[Qwen2.5] Support bnb quant for Qwen2.5 (vllm-project#9467)

opendatahub-io · Oct 18, 2024 · 154a8ae · 154a8ae
1 parent de4008e
commit 154a8ae
Showing 1 changed file with 8 additions and 0 deletions.
diff --git a/vllm/model_executor/models/qwen2.py b/vllm/model_executor/models/qwen2.py
@@ -364,6 +364,14 @@ class Qwen2ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
     ]
     embedding_modules = {}
     embedding_padding_modules = []
+    bitsandbytes_stacked_params_mapping = {
+        # shard_name, weight_name, index
+        "q_proj": ("qkv_proj", 0),
+        "k_proj": ("qkv_proj", 1),
+        "v_proj": ("qkv_proj", 2),
+        "gate_proj": ("gate_up_proj", 0),
+        "up_proj": ("gate_up_proj", 1),
+    }
 
     def __init__(
         self,