# Reproducibility

Reproducibility is a bedrock of scientific progress. By combining vLLM's [batch-invariant deterministic inference](https://vllm.ai/blog/2025-11-10-bitwise-consistent-train-inference) with Megatron-LM's deterministic mode, vime supports bitwise experiment reproduction.

To enable deterministic training, you need to first uninstall the flash attention 3 in the docker with `pip uninstall flash_attn_3 -y` and set:
```bash
  # vLLM config
  --vllm-enable-deterministic-inference
  --vllm-attention-backend flashinfer

  # megatron config
  --deterministic-mode
```

And set the following environment variables:

```bash
     "env_vars": {
        ...,
        "NCCL_ALGO": "Ring",
        "NVTE_ALLOW_NONDETERMINISTIC_ALGO": "0",
        "CUBLAS_WORKSPACE_CONFIG": ":4096:8"
     }
```

Here we provide the script to do RL training on Qwen2.5 0.5B model and GSM8K dataset with full deterministic.

For data and checkpoint preparation, please run:

```bash
# download
hf download --repo-type dataset zhuzilin/gsm8k --local-dir /root/gsm8k
hf download Qwen/Qwen2.5-0.5B-Instruct --local-dir /root/Qwen2.5-0.5B-Instruct

# convert ckpt
cd vime/
source scripts/models/qwen2.5-0.5B.sh
PYTHONPATH=/root/Megatron-LM/ python \
   tools/convert_hf_to_torch_dist.py \
   ${MODEL_ARGS[@]} \
   --hf-checkpoint /root/Qwen2.5-0.5B-Instruct \
   --save /root/Qwen2.5-0.5B-Instruct_torch_dist/
```

And to run training,

```bash
bash scripts/run-qwen2.5-0.5B-reproducibility.sh
```

For screen shots of the wandb, please refer to [pull#370](https://github.com/THUDM/slime/pull/370).