● Done. Pushed to https://github.com/Maya-AI-ML-Trainings-Cookbooks/maya-tts-llama32-1b-pretraining (private).

  - 143 files, clean single-commit history
  - .env included (with HF token, wandb key, R2 credentials)
  - All data files, checkpoints, outputs, wandb logs, llamafactory_ref excluded
  - Repo is private under the org

● Done. Private repo created and pushed:

  https://github.com/Maya-AI-ML-Trainings-Cookbooks/qwen3-asr-1.7b-phase2-sft

  - 51 files, fresh history (no old commits)
  - Includes .env with R2/wandb creds
  - Excludes: checkpoints, model weights, wandb logs, parquets, tar files, __pycache__
  - Private, under Maya-AI-ML-Trainings-Cookbooks org

● Done. Repo created and pushed:

  https://github.com/Maya-AI-ML-Trainings-Cookbooks/maya-asr-hybrid-fastconformer-rnnt-stage1

  - Private repo under Maya-AI-ML-Trainings-Cookbooks org
  - 80 files pushed (all code, configs, scripts, tests, .env)
  - Excluded: data/, artifacts/, experiments/, pretrained/, models/, tokenizers/ (200+ GB of large files)
  - Clean single commit, no prior history

● Done. Pushed to https://github.com/Maya-AI-ML-Trainings-Cookbooks/gemma3n-E2B-asr-ft (private).

  534 files, 64,740 lines. Includes:
  - All scripts, tools, configs, docs, benchmarks, artifacts (JSON/stats only)
  - Full LlamaFactory vendor with all patches (tar loader, FA2, buffer repair, collator)
  - No data files (parquets, tars, model weights, checkpoints, logs, venv excluded)