@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 5204ff5c3f
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, 00:45 UTC

[Bugfix] Fix Granite 3.0 MoE model loading (#12446)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

Cyrus Leung committed Jan 27, 2025 at 13:26 UTC 5204ff5c3feeb96e8a6eea65dfcb78395f90d4d8
1 file changed +1
vllm/model_executor/models/granitemoe.py
+1
@@ -348,6 +348,7 @@ class GraniteMoeForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
348
349 self.config = config
350 self.lora_config = lora_config
351 + self.quant_config = quant_config # Required by MixtralForCausalLM
352
353 self.model = GraniteMoeModel(vllm_config=vllm_config,
354 prefix=maybe_prefix(prefix, "model"))