default HF jobs to amp off

length-bucketed training and HF bootstrap

disable full embedding export in hf train jobs

support hf jobs simulator corpus launch

document pinned split revision for hf jobs

pin split revision for lmdb cache key

fix remote train processed cache path

document exact hf jobs api launch

sync current ml-intern source for hf jobs

Update scripts/remote_train.sh from ml-intern branch

Upload scripts/analyze_class_dist_v2.py

Upload scripts/train_real_te_v3.py

Upload scripts/train_real_te_v2.py

Upload scripts/analyze_class_dist.py

Upload scripts/train_real_te_standalone.py

Upload scripts/train_real_te_standalone.py

Upload scripts/run_real_te_patched.py

Delete scripts/train_real_te_standalone.py

Upload scripts/train_real_te_standalone.py

Upload scripts/train_real_te_standalone.py

Upload scripts/hf_job_real_te.py

Upload scripts/pipeline.py

Upload scripts/generate_synthetic.py

Upload scripts/etl_pipeline.py

Upload scripts/train.py