184 kB

Ctrl+K

1 contributor

History: 32 commits

nraptisss

Add nohup launcher for focused 4-layer training

7559485 verified 13 days ago

baseline_eval.py

13 kB
Add baseline evaluation script for Llama/GPT-4o-mini comparison 23 days ago
build_weak_layer_dataset.py

4.71 kB
Add weak-layer stage2 dataset builder 26 days ago
check_gpu.py

1.44 kB
Ensure CUDA GPU preflight and RTX 6000 Ada install path about 1 month ago
evaluate_grpo_correct.py

7.69 kB
Add correct GRPO evaluation script that loads SFT-merged base + GRPO adapter 17 days ago
evaluate_model.py

10.1 kB
Speed up and resume OOD evaluation with batched dynamic generation 27 days ago
evaluate_semantic_o1_a1.py

10.1 kB
Fix semantic evaluator to recover metadata by prediction id 23 days ago
install_rtx6000ada.sh

1.33 kB
Ensure CUDA GPU preflight and RTX 6000 Ada install path about 1 month ago
merge_adapter.py

1.36 kB
Add RTX 6000 Ada QLoRA training and evaluation repo about 1 month ago
nohup_eval.sh

2.28 kB
Fix nohup evaluator to support merged model paths 26 days ago
nohup_focused.sh

1.19 kB
Add nohup launcher for focused 4-layer training 13 days ago
nohup_grpo.sh

1.71 kB
Add GRPO nohup launch script for RTX 6000 Ada / A100 22 days ago
nohup_new_run.sh

2.5 kB
Ensure CUDA GPU preflight and RTX 6000 Ada install path about 1 month ago
nohup_resume.sh

1.74 kB
Ensure CUDA GPU preflight and RTX 6000 Ada install path about 1 month ago
nohup_stage2_weak.sh

3.02 kB
Add stage2 weak-layer nohup runner 26 days ago
normalize_eval_metrics.py

6.38 kB
Add normalized evaluator for existing predictions 27 days ago
package_results.py

5.7 kB
Add results packaging script 25 days ago
reproduce_stage1_eval.sh

2.35 kB
Add stage1 evaluation reproduction script 24 days ago
run_all_baselines.sh

6.42 kB
Upload scripts/run_all_baselines.sh 23 days ago
run_rtx6000ada.sh

934 Bytes
Add RTX 6000 Ada QLoRA training and evaluation repo about 1 month ago
run_zero_shot_baseline.sh

1.53 kB
Add sampled zero-shot baseline runner 24 days ago
sample_failure_examples.py

6.68 kB
Add qualitative failure example sampler 25 days ago
status_run.sh

1.06 kB
Add nohup run management and resumable checkpoint support about 1 month ago
train_continue_adapter.py

8.75 kB
Fix stage2 push_to_hub dataset metadata validation 26 days ago
train_focused.py

6.57 kB
Add focused 4-layer SFT script: train only on tmf921/camara/3gpp/etsi_zsm (removes weak O1/A1/lifecycle layers) 13 days ago
train_grpo.py

17 kB
Tune GRPO hyperparams for RTX 6000 Ada 48GB: reduce steps/grad_accum/completion_length for ~6h runtime, prevent OOM 19 days ago
train_grpo_v2.py

16.2 kB
Fix GRPO v2: lower temp=0.3, dense reward shaping, higher beta=0.1 to stay near SFT, G=2 safe for 48GB 19 days ago
train_grpo_v3.py

15.1 kB
GRPO v3: fix truncation — use full 1536 completion length + G=4, we have 42GB headroom 18 days ago
train_qlora.py

10.8 kB
Harden Trackio Space validation to avoid startup crash about 1 month ago
train_rft.py

16.5 kB
Fix RFT: batch generation, reduce to 200 prompts x 8 samples (~24h feasible on RTX 6000 Ada) 16 days ago