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 15, 16:00 UTC
[Bugfix] Re-enable support for `ChatGLMForConditionalGeneration` (#16187)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Apr 7, 2025 at 23:15 UTC
027b204ff15e803190775e04d39973606a3a7021
3 files changed
+5
-2
docs/source/models/supported_models.md
+2
-2
@@ -233,9 +233,9 @@ See [this page](#generative-models) for more information on how to use generativ
233
* `facebook/bart-base`, `facebook/bart-large-cnn`, etc.
234
*
235
*
236
-- * `ChatGLMModel`
236
+- * `ChatGLMModel`, `ChatGLMForConditionalGeneration`
237
* ChatGLM
238
- * `THUDM/chatglm2-6b`, `THUDM/chatglm3-6b`, etc.
238
+ * `THUDM/chatglm2-6b`, `THUDM/chatglm3-6b`, `ShieldLM-6B-chatglm3`, etc.
239
* ✅︎
240
* ✅︎
241
- * `CohereForCausalLM`, `Cohere2ForCausalLM`
tests/models/registry.py
+2
@@ -124,6 +124,8 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
124
"BloomForCausalLM": _HfExamplesInfo("bigscience/bloomz-1b1"),
125
"ChatGLMModel": _HfExamplesInfo("THUDM/chatglm3-6b",
126
trust_remote_code=True),
127
+ "ChatGLMForConditionalGeneration": _HfExamplesInfo("thu-coai/ShieldLM-6B-chatglm3", # noqa: E501
128
+ trust_remote_code=True),
129
"CohereForCausalLM": _HfExamplesInfo("CohereForAI/c4ai-command-r-v01",
130
trust_remote_code=True),
131
"Cohere2ForCausalLM": _HfExamplesInfo("CohereForAI/c4ai-command-r7b-12-2024", # noqa: E501
vllm/model_executor/models/registry.py
+1
@@ -43,6 +43,7 @@ _TEXT_GENERATION_MODELS = {
43
"BambaForCausalLM": ("bamba", "BambaForCausalLM"),
44
"BloomForCausalLM": ("bloom", "BloomForCausalLM"),
45
"ChatGLMModel": ("chatglm", "ChatGLMForCausalLM"),
46
+ "ChatGLMForConditionalGeneration": ("chatglm", "ChatGLMForCausalLM"),
47
"CohereForCausalLM": ("commandr", "CohereForCausalLM"),
48
"Cohere2ForCausalLM": ("commandr", "CohereForCausalLM"),
49
"DbrxForCausalLM": ("dbrx", "DbrxForCausalLM"),