Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Keep MSE validation tabular in dataset repo (#8)

Expose layer_roofline as loadable dataset config (#7)

Repair dataset card configs after MSE merge (#6)

Fix dataset card split metadata keys (#5)

Preserve MSE validation evidence for synthetic replay claims (#3)

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Remove predictions and roofline_quadrant configs, keep 3 configs only

Fix terminology: NCU profiles, not CUDA profiles

Clean up dataset card: remove slop, add concurrency filtering context

Remove workload trace configs from dataset card

Clean dataset card: positive framing, remove sparsity percentages and failure rates

Split multi-schema configs into individual configs (5 -> 7) for datasets lib compat

Fix dataset_info: num_rows -> num_examples for datasets lib compat

Fix dataset_info: add missing configs, omit features for multi-schema splits

Add kernel profiles (148k rows), workload traces, serving predictions, roofline quadrant data

Fix: PyTorch 2.10.0, CUDA 12.8 (was incorrectly CUDA 12.4)

v1.0: dataset card with full metadata, precision info, coverage sparsity, croissant alignment

Fix dataset card: accurate stats, failed request disclosure, croissant alignment, BF16 column fix

Clean up dataset card: simplify retention table, remove AI-generated prose

Update data with fresh R2 pull (5,743 rows) and add retention table to dataset card

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub