code-sprout-30m / status.json
Raymond1938-4's picture
Release experimental CodeSprout 30M checkpoint
61790a1 verified
Raw History Blame Contribute Delete
1.3 kB
{
"status": "completed",
"device": "NVIDIA GeForce RTX 3050 Laptop GPU",
"parameters": 29676544,
"config": {
"vocab_size": 8192,
"context": 512,
"width": 512,
"layers": 8,
"heads": 8,
"dropout": 0.1
},
"sequence": 256,
"microbatch": 4,
"accumulation": 4,
"starting_step": 3079,
"initial_validation_loss": {
"html": 2.1982025305430093,
"javascript": 4.3991193771362305,
"python": 5.0872650146484375,
"typescript": 3.6908501784006753
},
"wall_time_limit_seconds": 5960,
"training_examples": "language marker prepended to every sampled context",
"step": 6079,
"tokens_processed": 24899584,
"elapsed_seconds": 1683.8269999999902,
"validation_loss": {
"html": 0.3678888479868571,
"javascript": 1.280820369720459,
"python": 1.441411058108012,
"typescript": 0.8401213685671488
},
"mean_validation_loss": 0.9825604110956192,
"peak_vram_mb": 788.3,
"run_tokens_per_second": 27652.3,
"last_validation_loss": {
"html": 0.4167328824599584,
"javascript": 1.5458275477091472,
"python": 1.8703549702962239,
"typescript": 1.1160530249277751
},
"inference_checkpoint_step": 4324,
"inference_tokens_processed": 17711104,
"early_stopping": false
}