apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[benchmark]add qwen3-235b pd+ep yaml (#5225)

Browse Source
This commit is contained in:
xiegegege
2025-11-25 19:53:30 +08:00
committed by GitHub
parent df2be1cf16
commit eae34a416c
3 changed files with 27 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
benchmarks/yaml/eb45-32k-wint4-mtp-h800-tp4.yaml
View File

@@ -1,6 +1,6 @@
max_model_len: 32768 max_model_len: 32768
max_num_seqs: 96 max_num_seqs: 96
gpu_memory_utilization: 0.85 gpu_memory_utilization: 0.8
kv_cache_ratio: 0.71 kv_cache_ratio: 0.71
tensor_parallel_size: 4 tensor_parallel_size: 4
quantization: wint4 quantization: wint4

13
benchmarks/yaml/qwen3-235b-32k-fp8-tp1-dp4_decode.yaml Normal file
View File

@@ -0,0 +1,13 @@
max_model_len: 32768
max_num_seqs: 32
data_parallel_size: 4
tensor_parallel_size: 1
enable_expert_parallel: True
enable_prefix_caching: False
splitwise_role: decode
cache_transfer_protocol: "rdma"
rdma_comm_ports: "7671,7672,7673,7674"
pd_comm_port: "2335"
engine_worker_queue_port: "4582,4583,4584,4585"
graph_optimization_config:
use_cudagraph: False

13
benchmarks/yaml/qwen3-235b-32k-fp8-tp1-dp4_prefill.yaml Normal file
View File

@@ -0,0 +1,13 @@
max_model_len: 32768
max_num_seqs: 32
data_parallel_size: 4
tensor_parallel_size: 1
enable_expert_parallel: True
enable_prefix_caching: False
splitwise_role: prefill
cache_transfer_protocol: "rdma"
rdma_comm_ports: "7675,7676,7677,7678"
pd_comm_port: "2334"
engine_worker_queue_port: "4368,4369,4360,4361"
graph_optimization_config:
use_cudagraph: False