Spaces:

anugrah55
/

opensleuth-training-gemini-cli

Paused

opensleuth-training-gemini-cli / opensleuth_train

Ctrl+K

1 contributor

trainer v0.4: switch to Qwen2.5-3B-Instruct, dynamic task discovery, delegated probe sampling, difficulty-weighted rollouts, push to opensleuth-qwen2.5-3b-grpo-v2; sentinel cleared on FORCE_TRAIN=1.

78575eb verified 12 days ago

__init__.py

512 Bytes
trainer v0.4: switch to Qwen2.5-3B-Instruct, dynamic task discovery, delegated probe sampling, difficulty-weighted rollouts, push to opensleuth-qwen2.5-3b-grpo-v2; sentinel cleared on FORCE_TRAIN=1. 12 days ago
client.py

4.52 kB
trainer v0.4: switch to Qwen2.5-3B-Instruct, dynamic task discovery, delegated probe sampling, difficulty-weighted rollouts, push to opensleuth-qwen2.5-3b-grpo-v2; sentinel cleared on FORCE_TRAIN=1. 12 days ago
dataset.py

10.3 kB
trainer v0.4: switch to Qwen2.5-3B-Instruct, dynamic task discovery, delegated probe sampling, difficulty-weighted rollouts, push to opensleuth-qwen2.5-3b-grpo-v2; sentinel cleared on FORCE_TRAIN=1. 12 days ago
prompt.py

2.22 kB
Overhaul trainer: TRL GRPO with env-backed reward, Qwen2.5-0.5B 4bit+LoRA, slim PyTorch CUDA base, heartbeat HTTP for HF Spaces health probe 13 days ago
reward.py

3.37 kB
Overhaul trainer: TRL GRPO with env-backed reward, Qwen2.5-0.5B 4bit+LoRA, slim PyTorch CUDA base, heartbeat HTTP for HF Spaces health probe 13 days ago