final-iteration / training /hf_run_space_train_job.sh

Commit History

hf jobs: propagate TEST_ONLY env through launcher + remote runner
95d0045

anuragredbus commited on

training: use -- and bash -c to bypass hf CLI typer flag stealing
bcc27a5

anuragredbus commited on

training: default Space to ycwhencpp/final-iteration
6279175

anuragredbus commited on

training: split remote runner into in-repo script
0980a17

anuragredbus commited on

training: default flavor a10g-largex4 (4xA10G, 96GB VRAM)
ef79012

vaibhav12332112312 commited on

training: smoke-mode + hardcoded peak hint + valid tool IDs
1f72457

vaibhav12332112312 commited on

train: default HF Job flavor l4x1 -> l40sx1 (48GB VRAM)
76b19bd

vaibhav12332112312 commited on

train: per-step credit + drop replies + larger batches
9ee7a09

vaibhav12332112312 commited on