@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / dbdd3b5e5a
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 16, 09:30 UTC

[misc] comment to avoid future confusion about baichuan (#9620)

Signed-off-by: youkaichao <youkaichao@gmail.com>

youkaichao committed Oct 23, 2024 at 09:14 UTC dbdd3b5e5ace989923a5abb549780564980bc11e
2 files changed +10 -4
vllm/model_executor/models/baichuan.py
+6 -2
@@ -432,7 +432,9 @@ class BaiChuanBaseForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
432
433
434 class BaichuanForCausalLM(BaiChuanBaseForCausalLM):
435 - """Baichuan 13B and Baichuan2 7B/13B."""
435 + """Baichuan 13B and Baichuan2 7B/13B.
436 + NOTE: the class name has a lower case 'c'.
437 + """
438
439 def __init__(
440 self,
@@ -450,7 +452,9 @@ class BaichuanForCausalLM(BaiChuanBaseForCausalLM):
452
453
454 class BaiChuanForCausalLM(BaiChuanBaseForCausalLM):
453 - """Baichuan 7B."""
455 + """Baichuan 7B.
456 + NOTE: the class name has an upper case 'C'.
457 + """
458
459 def __init__(
460 self,
vllm/model_executor/models/registry.py
+4 -2
@@ -26,8 +26,10 @@ _TEXT_GENERATION_MODELS = {
26 "AquilaModel": ("llama", "LlamaForCausalLM"),
27 "AquilaForCausalLM": ("llama", "LlamaForCausalLM"), # AquilaChat2
28 "ArcticForCausalLM": ("arctic", "ArcticForCausalLM"),
29 - "BaiChuanForCausalLM": ("baichuan", "BaiChuanForCausalLM"), # baichuan-7b
30 - "BaichuanForCausalLM": ("baichuan", "BaichuanForCausalLM"), # baichuan-13b
29 + # baichuan-7b, upper case 'C' in the class name
30 + "BaiChuanForCausalLM": ("baichuan", "BaiChuanForCausalLM"),
31 + # baichuan-13b, lower case 'c' in the class name
32 + "BaichuanForCausalLM": ("baichuan", "BaichuanForCausalLM"),
33 "BloomForCausalLM": ("bloom", "BloomForCausalLM"),
34 # ChatGLMModel supports multimodal
35 "CohereForCausalLM": ("commandr", "CohereForCausalLM"),