chore: import upstream snapshot with attribution

2026-07-13 13:34:58 +08:00
commit a203934033
1368 changed files with 175001 additions and 0 deletions
@@ -0,0 +1,44 @@
+# Internal vLLM
+
+# pip install math_verify # reward function
+# pip install -U trl
+# note: Note: The parameters of each node need to be consistent.
+export CUDA_VISIBLE_DEVICES=0,1,2,3
+export NNODES=2
+export NODE_RANK=0
+export MASTER_ADDR=127.0.0.1
+export MASTER_PORT=29500
+export NPROC_PER_NODE=4
+
+swift rlhf \
+    --rlhf_type grpo \
+    --model Qwen/Qwen2.5-Math-7B \
+    --reward_funcs accuracy format \
+    --use_vllm true \
+    --vllm_mode colocate \
+    --vllm_gpu_memory_utilization 0.5 \
+    --vllm_max_model_len 4096 \
+    --tuner_type full \
+    --torch_dtype bfloat16 \
+    --dataset 'AI-MO/NuminaMath-TIR#5000' \
+    --load_from_cache_file true \
+    --max_completion_length 2048 \
+    --num_train_epochs 1 \
+    --per_device_train_batch_size 1 \
+    --per_device_eval_batch_size 1 \
+    --learning_rate 1e-6 \
+    --gradient_accumulation_steps 2 \
+    --eval_steps 200 \
+    --save_steps 200 \
+    --save_total_limit 2 \
+    --logging_steps 5 \
+    --max_length 4096 \
+    --output_dir output \
+    --warmup_ratio 0.05 \
+    --dataloader_num_workers 4 \
+    --dataset_num_proc 4 \
+    --num_generations 8 \
+    --temperature 0.9 \
+    --system 'examples/train/grpo/prompt.txt' \
+    --deepspeed zero2 \
+    --log_completions true