93 lines
2.7 KiB
YAML
93 lines
2.7 KiB
YAML
# ==========================================
|
|
# ACTUAL TEST CASES
|
|
# ==========================================
|
|
|
|
test_cases:
|
|
- name: "Qwen2.5-VL-7B-Instruct-epd"
|
|
model: "Qwen/Qwen2.5-VL-7B-Instruct"
|
|
service_mode: "epd"
|
|
envs:
|
|
ENCODE_PORT: "DEFAULT_PORT"
|
|
PD_PORT: "DEFAULT_PORT"
|
|
PROXY_PORT: "DEFAULT_PORT"
|
|
epd_server_cmds:
|
|
- - "--port"
|
|
- "$ENCODE_PORT"
|
|
- "--model"
|
|
- "Qwen/Qwen2.5-VL-7B-Instruct"
|
|
- "--gpu-memory-utilization"
|
|
- "0.01"
|
|
- "--tensor-parallel-size"
|
|
- "1"
|
|
- "--enforce-eager"
|
|
- "--no-enable-prefix-caching"
|
|
- "--max-model-len"
|
|
- "10000"
|
|
- "--max-num-batched-tokens"
|
|
- "10000"
|
|
- "--max-num-seqs"
|
|
- "1"
|
|
- "--ec-transfer-config"
|
|
- '{"ec_connector_extra_config":{"shared_storage_path":"/dev/shm/epd/storage"},"ec_connector":"ECExampleConnector","ec_role": "ec_producer"}'
|
|
- - "--port"
|
|
- "$PD_PORT"
|
|
- "--model"
|
|
- "Qwen/Qwen2.5-VL-7B-Instruct"
|
|
- "--gpu-memory-utilization"
|
|
- "0.95"
|
|
- "--tensor-parallel-size"
|
|
- "1"
|
|
- "--enforce-eager"
|
|
- "--max-model-len"
|
|
- "10000"
|
|
- "--max-num-batched-tokens"
|
|
- "10000"
|
|
- "--max-num-seqs"
|
|
- "128"
|
|
- "--ec-transfer-config"
|
|
- '{"ec_connector_extra_config":{"shared_storage_path":"/dev/shm/epd/storage"},"ec_connector":"ECExampleConnector","ec_role": "ec_consumer"}'
|
|
epd_proxy_args:
|
|
- "--host"
|
|
- "127.0.0.1"
|
|
- "--port"
|
|
- "$PROXY_PORT"
|
|
- "--encode-servers-urls"
|
|
- "http://localhost:$ENCODE_PORT"
|
|
- "--decode-servers-urls"
|
|
- "http://localhost:$PD_PORT"
|
|
- "--prefill-servers-urls"
|
|
- "disable"
|
|
test_content:
|
|
benchmarks:
|
|
warm_up:
|
|
case_type: performance
|
|
dataset_path: vllm-ascend/textvqa-perf-1080p
|
|
request_conf: vllm_api_stream_chat
|
|
dataset_conf: textvqa/textvqa_gen_base64
|
|
num_prompts: 50
|
|
max_out_len: 20
|
|
batch_size: 32
|
|
request_rate: 0
|
|
baseline: 1
|
|
threshold: 0.97
|
|
acc:
|
|
case_type: accuracy
|
|
dataset_path: vllm-ascend/textvqa-lite
|
|
request_conf: vllm_api_stream_chat
|
|
dataset_conf: textvqa/textvqa_gen_base64
|
|
max_out_len: 2048
|
|
batch_size: 128
|
|
baseline: 82.05
|
|
threshold: 5
|
|
perf:
|
|
case_type: performance
|
|
dataset_path: vllm-ascend/textvqa-perf-1080p
|
|
request_conf: vllm_api_stream_chat
|
|
dataset_conf: textvqa/textvqa_gen_base64
|
|
num_prompts: 512
|
|
max_out_len: 256
|
|
batch_size: 128
|
|
request_rate: 0
|
|
baseline: 1
|
|
threshold: 0.97
|