10 days ago
-
Card: bands written before the run; end of second window specified after 7 of 10 values
11 days ago
-
Card: resume check stated with both computations
-
Card: history note states facts only
-
Evaluation history of 1.00 and 1.01 (measured values only)
-
Card: final evaluation of step 2860; training loss vs. windows fixed before the run
-
Card: component shares of the training split
-
Card: exact step ranges (pack-01 from step 1802, decay from step 2290), stream shares to 0.01 %
-
Add train_gpt_ref_r23.py (step 2860)
-
Add tokenizer.json (step 2860)
-
Add model.safetensors (step 2860)
-
Add load_slayer.py (step 2860)
-
Add config.json (step 2860)
-
SHA256SUMS of the release files
-
Model card (training facts; results to follow)
-
initial commit
