Card: describe run/ folder
Browse files
README.md
CHANGED
|
@@ -31,6 +31,13 @@ model card: [SlayerLab/Slayer1B-1.01](https://huggingface.co/SlayerLab/Slayer1B-
|
|
| 31 |
`torch.save` dictionary of the trainer `train_gpt_ref_r23.py` (in the 1.00 model repository): `model`, `opt` (Muon +
|
| 32 |
AdamW), `step`, `config`. Load with `torch.load(path, map_location="cpu", weights_only=True)`.
|
| 33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
## SHA-256
|
| 35 |
|
| 36 |
Recorded off the training node at the moment each checkpoint was written, before upload; files are downloaded back and
|
|
|
|
| 31 |
`torch.save` dictionary of the trainer `train_gpt_ref_r23.py` (in the 1.00 model repository): `model`, `opt` (Muon +
|
| 32 |
AdamW), `step`, `config`. Load with `torch.load(path, map_location="cpu", weights_only=True)`.
|
| 33 |
|
| 34 |
+
## Run record (`run/`)
|
| 35 |
+
|
| 36 |
+
`metrics.jsonl` (training loss and learning rate every 10 steps), `train.log`, `train.out`, `args.txt` (trainer
|
| 37 |
+
arguments), `checkpoints_SHA256SUMS` (written on the node at save time), data manifests and sums of both token streams
|
| 38 |
+
(`wsad_v2_*`: pack-00 stream, `wsad_p01_*`: pack-01 stream), `13gram_p01.json` (13-gram overlap of the pack-01 stream
|
| 39 |
+
with the evaluation sets), `pip_freeze.txt` (software environment).
|
| 40 |
+
|
| 41 |
## SHA-256
|
| 42 |
|
| 43 |
Recorded off the training node at the moment each checkpoint was written, before upload; files are downloaded back and
|