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"))