d12 RL GRPO on GSM8K (1 epoch, ~467 steps)

Dryrun Job Failed
UUID:93cb4777-f771-4359-9a85-865a4f73be40
Tao Linvia🐎 pentiumCreated 209 days agoUpdated 209 days ago

RL (GRPO) on GSM8K

Source: SFT checkpoint from MLflow run 2bffbfabb3024458ae2778a12f320163

Config: upstream defaults — GSM8K (7473 train, 16 examples/step = 467 steps/epoch), GRPO with DAPO-style token-level normalization. num_epochs=2 for valid dryrun ratio; upstream default is 1.

Note: num_epochs set to 2 (dryrun=1) so MLPatron can estimate cost. If 1 epoch is enough (check reward plateau), can cancel early or reduce.

Status: dryrun pending

Collapsed
Dryrun JobFailed
2
Funding 
3
Funded Job 
4
Done 

Finance

Fundings at snapshot$0.00

Individual funding and transaction records are not included in the public archive.

Configuration

Repository
https://github.com/mlpatron/mlpatron-nanochat
Version
master → 65c2c6a930c7ad4c2898d249e36dbcad823463f8
Image
ghcr.io/mlpatron/mlpatron-nanochat:latest
→ ghcr.io/mlpatron/mlpatron-nanochat@sha256:a142e86efdcd41e0355e70d46c341cbf9a86b2da9448459009d7b5e8209ffa61
Entrypoint
h100_d12_chat_rl
Resources
NVIDIA H100 80GB × 1·a3-highgpu-1g (26 vCPU / 234 GiB)
Parameters
eval_every=60·num_epochs=2(dryrun: 1)·save_every=60·tok_max_chars=2000000000·device_batch_size=8·examples_per_step=16·sft_mlflow_run_id=2bffbfabb3024458ae2778a12f320163·upload_checkpoint=1