# 改成miniconda export PATH=/root/miniconda3/condabin:/root/miniconda3/bin:$PATH from openai import OpenAI import time # ================= 配置区域 ================= API_KEY = "token-abc123" BASE_URL = "http://localhost:1054/v1" MODEL_NAME = "/mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen2.5-72B-Instruct" client = OpenAI(api_key=API_KEY, base_url=BASE_URL) while True: time.sleep(3) response = client.chat.completions.create( model=MODEL_NAME, messages=[ {"role": "user", "content": "你好"} ], temperature=0.7, ) CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/rl/search_words/step_10 --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-VL-32B-Instruct --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 32000 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/code/summary/v2dc_proj/OUTPUTS/llama_factory/qwen3/qwen3-32B/ai_search_summary_redone_32B_1020_20k_ssp15_ssp16_gemini_merged --port 1054 --dtype bfloat16 --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --enable-lora --lora-modules lora1=/mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary_merged --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary_style_ench_merged --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary_compliance_merged --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary_style_ench_wo_shortanswer --port 1055 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/step1/v1_merged --port 1054 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/step1/v2_merged --port 1055 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/ZhipuAI/GLM-4.6V --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/ai_search_summary_style_ench_wo_shortanswer_style_search_words_merged --port 1054 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/step2/v4_merged --port 1054 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code CUDA_VISIBLE_DEVICES=0,1,2,3 vllm serve /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/step2/v5_merged --port 1054 --dtype auto --tensor-parallel-size 4 --gpu-memory-utilization 0.9 --max-model-len 16384 --api-key token-abc123 --trust-remote-code # export HF_ENDPOINT="https://hf-mirror.com" # HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download openai/gpt-oss-120b --local-dir /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-235B-A22B-Instruct-2507 huggingface-cli download Qwen/Qwen3.5-397B-A17B --local-dir /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3.5-397B-A17B git clone https://oauth2:ms-9f8d516d-837f-4682-bf34-f6f8f481b951@www.modelscope.cn/datasets/qsq12138/insert_image.git llamafactory-cli train ... 2>&1 | tee xxx.txt CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-235B-A22B-Instruct-2507 --port 1054 --dtype auto --tensor-parallel-size 8 --gpu-memory-utilization 0.9 --max-model-len 8192 --api-key token-abc123 --trust-remote-code # 下载remote tunnels curl -Lk 'https://code.visualstudio.com/sha/download?build=stable&os=cli-alpine-x64' --output vscode_cli.tar.gz tar -xf vscode_cli.tar.gz # 代理地址 export https_proxy=10.140.24.177:3128 export https_proxy=10.140.15.68:3128 export http_proxy="http://10.140.15.68:3128" export https_proxy="http://10.140.15.68:3128" export HTTP_PROXY="$http_proxy" export HTTPS_PROXY="$https_proxy" ./code tunnel user logout ./code tunnel unregister ./code tunnel --accept-server-license-terms --name c1 # torch多机多卡分布式训练 FORCE_TORCHRUN=1 torchrun --nnodes=2 --nproc_per_node=1 --node_rank=0 --master_addr=10.148.5.59 --master_port=29600 src/llamafactory/launcher.py torchrun_test.yaml FORCE_TORCHRUN=1 torchrun --nnodes=2 --nproc_per_node=1 --node_rank=1 --master_addr=10.148.5.59 --master_port=29600 src/llamafactory/launcher.py torchrun_test.yaml # ray + vllm多机多卡部署 在master执行： ray stop --force || true ray start --head \ --node-ip-address=10.148.26.191 \ --port=6379 \ --dashboard-host=0.0.0.0 # 在master执行 ray stop --force || true ray start --head \ --node-ip-address=10.148.26.191 \ --port=6379 \ --dashboard-host=0.0.0.0 在worker执行： ray stop --force || true ray start --address=10.148.26.191:6379 \ --node-ip-address=10.148.26.190 在worker执行： ray stop --force || true ray start --address=10.148.1.31:6379 \ --node-ip-address=10.148.1.28 确认ray资源： ray status 在主节点执行： vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/deepseek-ai/DeepSeek-V3.2 \ --served-model-name deepseek-v3.2 \ --host 0.0.0.0 \ --port 1054 \ --dtype auto \ --tensor-parallel-size 8 \ --pipeline-parallel-size 2 \ --tokenizer-mode deepseek_v32 \ --reasoning-parser deepseek_v3 \ --distributed-executor-backend ray \ --gpu-memory-utilization 0.8 \ --max-model-len 8192 \ --api-key token-abc123 \ --trust-remote-code vllm serve /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-4B-Instruct-2507/ \ --served-model-name qwen \ --host 0.0.0.0 \ --port 1054 \ --dtype auto \ --tensor-parallel-size 2 \ --distributed-executor-backend ray \ --gpu-memory-utilization 0.9 \ --max-model-len 32000 \ --api-key token-abc123 \ --trust-remote-code python3 -m model_tools push --model_name glm-4.6v-qsq --model_path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/ZhipuAI/GLM-4.6V --model_format hf --precision bf16 --region tencent-ap-shanghai --user qinshengqian --token QST9280db23590123fbe42de551de0971ce python3 -m model_tools push --model_name glm-4.6v-qsq --model_path 模型目录 --model_format 模型格式 --precision 模型精度 --region tencent-ap-shanghai --user qinshengqian --token QST9280db23590123fbe42de551de0971ce curl http://10.148.28.162:1054/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer token-abc123" \ -d '{ "model": "deepseek-v3.2", "messages": [ {"role": "user", "content": "你好"} ], "temperature": 0, "max_tokens": 512 }' wget https://repo.anaconda.com/archive/Anaconda3-2024.02-1-Linux-x86_64.sh python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen2.5-32B-Instruct \ --served-model-name qwen \ --host 0.0.0.0 \ --port 1054 \ --dtype auto \ --pipeline-parallel-size 2 \ --context-length 32000 \ --api-key token-abc123 \ --mem-fraction-static 0.8 \ --trust-remote-code \ --nnodes 2 \ --node-rank 0 \ --dist-init-addr 10.148.25.165:29501 python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen2.5-32B-Instruct \ --served-model-name qwen \ --dtype auto \ --pipeline-parallel-size 2 \ --context-length 32000 \ --api-key token-abc123 \ --mem-fraction-static 0.8 \ --trust-remote-code \ --nnodes 2 \ --node-rank 1 \ --dist-init-addr 10.148.25.165:29501 python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/deepseek-ai/DeepSeek-V3.2 \ --served-model-name deepseek-v3.2 \ --host 0.0.0.0 \ --port 1054 \ --dtype auto \ --tensor-parallel-size 8 \ --pipeline-parallel-size 2 \ --context-length 8192 \ --api-key token-abc123 \ --mem-fraction-static 0.8 \ --trust-remote-code \ --nnodes 2 \ --node-rank 0 \ --dist-init-addr 10.148.28.162:6379 python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/deepseek-ai/DeepSeek-V3.2 \ --served-model-name deepseek-v3.2 \ --dtype auto \ --tensor-parallel-size 8 \ --pipeline-parallel-size 2 \ --context-length 8192 \ --api-key token-abc123 \ --mem-fraction-static 0.8 \ --trust-remote-code \ --nnodes 2 \ --node-rank 1 \ --dist-init-addr 10.148.28.162:6379 # 在节点0上 CUDA_VISIBLE_DEVICES=0,1,2,3 \ python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-4B-Instruct-2507 \ --served-model-name qwen \ --host 0.0.0.0 \ --port 1054 \ --dtype auto \ --tensor-parallel-size 4 \ --context-length 32000 \ --api-key token-abc123 \ --mem-fraction-static 0.9 \ --trust-remote-code \ --nnodes 2 \ --node-rank 0 \ --dist-init-addr 10.148.11.239:6379 # 在节点1上 CUDA_VISIBLE_DEVICES=0,1,2,3 \ python -m sglang.launch_server \ --model-path /mnt/tidal-alsh01/usr/dawo/dawo/model_ori/Qwen/Qwen3-4B-Instruct-2507 \ --served-model-name qwen \ --host 0.0.0.0 \ --port 1055 \ --dtype auto \ --tensor-parallel-size 4 \ --pipeline-parallel-size 2 \ --context-length 32000 \ --api-key token-abc123 \ --mem-fraction-static 0.9 \ --trust-remote-code \ --nnodes 2 \ --node-rank 1 \ --dist-init-addr 10.148.11.239:6379 export ANTHROPIC_BASE_URL=http://10.204.73.145:1054 export ANTHROPIC_AUTH_TOKEN=token-abc123 export ANTHROPIC_MODEL=deepseek-v3.2 export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRIFFIC=1 export OPENAI_BASE_URL="http://10.204.73.145:1054/v1" export OPENAI_API_KEY="token-abc123" export OPENAI_MODEL="deepseek-v3.2" python3 -m model_tools push --model_name oversea-summary-multimodal --model_path /mnt/tidal-alsh01/usr/dawo/qinshengqian/output/llama_factory/qwen3/qwen3-32B/round2/step2/v5_merged --model_format hf --precision bf16 --region tencent-ap-shanghai --user qinshengqian --token QST9280db23590123fbe42de551de0971ce huggingface-cli download mPLUG/GUI-Owl-1.5-32B-Instruct --local-dir /mnt/tidal-alsh01/usr/qinshengqian/ckpt/mPLUG/GUI-Owl-1.5-32B-Instruct huggingface-cli download mPLUG/GUI-Owl-1.5-32B-Instruct --local-dir /mnt/tidal-alsh01/usr/qinshengqian/ckpt/mPLUG/GUI-Owl-1.5-32B-Thinking huggingface-cli upload-large-folder Shengqian12138/c1 ./claude_test --repo-type model