EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[CI] Add multi-nodes longseq configs of DeepSeek-R1-W8A8 & Qwen3-235B-W8A8 (#5381)

Browse Source 
### What this PR does / why we need it?
add DeepSeek-R1-W8A8 and Qwen3-235B-W8A8 configs in multi-nodes and
longseq (PCP&DCP) scenario

- vLLM version: release/v0.13.0
- vLLM main:
bc0a5a0c08
---------
Signed-off-by: daishixun <dsxsteven@sina.com>
This commit is contained in:
dsxsteven
2026-01-04 10:38:40 +08:00
committed by GitHub
parent 799b41a9f4
commit 3c7e6c6817
3 changed files with 208 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
.github/workflows/nightly_test_a3.yaml vendored
View File

@@ -68,6 +68,12 @@ jobs:
- name: multi-node-qwenw8a8-2node-eplb
config_file_path: Qwen3-235B-W8A8-EPLB.yaml
size: 2
- name: multi-node-deepseek-r1-w8a8-longseq
config_file_path: DeepSeek-R1-W8A8-longseq.yaml
size: 2
- name: multi-node-qwenw8a8-2node-longseq
config_file_path: Qwen3-235B-W8A8-longseq.yaml
size: 2
uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
with:
soc_version: a3

109
tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-longseq.yaml Normal file
View File

@@ -0,0 +1,109 @@
test_name: "test DeepSeek-R1-W8A8-longseq disaggregated_prefill"
model: "vllm-ascend/DeepSeek-R1-0528-W8A8"
num_nodes: 2
npu_per_node: 16
env_common:
VLLM_USE_MODELSCOPE: true
HCCL_BUFFSIZE: 1024
SERVER_PORT: 8080
OMP_PROC_BIND: false
OMP_NUM_THREADS: 10
PYTORCH_NPU_ALLOC_CONF: expandable_segments:True
HCCL_DETERMINISTIC: True
TASK_QUEUE_ENABLE: 1
HCCL_OP_RETRY_ENABLE: "L0:0, L1:0"
disaggregated_prefill:
enabled: true
prefiller_host_index: [0]
decoder_host_index: [1]
deployment:
-
server_cmd: >
vllm serve vllm-ascend/DeepSeek-R1-0528-W8A8
--host 0.0.0.0
--port $SERVER_PORT
--data-parallel-size 1
--decode-context-parallel-size 8
--prefill-context-parallel-size 2
--tensor-parallel-size 8
--cp-kv-cache-interleave-size 128
--enforce-eager
--enable-expert-parallel
--seed 1024
--quantization ascend
--max-num-seqs 4
--max-model-len 32768
--max-num-batched-tokens 16384
--trust-remote-code
--gpu-memory-utilization 0.9
--enable-chunked-prefill
--speculative-config '{"num_speculative_tokens": 3, "method":"mtp"}'
--kv-transfer-config
'{"kv_connector": "MooncakeConnectorV1",
"kv_role": "kv_producer",
"kv_port": "30000",
"engine_id": "0",
"kv_connector_module_path": "vllm_ascend.distributed.mooncake_connector",
"kv_connector_extra_config": {
"prefill": {
"dp_size": 1,
"tp_size": 8
},
"decode": {
"dp_size": 2,
"tp_size": 8
}
}
}'
-
server_cmd: >
vllm serve vllm-ascend/DeepSeek-R1-0528-W8A8
--host 0.0.0.0
--port $SERVER_PORT
--data-parallel-size 2
--decode-context-parallel-size 2
--prefill-context-parallel-size 1
--tensor-parallel-size 8
--cp-kv-cache-interleave-size 128
--enable-expert-parallel
--seed 1024
--quantization ascend
--max-num-seqs 4
--max-model-len 32768
--max-num-batched-tokens 256
--trust-remote-code
--gpu-memory-utilization 0.9
--compilation_config '{"cudagraph_capture_sizes":[4,8,12,16],"cudagraph_mode": "FULL_DECODE_ONLY"}'
--enable-chunked-prefill
--speculative-config '{"num_speculative_tokens": 3, "method":"mtp"}'
--kv-transfer-config
'{"kv_connector": "MooncakeConnectorV1",
"kv_role": "kv_consumer",
"kv_port": "30100",
"engine_id": "1",
"kv_connector_module_path": "vllm_ascend.distributed.mooncake_connector",
"kv_connector_extra_config": {
"prefill": {
"dp_size": 1,
"tp_size": 8
},
"decode": {
"dp_size": 2,
"tp_size": 8
}
}
}'
benchmarks:
acc:
case_type: accuracy
dataset_path: vllm-ascend/gsm8k
request_conf: vllm_api_general_chat
dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_chat_prompt
max_out_len: 32768
batch_size: 512
baseline: 95
threshold: 5

93
tests/e2e/nightly/multi_node/config/models/Qwen3-235B-W8A8-longseq.yaml Normal file
View File

@@ -0,0 +1,93 @@
test_name: "test Qwen3-235B-A22B-W8A8-longseq disaggregated_prefill"
model: "vllm-ascend/Qwen3-235B-A22B-W8A8"
num_nodes: 2
npu_per_node: 16
env_common:
VLLM_USE_MODELSCOPE: true
OMP_PROC_BIND: false
OMP_NUM_THREADS: 100
HCCL_BUFFSIZE: 1024
SERVER_PORT: 8080
NUMEXPR_MAX_THREADS: 128
disaggregated_prefill:
enabled: true
prefiller_host_index: [0]
decoder_host_index: [1]
deployment:
-
server_cmd: >
vllm serve "vllm-ascend/Qwen3-235B-A22B-W8A8"
--host 0.0.0.0
--port $SERVER_PORT
--data-parallel-size 1
--decode-context-parallel-size 2
--prefill-context-parallel-size 2
--tensor-parallel-size 8
--cp-kv-cache-interleave-size 128
--seed 1024
--enforce-eager
--enable-expert-parallel
--max-num-seqs 16
--max-model-len 8192
--max-num-batched-tokens 8192
--quantization ascend
--trust-remote-code
--no-enable-prefix-caching
--gpu-memory-utilization 0.9
--kv-transfer-config
'{"kv_connector": "MooncakeConnectorV1",
"kv_role": "kv_producer",
"kv_port": "30000",
"engine_id": "0",
"kv_connector_module_path": "vllm_ascend.distributed.mooncake_connector",
"kv_connector_extra_config": {
"prefill": {
"dp_size": 1,
"tp_size": 8
},
"decode": {
"dp_size": 2,
"tp_size": 8
}
}
}'
-
server_cmd: >
vllm serve "vllm-ascend/Qwen3-235B-A22B-W8A8"
--host 0.0.0.0
--port $SERVER_PORT
--data-parallel-size 2
--decode-context-parallel-size 2
--prefill-context-parallel-size 1
--tensor-parallel-size 8
--cp-kv-cache-interleave-size 128
--seed 1024
--quantization ascend
--max-num-seqs 16
--max-model-len 8192
--max-num-batched-tokens 8192
--enable-expert-parallel
--trust-remote-code
--no-enable-prefix-caching
--gpu-memory-utilization 0.9
--compilation_config '{"cudagraph_mode": "FULL_DECODE_ONLY"}'
--kv-transfer-config
'{"kv_connector": "MooncakeConnectorV1",
"kv_role": "kv_consumer",
"kv_port": "30100",
"engine_id": "1",
"kv_connector_module_path": "vllm_ascend.distributed.mooncake_connector",
"kv_connector_extra_config": {
"prefill": {
"dp_size": 1,
"tp_size": 8
},
"decode": {
"dp_size": 2,
"tp_size": 8
}
}
}'
benchmarks: