File size: 5,361 Bytes
76bf1c6 1888e1e fdbb37d 1888e1e d455069 1888e1e acc1e1c d455069 1888e1e d455069 1888e1e d455069 1888e1e d455069 1888e1e d455069 1888e1e d455069 1888e1e 76bf1c6 1888e1e 76bf1c6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 |
---
license: apache-2.0
datasets:
- NeelNanda/pile-10k
---
## Model Details
This model is an int4 model with group_size 128 of [google/gemma-2b](https://huggingface.co/google/gemma-2b) generated by [intel/auto-round](https://github.com/intel/auto-round).
### Use the model
### INT4 Inference with ITREX on CPU
Install the latest [intel-extension-for-transformers](
https://github.com/intel/intel-extension-for-transformers)
```python
from intel_extension_for_transformers.transformers import AutoModelForCausalLM
from transformers import AutoTokenizer
quantized_model_dir = "Intel/gemma-2b-int4-inc"
model = AutoModelForCausalLM.from_pretrained(quantized_model_dir,
device_map="auto",
trust_remote_code=False,
use_neural_speed=False,
)
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir, use_fast=True)
print(tokenizer.decode(model.generate(**tokenizer("There is a girl who likes adventure,", return_tensors="pt").to(model.device),max_new_tokens=50)[0]))
"""
<bos>There is a girl who likes adventure, and she is a girl who likes to travel. She is a girl who likes to explore the world and see new things. She is a girl who likes to meet new people and learn about their cultures. She is a girl who likes to take risks
"""
```
### INT4 Inference with AutoGPTQ's kernel
```python
##pip install auto-gptq
from transformers import AutoModelForCausalLM, AutoTokenizer
quantized_model_dir = "Intel/gemma-2b-int4-inc"
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir)
model = AutoModelForCausalLM.from_pretrained(quantized_model_dir,
device_map="auto",
trust_remote_code=False,
)
tokenizer = AutoTokenizer.from_pretrained(quantized_model_dir, use_fast=True)
text = "There is a girl who likes adventure,"
inputs = tokenizer(text, return_tensors="pt").to(model.device)
print(tokenizer.decode(model.generate(**inputs, max_new_tokens=50)[0]))
##<bos>There is a girl who likes adventure, and she is a girl who likes to travel. She is a girl who likes to explore the world and see new things. She is a girl who likes to meet new people and learn about their cultures. She is a girl who likes to take risks
```
### Evaluate the model
pip3 install lm-eval==0.4.2
pip install auto-gptq
Please note that there is a discrepancy between the baseline result and the official data, which is a known issue within the official model card community.
```bash
lm_eval --model hf --model_args pretrained="Intel/gemma-2b-int4-inc",autogptq=True,gptq_use_triton=True --device cuda:0 --tasks lambada_openai,hellaswag,piqa,winogrande,truthfulqa_mc1,openbookqa,boolq,arc_easy,arc_challenge,mmlu --batch_size 16
```
| Metric | BF16 | FP16 | AutoRound v0.1 | AutoRound v0.2 |
| -------------- | ------ | ------ | -------------- | -------------- |
| Avg. | 0.5263 | 0.5277 | 0.5235 | 0.5248 |
| mmlu | 0.3287 | 0.3287 | 0.3297 | 0.3309 |
| lambada_openai | 0.6344 | 0.6375 | 0.6307 | 0.6379 |
| hellaswag | 0.5273 | 0.5281 | 0.5159 | 0.5184 |
| winogrande | 0.6504 | 0.6488 | 0.6543 | 0.6575 |
| piqa | 0.7671 | 0.7720 | 0.7612 | 0.7606 |
| truthfulqa_mc1 | 0.2203 | 0.2203 | 0.2203 | 0.2191 |
| openbookqa | 0.2980 | 0.3020 | 0.3000 | 0.3060 |
| boolq | 0.6927 | 0.6936 | 0.6939 | 0.6966 |
| arc_easy | 0.7420 | 0.7403 | 0.7353 | 0.7357 |
| arc_challenge | 0.4019 | 0.4061 | 0.3933 | 0.3857 |
Here is the sample command to reproduce the model
```bash
git clone https://github.com/intel/auto-round
cd auto-round/examples/language-modeling
pip install -r requirements.txt
python3 main.py \
--model_name google/gemma-2b \
--device 0 \
--group_size 128 \
--bits 4 \
--iters 400 \
--use_quant_input \
--model_dtype "float16"
--deployment_device 'gpu' \
--disable_low_gpu_mem_usage \
--output_dir "./tmp_autoround"
```
## Ethical Considerations and Limitations
The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.
Therefore, before deploying any applications of the model, developers should perform safety testing.
## Caveats and Recommendations
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.
Here are a couple of useful links to learn more about Intel's AI software:
* Intel Neural Compressor [link](https://github.com/intel/neural-compressor)
* Intel Extension for Transformers [link](https://github.com/intel/intel-extension-for-transformers)
## Disclaimer
The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes. |