Spaces:

TheCrustaceans
/

Pm-ops

Sleeping

App Files Files Community

Pm-ops / training

460 kB

Ctrl+K

Ctrl+K

4 contributors

History: 23 commits

SavK1's picture

fix(v4): increase T4 SFT to 50 eps, add action trace to eval output

16705aa 4 months ago

Competent Pink Train (1).ipynb
257 kB
fix(notebook): rewrite run_grpo_episode with runbook-compliance reward 4 months ago
__init__.py
0 Bytes
writing the GRPO script to train qwen1.7 on A100 GRPO 4 months ago
dataset.py
3.77 kB
fix(grpo): fresh reward design — runbook-compliance scoring 4 months ago
pm_ops_trainer.py
8.26 kB
fix(grpo): eliminate zero-advantage collapse from stale rewards 4 months ago
prompts.py
4.37 kB
fix: format runbook as readable key-value pairs for model 4 months ago
rewards.py
4.18 kB
fix(grpo): fresh reward design — runbook-compliance scoring 4 months ago
rollout.py
16.8 kB
fix(grpo): fresh reward design — runbook-compliance scoring 4 months ago
smoke_test.py
11.5 kB
modifying the training script to be more memory efficient: Unsloth, also reduced the number of steps to 15 instead of 40 4 months ago
train.ipynb
22.5 kB
modifying the training script to be more memory efficient: Unsloth, also reduced the number of steps to 15 instead of 40 4 months ago
train_v2.ipynb
24.6 kB
feat(training): PMOpsGRPOTrainer — override _calculate_rewards to inject rollout rewards directly 4 months ago
train_v3.ipynb
37.9 kB
fix(grpo): fresh reward design — runbook-compliance scoring 4 months ago
train_v4.ipynb
37.1 kB
fix(v4): increase T4 SFT to 50 eps, add action trace to eval output 4 months ago
triage_dataset.jsonl
32.6 kB
writing the GRPO script to train qwen1.7 on A100 GRPO 4 months ago