z3st/modelhub-submit-agent
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix: match platform auto-generated configParams format

Browse Source 
- Remove /bin/bash -ic wrapper from sut_config.command
- Use list format for all command arguments (matches platform auto-gen)
- Add nv_framework field
- Remove unnecessary fields: docker_image, nv_docker_image, modelFormat, model, cpu_num
- Fix ref_config to match platform format (port 80, max_model_len 4096)
This commit is contained in:
z3st
2026-07-23 22:35:50 +08:00
parent 12bf6d637f
commit 1811a66aee
1 changed files with 20 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
main.py
View File

@@ -243,16 +243,15 @@ def check_queue_available(gpu: str, token: str) -> int:
def build_config_params(gpu: str) -> str:
"""构建 YAML 配置"""
"""构建 YAML 配置 - 完全匹配平台自动生成的格式"""
config = GPU_CONFIGS.get(gpu, {})
framework = config.get('framework', 'vllm')
docker_image = config.get('docker_image', '')
if framework == 'llama.cpp':
# hygon_k100-ai 使用 llama.cpp
params = {
'framework': 'llama.cpp',
'docker_image': docker_image,
'nv_docker_image': docker_image,
'nv_framework': 'llama.cpp',
'api': 'completion',
'max_tokens': 1024,
'temperature': 0.7,
@@ -260,22 +259,21 @@ def build_config_params(gpu: str) -> str:
'top_p': 0.9,
'lang': 'zh',
'max_model_len': 4096,
'modelFormat': 'GGUF',
'sut_config': {
'gpu_num': 1,
'values': {
'command': [
'/bin/bash', '-ic',
'llama-server --model /model --alias llm --threads 20 '
'--n-gpu-layers 999 --prio 3 --min_p 0.01 '
'--ctx-size 4096 --host 0.0.0.0 --port 8000 --jinja --flash-attn off'
'llama-server', '--model', '/model', '--alias', 'llm',
'--threads', '20', '--n-gpu-layers', '999', '--prio', '3',
'--min_p', '0.01', '--ctx-size', '4096',
'--host', '0.0.0.0', '--port', '8000',
'--jinja', '--flash-attn', 'off',
]
}
},
'ref_config': {
'gpu_num': 1,
'values': {
'cpu_num': 2,
'command': [
'llama-server', '--model', '/model', '--alias', 'llm',
'--threads', '20', '--n-gpu-layers', '999',
@@ -283,13 +281,12 @@ def build_config_params(gpu: str) -> str:
]
}
},
'model': 'llm',
}
else:
# vllm (P800, Ascend, MetaX)
params = {
'framework': 'vllm',
'docker_image': docker_image,
'nv_docker_image': docker_image,
'nv_framework': 'vllm',
'api': 'completion',
'max_tokens': 1024,
'temperature': 0.7,
@@ -297,20 +294,9 @@ def build_config_params(gpu: str) -> str:
'top_p': 0.9,
'lang': 'zh',
'max_model_len': 2048,
'modelFormat': 'HuggingFace',
'sut_config': {
'gpu_num': 1,
'values': {
'command': [
'/bin/bash', '-ic',
'vllm serve /model --port 8000 --served-model-name llm --max-model-len 2048 --dtype auto --gpu-memory-utilization 0.95 -tp 1 --enforce-eager --trust-remote-code'
]
}
},
'ref_config': {
'gpu_num': 1,
'values': {
'cpu_num': 2,
'command': [
'vllm', 'serve', '/model', '--port', '8000',
'--served-model-name', 'llm', '--max-model-len', '2048',
@@ -319,7 +305,16 @@ def build_config_params(gpu: str) -> str:
]
}
},
'model': 'llm',
'ref_config': {
'gpu_num': 1,
'values': {
'command': [
'vllm', 'serve', '/model', '--port', '80',
'--served-model-name', 'llm', '--max-model-len', '4096',
'--enforce-eager', '--trust-remote-code', '-tp', '1',
]
}
},
}
return yaml.dump(params, default_flow_style=False, allow_unicode=True, width=1000)