[Misc]Reduce BNB static variable (#9987)

Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
2024-11-05 01:04:40 +08:00
parent 8d72bb20fa
commit fb2716d641
10 changed files with 20 additions and 46 deletions
--- a/vllm/model_executor/models/falcon.py
+++ b/vllm/model_executor/models/falcon.py
@@ -401,8 +401,6 @@ class FalconForCausalLM(nn.Module, SupportsPP):
        ".dense_h_to_4h.",
        ".dense_4h_to_h.",
    ]
-    # in TP, these weights are partitioned along the column dimension (dim=-1)
-    column_parallel_weights_modules = [".dense_4h_to_h.", ".dense."]

    def __init__(
        self,