Yesterday
-
Release cooldown from step 19,000 (8,000 steps): 7-task average 48.43; add run.py
-
Replace table images with markdown tables; list all datasets incl. later runs
-
Datasets: add 75M pretraining shares and earlier 75M chat sources
-
Table-only prerenders, progress table image, no duplicated tables
-
Rework model card: link all datasets, add author Dominik Andrzejewski
2 days ago
-
MP-01 100M base: revised model card, each table shown once
-
Remove duplicate chart files
-
MP-01 100M base: prerendered compute and average tables placed through the card
-
MP-01 100M base: highlight best score per column in benchmark table
-
MP-01 100M base: model card with comparison charts, average and benchmark tables, logo
-
MP-01 100M base: cooldown checkpoint (step 4000), new model card with compute disclosure, comparison and logo
-
Add MP-01 (9500) 100M base: weights, config, tokenizer, model code, model card and eval results
-
initial commit
