vLLM - 查看模型是否支持

---

支持的模型：https://docs.vllm.ai/en/latest/models/supported_models.html

---

要确定是否支持给定模型，您可以检查HF存储库中的config.json文件。

如果"architectures"字段包含下面列出的模型架构，那么理论上应该支持它。

---

查看模型架构

查看 模型的 config.json 中的 architectures

cat ~/.cache/huggingface/hub/models--deepseek-ai--DeepSeek-R1-Distill-Qwen-32B/snapshots/3865e12a1eb7cbd641ab3f9dfc28c588c6b0c1e9

---

DeepSeek-R1-Distill-Qwen-32B architectures 为 Qwen2ForCausalLM

{
  "architectures": [
    "Qwen2ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "bos_token_id": 151643,
  "eos_token_id": 151643,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 27648,
  "max_position_embeddings": 131072,
  "max_window_layers": 64,
  "model_type": "qwen2",
  "num_attention_heads": 40,
  "num_hidden_layers": 64,
  "num_key_value_heads": 8,
  "rms_norm_eps": 1e-05,
  "rope_theta": 1000000.0,
  "sliding_window": 131072,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.43.1",
  "use_cache": true,
  "use_sliding_window": false,
  "vocab_size": 152064
}

---

**Alibaba-NLP–gte-Qwen2-7B-instruct ** – Qwen2ForCausalLM

{
  "architectures": [
    "Qwen2ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoModel": "modeling_qwen.Qwen2Model",
    "AutoModelForCausalLM": "modeling_qwen.Qwen2ForCausalLM",
    "AutoModelForSequenceClassification": "modeling_qwen.Qwen2ForSequenceClassification"
  },
  "bos_token_id": 151643,
  "eos_token_id": 151643,
  "hidden_act": "silu",
  "hidden_size": 3584,
  "initializer_range": 0.02,
  "intermediate_size": 18944,
  "max_position_embeddings": 131072,
  "max_window_layers": 28,
  "model_type": "qwen2",
  "num_attention_heads": 28,
  "num_hidden_layers": 28,
  "num_key_value_heads": 4,
  "rms_norm_eps": 1e-06,
  "rope_theta": 1000000.0,
  "sliding_window": 131072,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.41.2",
  "use_cache": true,
  "use_sliding_window": false,
  "vocab_size": 151646
}

---

代码检查是否支持

就是直接运行

from vllm import LLM

# For generative models (task=generate) only
llm = LLM(model=..., task="generate")  # Name or path of your model
output = llm.generate("Hello, my name is")
print(output)

# For pooling models (task={embed,classify,reward,score}) only
llm = LLM(model=..., task="embed")  # Name or path of your model
output = llm.encode("Hello, my name is")
print(output)

---

from vllm import LLM

model_name = 'deepseek-ai/DeepSeek-R1-Distill-Qwen-32B'
model_name = 'Qwen/Qwen2.5-Coder-7B-Instruct' 
model_name = 'Qwen/Qwen2.5-Coder-32B-Instruct-GPTQ-Int4'

llm = LLM(model=model_name, task="generate")  # Name or path of your model
output = llm.generate("Hello, my name is")
print(output)

---