biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Doc] Update list of supported models (#8987)

Browse Source
This commit is contained in:
Cyrus Leung
2024-10-02 00:35:39 +08:00
committed by GitHub
parent 35bd215168
commit 4f341bd4bf
2 changed files with 21 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/model_executor/models/__init__.py
View File

@@ -12,6 +12,7 @@ logger = init_logger(__name__)
_GENERATION_MODELS = {
"AquilaModel": ("llama", "LlamaForCausalLM"),
"AquilaForCausalLM": ("llama", "LlamaForCausalLM"), # AquilaChat2
"ArcticForCausalLM": ("arctic", "ArcticForCausalLM"),
"BaiChuanForCausalLM": ("baichuan", "BaiChuanForCausalLM"), # baichuan-7b
"BaichuanForCausalLM": ("baichuan", "BaichuanForCausalLM"), # baichuan-13b
"BloomForCausalLM": ("bloom", "BloomForCausalLM"),
@@ -30,9 +31,11 @@ _GENERATION_MODELS = {
"GPTBigCodeForCausalLM": ("gpt_bigcode", "GPTBigCodeForCausalLM"),
"GPTJForCausalLM": ("gpt_j", "GPTJForCausalLM"),
"GPTNeoXForCausalLM": ("gpt_neox", "GPTNeoXForCausalLM"),
"GraniteForCausalLM": ("granite", "GraniteForCausalLM"),
"InternLMForCausalLM": ("llama", "LlamaForCausalLM"),
"InternLM2ForCausalLM": ("internlm2", "InternLM2ForCausalLM"),
"JAISLMHeadModel": ("jais", "JAISLMHeadModel"),
"JambaForCausalLM": ("jamba", "JambaForCausalLM"),
"LlamaForCausalLM": ("llama", "LlamaForCausalLM"),
# For decapoda-research/llama-*
"LLaMAForCausalLM": ("llama", "LlamaForCausalLM"),
@@ -52,6 +55,7 @@ _GENERATION_MODELS = {
"PersimmonForCausalLM": ("persimmon", "PersimmonForCausalLM"),
"PhiForCausalLM": ("phi", "PhiForCausalLM"),
"Phi3ForCausalLM": ("phi3", "Phi3ForCausalLM"),
"Phi3SmallForCausalLM": ("phi3_small", "Phi3SmallForCausalLM"),
"PhiMoEForCausalLM": ("phimoe", "PhiMoEForCausalLM"),
"Qwen2ForCausalLM": ("qwen2", "Qwen2ForCausalLM"),
"Qwen2MoeForCausalLM": ("qwen2_moe", "Qwen2MoeForCausalLM"),
@@ -62,14 +66,11 @@ _GENERATION_MODELS = {
"StableLmForCausalLM": ("stablelm", "StablelmForCausalLM"),
"Starcoder2ForCausalLM": ("starcoder2", "Starcoder2ForCausalLM"),
"SolarForCausalLM": ("solar", "SolarForCausalLM"),
"ArcticForCausalLM": ("arctic", "ArcticForCausalLM"),
"XverseForCausalLM": ("xverse", "XverseForCausalLM"),
"Phi3SmallForCausalLM": ("phi3_small", "Phi3SmallForCausalLM"),
# NOTE: The below models are for speculative decoding only
"MedusaModel": ("medusa", "Medusa"),
"EAGLEModel": ("eagle", "EAGLE"),
"MLPSpeculatorPreTrainedModel": ("mlp_speculator", "MLPSpeculator"),
"JambaForCausalLM": ("jamba", "JambaForCausalLM"),
"GraniteForCausalLM": ("granite", "GraniteForCausalLM")
}
_EMBEDDING_MODELS = {