apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-04 16:22:57 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Adapt for iluvatar gpu (#2684)

Browse Source
This commit is contained in:
liddk1121
2025-07-07 16:53:14 +08:00
committed by GitHub
parent 2579e8fea8
commit 1b54a2831e
50 changed files with 4485 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
fastdeploy/model_executor/layers/attention/attention.py
View File

@@ -67,7 +67,7 @@ class Attention(nn.Layer):
self.num_heads: int = fd_config.model_config.num_attention_heads // fd_config.parallel_config.tensor_parallel_degree
self.head_dim: int = fd_config.model_config.head_dim
self.kv_num_heads: int = \
fd_config.model_config.num_key_value_heads // fd_config.parallel_config.tensor_parallel_degree
max(1, fd_config.model_config.num_key_value_heads // fd_config.parallel_config.tensor_parallel_degree)
self.layer_id: int = layer_id
self.v_head_dim: int = v_head_dim if v_head_dim > 0 else self.head_dim
self.rope_type: str = rope_type