Spaces:

ianalin123
/

origami_env

Sleeping

App Files Files Community

origami_env / training

80 kB

Ctrl+K

Ctrl+K

4 contributors

History: 28 commits

ianalin123's picture

feat(v3): hybrid GRPO + SFT training with expert beam search

4d27d34 5 months ago

Dockerfile.train

855 Bytes
feat: Railway deployment + multi-task GRPO + Modal B200 training 5 months ago
__init__.py

0 Bytes
Origami RL environment with Three.js viewer 5 months ago
curriculum.py

574 Bytes
fix(v3): include multi-step tasks from curriculum start 5 months ago
env_pool.py

1.09 kB
feat(v3): implement core building blocks for multi-step RL training 5 months ago
expert_search.py

5.43 kB
feat(v3): hybrid GRPO + SFT training with expert beam search 5 months ago
gigpo.py

3.76 kB
feat(v3): implement core building blocks for multi-step RL training 5 months ago
prompt_builder.py

3.26 kB
fix(v3): prompt perturbation to break deterministic outputs 5 months ago
reward.py

9.33 kB
feat(v2): add extract_crease_json and valid_crease reward to training/reward.py 5 months ago
rollout.py

3.62 kB
fix(v3): guarantee reward variance with num_per_task and higher temp 5 months ago
train_grpo.py

12.4 kB
feat(v2): update train_grpo.py for step-level prompts and per_step_reward 5 months ago
train_origami.ipynb

19.8 kB
revert: restore train_origami.ipynb to Prasanna's exact version 5 months ago
train_v3.py

19.3 kB
feat(v3): hybrid GRPO + SFT training with expert beam search 5 months ago
trajectory.py

596 Bytes
feat(v3): implement core building blocks for multi-step RL training 5 months ago