GPTQ
Collection
quantized LLMs by AutoGPTQ
•
31 items
•
Updated
•
2
MaziyarPanahi/UNA-34Beagles-32K-bf16-v1-GPTQ is a quantized (GPTQ) version of one-man-army/UNA-34Beagles-32K-bf16-v1
pip install --upgrade accelerate auto-gptq transformers
from transformers import AutoTokenizer, pipeline
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
import torch
model_id = "MaziyarPanahi/UNA-34Beagles-32K-bf16-v1-GPTQ"
quantize_config = BaseQuantizeConfig(
bits=4,
group_size=128,
desc_act=False
)
model = AutoGPTQForCausalLM.from_quantized(
model_id,
use_safetensors=True,
device="cuda:0",
quantize_config=quantize_config)
tokenizer = AutoTokenizer.from_pretrained(model_id)
pipe = pipeline(
"text-generation",
model=model,
tokenizer=tokenizer,
max_new_tokens=512,
temperature=0.7,
top_p=0.95,
repetition_penalty=1.1
)
outputs = pipe("What is a large language model?")
print(outputs[0]["generated_text"])
Detailed results can be found here
Metric | Value |
---|---|
Avg. | 29.15 |
AI2 Reasoning Challenge (25-Shot) | 26.11 |
HellaSwag (10-Shot) | 26.29 |
MMLU (5-Shot) | 24.43 |
TruthfulQA (0-shot) | 47.27 |
Winogrande (5-shot) | 50.83 |
GSM8k (5-shot) | 0.00 |
Base model
one-man-army/UNA-34Beagles-32K-bf16-v1