Skip to content

Distributed RL lacked a turnkey example reproducing the 8K-generation OpenMathInstruct-2 GRPO math setup with MaxText training and disaggregated vLLM rollouts. - #3448

Open
copybara-service[bot] wants to merge 1 commit into
mainfrom
test_1026002365

Conversation

@copybara-service

Copy link
Copy Markdown

Distributed RL lacked a turnkey example reproducing the 8K-generation OpenMathInstruct-2 GRPO math setup with MaxText training and disaggregated vLLM rollouts.

Add the openmath_dist example (openmath.py, openmath.yaml, launcher.sh, k8s_launcher.sh) and 64-chip GKE recipes for TPU v5p (v5p_openmath_qwen3_0p6b.sh) and Ironwood TPU v7x (v7x_openmath_qwen3_0p6b.sh). Prompts render <reasoning> and <answer> instructions through the tokenizer chat template after deterministic shuffling and 4096-token length filtering. Completions score exact tag formatting and symbolic LaTeX equivalence in persistent worker subprocesses with per-call timeouts. Timed-out workers are terminated and replaced so hung sympy expressions cannot stall rollouts.

Configure strict on-policy GRPO (off_policy_steps=0, beta=0.0, epsilon=0.2) with FP32 master weights, a 20-step cosine schedule, and Raiden weight sync from Pathways trainers to 12 disaggregated vLLM rollout replicas:

  • v5p_openmath_qwen3_0p6b.sh: tpuv5:2x2x4 trainer ((16,1,1) mesh, 12K-token sequence packing, 512-token Splash Attention tiles) paired with 12 tpuv5:2x2x1 rollout replicas ((4,1) mesh, round-robin DP routing).
  • v7x_openmath_qwen3_0p6b.sh: tpu7x:2x2x4 trainer ((32,1,1) mesh, 17408-token sequence packing, Tokamax Splash Attention with 1024-token tiles and diagonal mask skip, remat_policy="none") paired with 12 tpu7x:2x2x1 rollout replicas (intra-chip TP=2, DP=4 (4,2) mesh, batched RPA kernel, parallel AOT precompilation, top-k prefiltering, and least-loaded DP routing).

… `OpenMathInstruct-2` GRPO math setup with MaxText training and disaggregated vLLM rollouts.

Add the `openmath_dist` example (`openmath.py`, `openmath.yaml`, `launcher.sh`, `k8s_launcher.sh`) and 64-chip GKE recipes for TPU v5p (`v5p_openmath_qwen3_0p6b.sh`) and Ironwood TPU v7x (`v7x_openmath_qwen3_0p6b.sh`). Prompts render `<reasoning>` and `<answer>` instructions through the tokenizer chat template after deterministic shuffling and 4096-token length filtering. Completions score exact tag formatting and symbolic LaTeX equivalence in persistent worker subprocesses with per-call timeouts. Timed-out workers are terminated and replaced so hung `sympy` expressions cannot stall rollouts.

Configure strict on-policy GRPO (`off_policy_steps=0`, `beta=0.0`, `epsilon=0.2`) with FP32 master weights, a 20-step cosine schedule, and Raiden weight sync from Pathways trainers to 12 disaggregated vLLM rollout replicas:
- `v5p_openmath_qwen3_0p6b.sh`: `tpuv5:2x2x4` trainer (`(16,1,1)` mesh, 12K-token sequence packing, 512-token Splash Attention tiles) paired with 12 `tpuv5:2x2x1` rollout replicas (`(4,1)` mesh, round-robin DP routing).
- `v7x_openmath_qwen3_0p6b.sh`: `tpu7x:2x2x4` trainer (`(32,1,1)` mesh, 17408-token sequence packing, Tokamax Splash Attention with 1024-token tiles and diagonal mask skip, `remat_policy="none"`) paired with 12 `tpu7x:2x2x1` rollout replicas (intra-chip `TP=2, DP=4` `(4,2)` mesh, batched RPA kernel, parallel AOT precompilation, top-k prefiltering, and least-loaded DP routing).

PiperOrigin-RevId: 1026002365

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant