Model card: explicit zero-based/native indices; effective batch split

README.md

@@ -24,7 +24,7 @@ the weights is pending the owners' decision. Everything up to `ckpt_6750.pt` is
 | Steps | 7150 in total × 1 048 576 tokens = 7 497 318 400 tokens (+1430 steps = +1 499 463 680 tokens over 1.02) |
 | Schedule | **constant lr 6e-4 to the end, no final decay** — the decay is deferred to the end of the whole Slayer1B line (pre-training continues on the whole prepared corpus) |
 | Data | continues the pack-03 stream of [SlayerLab/slayer-pl-8x3b](https://huggingface.co/datasets/SlayerLab/slayer-pl-8x3b) (revision `d50df04a`) used by 1.02 from step 5390: steps 5721–7150 |
-| Hardware | steps 5721–6750: 4 × NVIDIA RTX 4090 (24 GB); steps 6751–7150: 4 × NVIDIA L40S (48 GB), resumed from `ckpt_6750.pt` (sha256 `970518a1…`) with the same trainer, recipe, data order and global batch. The data order of the resume was verified against the original run (batch hash of step 6751 and 7149 identical). DDP; global batch 256 × 4096 tokens as in 1.00–1.02 |
+| Hardware | steps 5721–6750: 4 × NVIDIA RTX 4090 (24 GB); steps 6751–7150: 4 × NVIDIA L40S (48 GB), resumed from `ckpt_6750.pt` (sha256 `970518a1…`) with the same trainer, recipe, data order and global batch. The data order of the resume was verified against the original run (batch hash of zero-based iterations 6751 and 7149, i.e. native steps 6752 and 7150, identical). DDP; global batch 256 × 4096 tokens as in 1.00–1.02 (effective: 4 ranks × micro-batch 2 × gradient accumulation 32) |
 | Checkpoints | every 150 steps from 5850 and the final step 7150; all are constant-lr states |
 | Trainer | `train_gpt_ref_r24_1.py`: the 1.02 trainer, releasing the checkpoint tensors from GPU memory after resuming (needed on 24 GB cards); the model definition is unchanged |
 | Training curve | steps up to 6850: https://track.fabryka.ai/run/7ca4f84f-3b07-4525-8cde-f3a936dc3425 · resume 6750–7150: https://track.fabryka.ai/run/5e362bba-53e0-5e5e-91b1-938ca2e3795a |
@@ -38,8 +38,8 @@ https://track.fabryka.ai/run/5c53b0cc-c947-5cd5-806a-0ace4b360ded); those checkp
 A leak scan of the whole training superset (SlayerLab/slayer-pl-8x3b@d50df04a, 22 603 122 documents) found n-gram
 overlap with BoolQ, MMLU and ARC test items in the English shared part (`shared/en`: 1428 / 1429 / 1345 documents).
 Overlap is not a measured contamination rate; the assessment is in progress. Read the BoolQ, MMLU and ARC numbers with
-this caveat. From pack-04 on (step 7192), the 5133 matching documents are excluded from new training data; earlier
-steps, including 1.00–1.03, used the unfiltered data.
+this caveat. From pack-04 on (first filtered update: zero-based 7192, native step 7193), the 5133 matching documents
+are excluded from new training data; earlier steps, including 1.00–1.03, used the unfiltered data.
 
 ## File format