d12 RL GRPO on GSM8K (1 epoch, skip-dryrun)

Cancelled
UUID:6766f2e9-2e8d-4df5-8933-da1a0797c883
Tao Linvia🐎 pentiumCreated 207 days agoUpdated 207 days ago

RL (GRPO) on GSM8K (skip-dryrun)

Source: SFT checkpoint (MLflow 2bffbfabb3024458ae2778a12f320163)

Config: upstream defaults — 1 epoch, 467 steps, GRPO with DAPO normalization

Using skip-dryrun because RL sampling exceeds 900s dryrun timeout. max_time=1800s.

Status: awaiting funding

Collapsed
1
Dryrun JobSkipped
—
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=1·save_every=60·tok_max_chars=2000000000·device_batch_size=8·examples_per_step=16·sft_mlflow_run_id=2bffbfabb3024458ae2778a12f320163·upload_checkpoint=1
Max Time
0.5h(dryrun skipped)