web applications.\n\n---\n\n### \u26a1 Example Usage\n\nUse the `transformers` library pipeline for easy text generation. Ensure you have `transformers`, `torch`, and `accelerate` installed.\n\n```python\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, pipeline\nimport torch\n\n# Define model ID\nmodel_id = \"VicharAI/ViCoder-html-32B-preview\"\n\n# Load tokenizer and model\n# Use bfloat16 for faster inference if your GPU supports it\ntokenizer = AutoTokenizer.from_pretrained(model_id)\nmodel = AutoModelForCausalLM.from_pretrained(\n model_id,\n torch_dtype=torch.bfloat16, # Or torch.float16 if bfloat16 is not supported\n device_map=\"auto\" # Automatically distribute across available GPUs/CPU\n)\n\nmessages = [\n {\"role\": \"user\", \"content\": \"A modern, sleek landing page for a company focusing on open-source LLM solutions\"},\n]\ninputs = tokenizer.apply_chat_template(\n messages,\n tokenize = True,\n add_generation_prompt = True,\n return_tensors = \"pt\",\n).to(\"cuda\")\n\nfrom transformers import TextStreamer\ntext_streamer = TextStreamer(tokenizer, skip_prompt = True)\n_ = model.generate(input_ids = inputs, streamer = text_streamer, max_new_tokens = 16000,\n use_cache = True, temperature = 0.7, min_p = 0.1, repetition_penalty=1.1)\n```\n\n---\n\n### \u2728 Output Sample\n\n\n```html\n\n\n\n \n Our Love Story - Surprise Website\n \n \n \n