Fetching metadata from the HF Docker repository...
-
update: results table, 0.5B model links, citation year 2026
Jayant-Kernel
-
docs: detailed README with curriculum, reward table, results, usage
Jayant-Kernel
-
rollback: revert to last working Dockerfile and train.py
Jayant-Kernel
-
fix: proper GRPO with trl 0.12.2 no-deps + force hub downgrade
Jayant-Kernel
-
fix: custom training loop without TRL dependency
Jayant-Kernel
-
fix: force reinstall huggingface_hub 0.24.7 after deceit_env
Jayant-Kernel
-
fix: pin huggingface_hub 0.24.7, install trl with --no-deps
Jayant-Kernel
-
fix: trl 0.12.2 has GRPOTrainer, pin all deps before trl install
Jayant-Kernel
-
fix: try multiple import paths for GRPOConfig
Jayant-Kernel
-
fix: install transformers 4.46.0 BEFORE trl so trl doesnt upgrade it
Jayant-Kernel
-
fix: bust docker cache force reinstall trl 0.11.4
Jayant-Kernel
-
fix: trl 0.11.4 + transformers 4.46.0 + processing_class
Jayant-Kernel
-
fix: trl 0.9.4 + transformers 4.41.2 compatible versions
Jayant-Kernel
-
fix: add torch version check in Dockerfile
Jayant-Kernel
-
fix: remove tokenizer arg from GRPOTrainer
Jayant-Kernel
-
fix: tokenizer not processing_class, torch cu121 for GPU
Jayant-Kernel
-
fix: correct trl version with GRPOConfig
Jayant-Kernel
-
fix: trl 0.13.0, remove verify steps
Jayant-Kernel
-
fix: cu124 not cu118 for A100 CUDA 12.9 driver
Jayant-Kernel
-
fix: trl 0.12.2 + torch 2.4.0
Jayant-Kernel
-
fix: CPU fallback when no GPU detected
Jayant-Kernel
-
fix: trl 0.15.0 definitely has GRPOConfig
Jayant-Kernel
-
fix: trl 0.9.6 + bitsandbytes 0.43.1 cu118
Jayant-Kernel
-
fix: trl 0.8.6 has GRPOConfig, compatible with torch 2.1.2
Jayant-Kernel
-
fix: remove multiline python heredoc from Dockerfile
Jayant-Kernel
-
fix: find deceit_env package location and copy data correctly
Jayant-Kernel
-
fix: back to python:3.10-slim for GPU, fix deceit_env path
Jayant-Kernel
-
fix: deceit_env module path and PYTHONPATH
Jayant-Kernel
-
fix: nvidia cuda base with python3.10 installed
Jayant-Kernel
-
fix: use huggingface transformers-pytorch-gpu base image
Jayant-Kernel
-
fix: revert to torch 2.1.0 cu121 with trl 0.7.4 - versions that worked before
Jayant-Kernel
-
fix: simplify dockerfile no version pinning
Jayant-Kernel
-
fix: accelerate 0.34.2 exists, 0.35.0 does not
Jayant-Kernel
-
fix: trl 0.12.0 has GRPOTrainer, compatible with torch 2.4.0
Jayant-Kernel
-
fix: pin trl 0.11.0 compatible with torch 2.4.0
Jayant-Kernel
-
fix: upgrade torch to 2.4.0 with CUDA 12.4 support
Jayant-Kernel
-
fix: run train.py instead of evaluate.py
Jayant-Kernel
-
improve: abstention penalty, better prompt, mixed curriculum, more steps
Jayant-Kernel
-
evaluate: switch to 0.5B model comparison, 200 episodes
Jayant-Kernel
-
fix: set N_EPISODES=200 constant (was still 30)
Jayant-Kernel
-
update: increase evaluation to 200 episodes per model
Jayant-Kernel
-
fix: parse_action confidence bug, numeric answers bug, missing reasoning field bug
Jayant-Kernel
-
fix: debug model output parsing in evaluation
Jayant-Kernel
-
add: evaluate 1.5B base vs trained, upload chart to HF Hub
Jayant-Kernel
-
update: 500 steps L1 + 300 steps L2, higher lr for 1.5B
Jayant-Kernel
-
fix: add matplotlib, split COPY into separate lines
Jayant-Kernel
-
fix: copy evaluate.py into Docker image
Jayant-Kernel
-
add: evaluate 1.5B base vs trained, upload charts
Jayant-Kernel
-
fix: remove misplaced import inside GRPOConfig args
Jayant-Kernel
-
fix: auto-detect bf16 support
Jayant-Kernel