apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[FDConfig]Remove splitwise_role and engine_worker_queue_port in FDConfig (#4147)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
Details
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
Details
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
Details
CE Compile Job / Show Code Archive Output (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
Details
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Details
Deploy GitHub Pages / deploy (push) Has been cancelled
Details

Browse Source 
* remove splitwise_role and engine_worker_queue_port

* fix xpu

* fix xpu

* fix xpu

* fix unittest

* resolve conflct
This commit is contained in:
YuanRisheng
2025-09-19 17:01:52 +08:00
committed by GitHub
parent ee9d8a840a
commit 24180fba0a
23 changed files with 129 additions and 89 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
fastdeploy/worker/gpu_model_runner.py
View File

@@ -193,7 +193,7 @@ class GPUModelRunner(ModelRunnerBase):
"""
if_only_prefill = True
decode_exists = None
if self.fd_config.parallel_config.use_ep and self.fd_config.parallel_config.splitwise_role == "mixed":
if self.fd_config.parallel_config.use_ep and self.fd_config.scheduler_config.splitwise_role == "mixed":
only_prefill_batch_list = []
decode_exists = self.exist_decode()
paddle.distributed.all_gather_object(only_prefill_batch_list, not decode_exists)
@@ -211,7 +211,7 @@ class GPUModelRunner(ModelRunnerBase):
if_only_decode = True
prefill_exists = None
# mix ep in single node
if self.fd_config.parallel_config.use_ep and self.fd_config.parallel_config.splitwise_role == "mixed":
if self.fd_config.parallel_config.use_ep and self.fd_config.scheduler_config.splitwise_role == "mixed":
only_decode_batch_list = []
prefill_exists = self.exist_prefill()
paddle.distributed.all_gather_object(only_decode_batch_list, not prefill_exists)
@@ -1103,8 +1103,8 @@ class GPUModelRunner(ModelRunnerBase):
# Update config about moe for better performance
# TODO(wanglongzhi):Modifying the config at runtime is not appropriate; it needs to be moved to forward_meta. It will be used in MoEMethodBase.apply()
if self.fd_config.parallel_config.use_ep and self.fd_config.parallel_config.splitwise_role == "mixed":
self.fd_config.parallel_config.moe_phase.phase = "decode" if if_only_decode else "prefill"
if self.fd_config.parallel_config.use_ep and self.fd_config.scheduler_config.splitwise_role == "mixed":
self.fd_config.model_config.moe_phase.phase = "decode" if if_only_decode else "prefill"
# Update Batch type for cuda graph for only_prefill_batch
only_prefill_use_cudagraph = self.use_cudagraph and self.cudagraph_only_prefill and self.only_prefill()
@@ -1145,7 +1145,9 @@ class GPUModelRunner(ModelRunnerBase):
kv_cache_scale_shape = [kv_cache_shape[0], kv_cache_shape[1], kv_cache_shape[2]]
local_rank = self.local_rank % self.parallel_config.tensor_parallel_size
if not profile and (self.cache_config.enable_prefix_caching or self.parallel_config.splitwise_role != "mixed"):
if not profile and (
self.cache_config.enable_prefix_caching or self.scheduler_config.splitwise_role != "mixed"
):
cache_kvs_list = []
for i in range(self.model_config.num_hidden_layers):
key_cache = paddle.empty(shape=[], dtype=cache_type)
@@ -1711,7 +1713,7 @@ class GPUModelRunner(ModelRunnerBase):
stop_seqs_len=self.share_inputs["stop_seqs_len"],
)
if self.speculative_config.method in ["mtp"] and self.parallel_config.splitwise_role == "prefill":
if self.speculative_config.method in ["mtp"] and self.scheduler_config.splitwise_role == "prefill":
skip_save_output = True
else:
skip_save_output = False