Downloads · 30 days
0
agurung/colar-gemma-3-4b-ff-rl-1b3gv8q3-ckpt
colar-gemma-3-4b-ff-rl-1b3gv8q3-ckpt is a machine learning model from agurung. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
This is a raw PyTorch-Lightning training checkpoint (last.ckpt), archived as a byte-for-byte backup. It is not the clean hf + extrastate.pt export layout.
Downloads · 30 days
0
Access
Public
Updated Jun 28, 2026
Repo size
24.4 GB
Likes
0
Public
Click a slice to open those files.
.ckpt24.4 GB · 100%
From the Hugging Face model README
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.
google/gemma-3-4b-itLitCoLaR), config colar_ff_gemma3_rl (active_embed_std variant; dir colar_ff_gemma3_rl_active_embed_std)1b3gv8q3 (epoch 10, global_step 6376, PL 2.6.1)/mnt/disk/baseline_colar/colar_hqaakpve_exact/logs/colar_ff_gemma3_rl_active_embed_std/colar-experiments/1b3gv8q3/checkpoints/last.ckptlast.ckpt sha256: 36b7036bbd31b951078ff55ddb0f4eb0a15dfa9b0bdd93af46d1544dfd12f61fThe state_dict contains the full LLM backbone and the CoLaR latent extra
parameters (latent_policy.fc/mean/log_std, hidden=2560), plus optimizer /
trainer state. See backup_metadata.json.
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_ff_gemma3_rl --dataset <dataset_cfg> --trainer default
Archived 2026-06-28 because no prior Hub backup existed for this run.