mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-05 08:37:06 +08:00
[feat] support fa3 backend for pd disaggregated (#2695)
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
* support fa3 backend run in pd disaggregated * support fa3 backend run in pd disaggregated * support fa3 backend run in pd disaggregated * support fa3 backend run in pd disaggregated * delete use_fast_ffn
This commit is contained in:
@@ -214,7 +214,7 @@ def load_tp_checkpoint_v1(
|
||||
need_tp = True if tensor_parallel_filtered_map else False
|
||||
state_dict = {}
|
||||
for key, weight in weights_iterator:
|
||||
paddle.device.cuda.synchronize()
|
||||
paddle.device.synchronize()
|
||||
if need_tp and key in tensor_parallel_filtered_map:
|
||||
action = tensor_parallel_filtered_map.pop(key)
|
||||
tensor = action(weight).clone()
|
||||
|
Reference in New Issue
Block a user