ModelCloud/DeepSeek-V2-Chat-0628-gptq-4bit

This model has been quantized using GPTQModel.

• bits: 4
• group_size: 128
• desc_act: true
• static_groups: false
• sym: true
• lm_head: false
• damp_percent: 0.01
• 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"

You can use GPTQModel for model inference.

import torch
from transformers import AutoTokenizer, GenerationConfig
from gptqmodel import GPTQModel
model_name = "ModelCloud/DeepSeek-V2-Chat-0628-gptq-4bit"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
# `max_memory` should be set based on your devices
max_memory = {i: "75GB" for i in range(2)}
# `device_map` cannot be set to `auto`
model = GPTQModel.from_quantized(model_name, trust_remote_code=True, device_map="sequential", max_memory=max_memory, torch_dtype=torch.float16, attn_implementation="eager")

model.generation_config = GenerationConfig.from_pretrained(model_name)
model.generation_config.pad_token_id = model.generation_config.eos_token_id

messages = [
    {"role": "user", "content": "Write a piece of quicksort code in C++"}
]
input_tensor = tokenizer.apply_chat_template(messages, 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)