frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc][Attention][Quantization] init property earlier (#13733)

Browse Source 
Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-02-25 11:19:30 +08:00
committed by GitHub
parent 1e15aaef56
commit ab1091d5f2
1 changed files with 5 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/attention/layer.py
View File

@ -85,6 +85,11 @@ class Attention(nn.Module):
self._k_scale_float = 1.0
self._v_scale_float = 1.0
self.num_heads = num_heads
self.head_size = head_size
self.num_kv_heads = num_kv_heads
self.sliding_window = sliding_window
quant_method = quant_config.get_quant_method(
self, prefix=prefix) if quant_config else None
if quant_method is not None:
@ -116,10 +121,6 @@ class Attention(nn.Module):
alibi_slopes, sliding_window, kv_cache_dtype,
blocksparse_params, logits_soft_cap, attn_type,
**extra_impl_args)
self.num_heads = num_heads
self.head_size = head_size
self.num_kv_heads = num_kv_heads
self.sliding_window = sliding_window
self.backend = backend_name_to_enum(attn_backend.get_name())
self.dtype = dtype