|
--- |
|
tags: |
|
- gptq |
|
- 4bit |
|
- int4 |
|
- gptqmodel |
|
- modelcloud |
|
- llama-3.1 |
|
- 8b |
|
- instruct |
|
license: llama3.1 |
|
--- |
|
This model has been quantized using [GPTQModel](https://github.com/ModelCloud/GPTQModel). |
|
|
|
- **bits**: 4 |
|
- **group_size**: 128 |
|
- **desc_act**: true |
|
- **static_groups**: false |
|
- **sym**: true |
|
- **lm_head**: false |
|
- **damp_percent**: 0.005 |
|
- **true_sequential**: true |
|
- **model_name_or_path**: "" |
|
- **model_file_base_name**: "model" |
|
- **quant_method**: "gptq" |
|
- **checkpoint_format**: "gptq" |
|
- **meta**： |
|
- **quantizer**: "gptqmodel:0.9.9-dev0" |
|
|
|
## Example: |
|
```python |
|
from transformers import AutoTokenizer |
|
from gptqmodel import GPTQModel |
|
|
|
model_name = "ModelCloud/Meta-Llama-3.1-8B-Instruct-gptq-4bit" |
|
|
|
prompt = [{"role": "user", "content": "I am in Shanghai, preparing to visit the natural history museum. Can you tell me the best way to"}] |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
|
|
model = GPTQModel.from_quantized(model_name) |
|
|
|
input_tensor = tokenizer.apply_chat_template(prompt, add_generation_prompt=True, return_tensors="pt") |
|
outputs = model.generate(input_ids=input_tensor.to(model.device), max_new_tokens=100) |
|
result = tokenizer.decode(outputs[0][input_tensor.shape[1]:], skip_special_tokens=True) |
|
|
|
print(result) |
|
``` |
|
|
|
## lm-eval benchmark |
|
``` |
|
| Tasks |Version|Filter|n-shot| Metric | |Value | |Stderr| |
|
|---------------------------------------|------:|------|-----:|----------|---|-----:|---|-----:| |
|
|arc_challenge | 1|none | 0|acc |↑ |0.5171|± |0.0146| |
|
| | |none | 0|acc_norm |↑ |0.5290|± |0.0146| |
|
|arc_easy | 1|none | 0|acc |↑ |0.8068|± |0.0081| |
|
| | |none | 0|acc_norm |↑ |0.7837|± |0.0084| |
|
|boolq | 2|none | 0|acc |↑ |0.8232|± |0.0067| |
|
|hellaswag | 1|none | 0|acc |↑ |0.5787|± |0.0049| |
|
| | |none | 0|acc_norm |↑ |0.7765|± |0.0042| |
|
|lambada_openai | 1|none | 0|acc |↑ |0.7091|± |0.0063| |
|
| | |none | 0|perplexity|↓ |3.6297|± |0.0805| |
|
|mmlu | 1|none | |acc |↑ |0.6421|± |0.0039| |
|
| - humanities | 1|none | |acc |↑ |0.5932|± |0.0069| |
|
| - formal_logic | 0|none | 0|acc |↑ |0.4206|± |0.0442| |
|
| - high_school_european_history | 0|none | 0|acc |↑ |0.7030|± |0.0357| |
|
| - high_school_us_history | 0|none | 0|acc |↑ |0.8039|± |0.0279| |
|
| - high_school_world_history | 0|none | 0|acc |↑ |0.8228|± |0.0249| |
|
| - international_law | 0|none | 0|acc |↑ |0.7686|± |0.0385| |
|
| - jurisprudence | 0|none | 0|acc |↑ |0.7685|± |0.0408| |
|
| - logical_fallacies | 0|none | 0|acc |↑ |0.7914|± |0.0319| |
|
| - moral_disputes | 0|none | 0|acc |↑ |0.7110|± |0.0244| |
|
| - moral_scenarios | 0|none | 0|acc |↑ |0.4536|± |0.0167| |
|
| - philosophy | 0|none | 0|acc |↑ |0.6913|± |0.0262| |
|
| - prehistory | 0|none | 0|acc |↑ |0.7037|± |0.0254| |
|
| - professional_law | 0|none | 0|acc |↑ |0.4739|± |0.0128| |
|
| - world_religions | 0|none | 0|acc |↑ |0.7953|± |0.0309| |
|
| - other | 1|none | |acc |↑ |0.7036|± |0.0079| |
|
| - business_ethics | 0|none | 0|acc |↑ |0.6400|± |0.0482| |
|
| - clinical_knowledge | 0|none | 0|acc |↑ |0.7094|± |0.0279| |
|
| - college_medicine | 0|none | 0|acc |↑ |0.6358|± |0.0367| |
|
| - global_facts | 0|none | 0|acc |↑ |0.3400|± |0.0476| |
|
| - human_aging | 0|none | 0|acc |↑ |0.6457|± |0.0321| |
|
| - management | 0|none | 0|acc |↑ |0.8544|± |0.0349| |
|
| - marketing | 0|none | 0|acc |↑ |0.8761|± |0.0216| |
|
| - medical_genetics | 0|none | 0|acc |↑ |0.7300|± |0.0446| |
|
| - miscellaneous | 0|none | 0|acc |↑ |0.8148|± |0.0139| |
|
| - nutrition | 0|none | 0|acc |↑ |0.7092|± |0.0260| |
|
| - professional_accounting | 0|none | 0|acc |↑ |0.5071|± |0.0298| |
|
| - professional_medicine | 0|none | 0|acc |↑ |0.7316|± |0.0269| |
|
| - virology | 0|none | 0|acc |↑ |0.5000|± |0.0389| |
|
| - social sciences | 1|none | |acc |↑ |0.7390|± |0.0077| |
|
| - econometrics | 0|none | 0|acc |↑ |0.4561|± |0.0469| |
|
| - high_school_geography | 0|none | 0|acc |↑ |0.8333|± |0.0266| |
|
| - high_school_government_and_politics| 0|none | 0|acc |↑ |0.8808|± |0.0234| |
|
| - high_school_macroeconomics | 0|none | 0|acc |↑ |0.6231|± |0.0246| |
|
| - high_school_microeconomics | 0|none | 0|acc |↑ |0.7437|± |0.0284| |
|
| - high_school_psychology | 0|none | 0|acc |↑ |0.8404|± |0.0157| |
|
| - human_sexuality | 0|none | 0|acc |↑ |0.7481|± |0.0381| |
|
| - professional_psychology | 0|none | 0|acc |↑ |0.6814|± |0.0189| |
|
| - public_relations | 0|none | 0|acc |↑ |0.6455|± |0.0458| |
|
| - security_studies | 0|none | 0|acc |↑ |0.7143|± |0.0289| |
|
| - sociology | 0|none | 0|acc |↑ |0.8259|± |0.0268| |
|
| - us_foreign_policy | 0|none | 0|acc |↑ |0.8200|± |0.0386| |
|
| - stem | 1|none | |acc |↑ |0.5601|± |0.0085| |
|
| - abstract_algebra | 0|none | 0|acc |↑ |0.3500|± |0.0479| |
|
| - anatomy | 0|none | 0|acc |↑ |0.6370|± |0.0415| |
|
| - astronomy | 0|none | 0|acc |↑ |0.7566|± |0.0349| |
|
| - college_biology | 0|none | 0|acc |↑ |0.7639|± |0.0355| |
|
| - college_chemistry | 0|none | 0|acc |↑ |0.4800|± |0.0502| |
|
| - college_computer_science | 0|none | 0|acc |↑ |0.5000|± |0.0503| |
|
| - college_mathematics | 0|none | 0|acc |↑ |0.3200|± |0.0469| |
|
| - college_physics | 0|none | 0|acc |↑ |0.4020|± |0.0488| |
|
| - computer_security | 0|none | 0|acc |↑ |0.7600|± |0.0429| |
|
| - conceptual_physics | 0|none | 0|acc |↑ |0.5574|± |0.0325| |
|
| - electrical_engineering | 0|none | 0|acc |↑ |0.6345|± |0.0401| |
|
| - elementary_mathematics | 0|none | 0|acc |↑ |0.4921|± |0.0257| |
|
| - high_school_biology | 0|none | 0|acc |↑ |0.7710|± |0.0239| |
|
| - high_school_chemistry | 0|none | 0|acc |↑ |0.5665|± |0.0349| |
|
| - high_school_computer_science | 0|none | 0|acc |↑ |0.7000|± |0.0461| |
|
| - high_school_mathematics | 0|none | 0|acc |↑ |0.4074|± |0.0300| |
|
| - high_school_physics | 0|none | 0|acc |↑ |0.4172|± |0.0403| |
|
| - high_school_statistics | 0|none | 0|acc |↑ |0.5278|± |0.0340| |
|
| - machine_learning | 0|none | 0|acc |↑ |0.4732|± |0.0474| |
|
|openbookqa | 1|none | 0|acc |↑ |0.3360|± |0.0211| |
|
| | |none | 0|acc_norm |↑ |0.4220|± |0.0221| |
|
|piqa | 1|none | 0|acc |↑ |0.7943|± |0.0094| |
|
| | |none | 0|acc_norm |↑ |0.7965|± |0.0094| |
|
|rte | 1|none | 0|acc |↑ |0.6968|± |0.0277| |
|
|truthfulqa_mc1 | 2|none | 0|acc |↑ |0.3439|± |0.0166| |
|
|winogrande | 1|none | 0|acc |↑ |0.7364|± |0.0124| |
|
|
|
| Groups |Version|Filter|n-shot|Metric| |Value | |Stderr| |
|
|------------------|------:|------|------|------|---|-----:|---|-----:| |
|
|mmlu | 1|none | |acc |↑ |0.6421|± |0.0039| |
|
| - humanities | 1|none | |acc |↑ |0.5932|± |0.0069| |
|
| - other | 1|none | |acc |↑ |0.7036|± |0.0079| |
|
| - social sciences| 1|none | |acc |↑ |0.7390|± |0.0077| |
|
| - stem | 1|none | |acc |↑ |0.5601|± |0.0085| |
|
``` |
|
|
