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 14, 23:45 UTC
remove GLM-4.5 quantization wrong Code (#21435)
Yuxuan Zhang committed
Jul 24, 2025 at 16:52 UTC
85bda9e7d05371af6bb9d0052b1eb2f85d3cde29
3 files changed
+2
-3
vllm/entrypoints/openai/tool_parsers/glm4_moe_tool_parser.py
+1
-1
@@ -20,7 +20,7 @@ from vllm.transformers_utils.tokenizer import AnyTokenizer
20
logger = init_logger(__name__)
21
22
23
-@ToolParserManager.register_module("glm4_moe")
23
+@ToolParserManager.register_module("glm45")
24
class Glm4MoeModelToolParser(ToolParser):
25
26
def __init__(self, tokenizer: AnyTokenizer):
vllm/model_executor/models/glm4_moe.py
-1
@@ -390,7 +390,6 @@ class Glm4MoeModel(nn.Module):
390
self.embed_tokens = VocabParallelEmbedding(
391
config.vocab_size,
392
config.hidden_size,
393
- quant_config=quant_config,
393
prefix=f"{prefix}.embed_tokens")
394
else:
395
self.embed_tokens = PPMissingLayer()
vllm/reasoning/glm4_moe_reasoning_parser.py
+1
-1
@@ -14,7 +14,7 @@ from vllm.reasoning import ReasoningParser, ReasoningParserManager
14
logger = init_logger(__name__)
15
16
17
-@ReasoningParserManager.register_module("glm4_moe")
17
+@ReasoningParserManager.register_module("glm45")
18
class Glm4MoeModelReasoningParser(ReasoningParser):
19
"""
20
Reasoning parser for the Glm4MoeModel model.