You can deploy this model and start earning money today!

agurung/
colar-gemma-3-1b-gsm-hard-rl-pfly40to-ckpt

Name: colar-gemma-3-1b-gsm-hard-rl-pfly40to-ckpt
Availability: InStock

Views

No views yet

CoLaR google/gemma-3-1b-it GSM-Hard post-RL — raw checkpoint archive

This is a raw PyTorch-Lightning training checkpoint (last.ckpt), archived as a byte-for-byte backup. It is not the clean hf + extra_state.pt export layout.

Base model: google/gemma-3-1b-it
Method: CoLaR (LitCoLaR), config colar_gsm_hard_gemma3_1b_rl
Task / stage: GSM-Hard / post-RL
Source run id: pfly40to (epoch 0, global_step 768, PL 2.6.1)
Original path: /mnt/disk/baseline_colar/colar_hqaakpve_exact/logs/colar_gsm_hard_gemma3_1b_rl/colar-experiments/pfly40to/checkpoints/last.ckpt
last.ckpt sha256: 90736df687416f95d9511d66a7408209074751da6a23a50771407de1028a32f0

Contents (verified before upload)

The state_dict contains the full LLM backbone and the CoLaR latent extra parameters (latent_policy.fc/mean/log_std, hidden=1152), plus optimizer / trainer state. See backup_metadata.json.

Why a raw ckpt

The archival host had no GPU and only 2.6 GB RAM, so the standard exporter (colar/convert_checkpoint_to_model.py, which loads the model) could not run. To produce the standard HF model + extra_state.pt, re-run that exporter on a GPU host:

python colar/convert_checkpoint_to_model.py --ckpt last.ckpt --out <dir> --format hf \
  --model colar_gsm_hard_gemma3_1b_rl --dataset <dataset_cfg> --trainer default

Archived 2026-06-28 because no prior Hub backup existed for this run.

agurung/colar-gemma-3-1b-gsm-hard-rl-pfly40to-ckpt

CoLaR google/gemma-3-1b-it GSM-Hard post-RL — raw checkpoint archive

Contents (verified before upload)

Why a raw ckpt

agurung/
colar-gemma-3-1b-gsm-hard-rl-pfly40to-ckpt