Model card — step 30000, PPL 3.65

Remove stale root-level training_args.bin (was from step 2000)

Remove stale root-level trainer_state.json (was from step 2000)

Remove stale root-level rng_state.pth (was from step 2000)

Remove stale root-level scheduler.pt (was from step 2000)

Remove stale root-level optimizer.pt (was from step 2000)

Merge step-30000 weights to repo root

Merge step-30000 weights to repo root

Training in progress, step 30000, checkpoint

Training in progress, step 30000

Training in progress, step 29500, checkpoint

Training in progress, step 29500

Training in progress, step 29000, checkpoint

Training in progress, step 29000

Training in progress, step 28500, checkpoint

Training in progress, step 28500

Training in progress, step 28000, checkpoint

Training in progress, step 28000

Training in progress, step 27500, checkpoint

Training in progress, step 27500

Training in progress, step 27000, checkpoint

Training in progress, step 27000

Training in progress, step 26500, checkpoint

Training in progress, step 26500

Training in progress, step 26000, checkpoint

Training in progress, step 26000

Training in progress, step 25500, checkpoint

Training in progress, step 25500

Training in progress, step 25000, checkpoint

Training in progress, step 25000

Training in progress, step 24500, checkpoint

Training in progress, step 24500

Training in progress, step 24000, checkpoint

Training in progress, step 24000

Training in progress, step 23500, checkpoint

Training in progress, step 23500

Training in progress, step 23000, checkpoint

Training in progress, step 23000

Training in progress, step 22500, checkpoint

Training in progress, step 22500

Training in progress, step 22000, checkpoint

Training in progress, step 22000

Training in progress, step 21500, checkpoint

Training in progress, step 21500

Training in progress, step 21000, checkpoint

Training in progress, step 21000

Training in progress, step 20500, checkpoint

Training in progress, step 20500

Training in progress, step 20000, checkpoint

Training in progress, step 20000