mirror of
https://github.com/NousResearch/atropos.git
synced 2026-04-19 12:57:58 +00:00
major refactor 4
This commit is contained in:
parent
aecad35ceb
commit
40a0c47fc3
4 changed files with 18 additions and 9 deletions
|
|
@ -249,10 +249,11 @@ run-api --port 8001
|
|||
python -u environments/gsm8k_server.py serve \
|
||||
--env.tokenizer_name $MODEL \
|
||||
--env.use_wandb=False \
|
||||
--env.rollout_server_url http://localhost:8001 \
|
||||
--openai.model_name $MODEL \
|
||||
--openai.base_url http://localhost:9001/v1 \
|
||||
--openai.server_type vllm \
|
||||
--server.port 8001
|
||||
--slurm false
|
||||
|
||||
# Terminal 3: Trainer (manages its own vLLM)
|
||||
CUDA_VISIBLE_DEVICES=0,1 python -m example_trainer.grpo \
|
||||
|
|
@ -280,10 +281,11 @@ CUDA_VISIBLE_DEVICES=2 python example_trainer/vllm_api_server.py \
|
|||
python -u environments/gsm8k_server.py serve \
|
||||
--env.tokenizer_name $MODEL \
|
||||
--env.use_wandb=False \
|
||||
--env.rollout_server_url http://localhost:8002 \
|
||||
--openai.model_name $MODEL \
|
||||
--openai.base_url http://localhost:9002/v1 \
|
||||
--openai.server_type vllm \
|
||||
--server.port 8002
|
||||
--slurm false
|
||||
|
||||
# Terminal 7: Trainer (attaches to vLLM)
|
||||
CUDA_VISIBLE_DEVICES=2 python -m example_trainer.grpo \
|
||||
|
|
@ -312,10 +314,11 @@ CUDA_VISIBLE_DEVICES=4 python example_trainer/vllm_api_server.py \
|
|||
python -u environments/gsm8k_server.py serve \
|
||||
--env.tokenizer_name $MODEL \
|
||||
--env.use_wandb=False \
|
||||
--env.rollout_server_url http://localhost:8003 \
|
||||
--openai.model_name $MODEL \
|
||||
--openai.base_url http://localhost:9003/v1 \
|
||||
--openai.server_type vllm \
|
||||
--server.port 8003
|
||||
--slurm false
|
||||
|
||||
# Terminal 11: Trainer
|
||||
CUDA_VISIBLE_DEVICES=5 python -m example_trainer.grpo \
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue