Skip to content

Actions: hwang2409/whitematter

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
134 workflow runs
134 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
fix: add missing <cstring> include in bpe_tokenizer.h CI #133: Commit 8018c20 pushed by hwang2409
3m 18s main
main
3m 18s
feat: BPE tokenization + weight tying + top-p sampling for NinoGPT CI #132: Commit 139911d pushed by hwang2409
6m 56s main
main
6m 56s
data: expand Nino dialogue to 447 entries (913K tokens) CI #131: Commit 61340a4 pushed by hwang2409
3m 10s main
main
3m 10s
fix: revert to fast hand-written helpers + shrink model for CPU training CI #130: Commit fdb4016 pushed by hwang2409
2m 48s main
main
2m 48s
fix: keep NinoGPT on CPU — GPU transfers slower than CPU-only training CI #129: Commit 626c05f pushed by hwang2409
3m 6s main
main
3m 6s
feat: GPU-accelerate NinoGPT via reshape-to-2D + framework matmul CI #128: Commit ec81a35 pushed by hwang2409
3m 7s main
main
3m 7s
data: add 264 Nino Nakano dialogue entries for GPT training CI #127: Commit a0378ee pushed by hwang2409
4m 4s main
main
4m 4s
fix: make pad2d differentiable to fix 150MB/batch memory leak CI #126: Commit 447a7c5 pushed by hwang2409
3m 1s main
main
3m 1s
fix: size-adaptive memory pool to prevent OOM on 224x224 training CI #125: Commit e293a8b pushed by hwang2409
3m 16s main
main
3m 16s
feat: memory-mapped data loading for ImageNette training CI #124: Commit e2e5694 pushed by hwang2409
3m 19s main
main
3m 19s
fix: add MemoryPool::trim() to prevent OOM on 224x224 training CI #123: Commit a0e0627 pushed by hwang2409
3m 12s main
main
3m 12s
fix: stream ImageNette preprocessing to disk — prevents OOM CI #122: Commit 9e3955f pushed by hwang2409
3m 12s main
main
3m 12s
docs: ARCHITECTURE.md — codebase structure for contributors CI #121: Commit 6c45ffe pushed by hwang2409
2m 52s main
main
2m 52s
cleanup: remove dead shadow cache code from CUDA backend CI #120: Commit bf3565b pushed by hwang2409
3m 36s main
main
3m 36s
fix: eliminate all compiler warnings across the codebase CI #119: Commit 3a297e6 pushed by hwang2409
3m 7s main
main
3m 7s
feat: ImageNette training — standard ResNet-18 on 224x224 images CI #118: Commit 23f26a1 pushed by hwang2409
2m 56s main
main
2m 56s
fix: save/load BatchNorm running stats in checkpoints CI #117: Commit 3428ccc pushed by hwang2409
3m 0s main
main
3m 0s
fix: read uint64 dims in export tool (matches checkpoint format) Build and Publish Wheels #1: Commit c469d70 pushed by hwang2409
1m 7s v1.0
v1.0
1m 7s
fix: read uint64 dims in export tool (matches checkpoint format) CI #116: Commit c469d70 pushed by hwang2409
3m 11s main
main
3m 11s
fix: export tool reads WMCP checkpoint format from training CI #115: Commit c31245c pushed by hwang2409
3m 8s main
main
3m 8s
feat: browser-based CIFAR-10 demo with ONNX Runtime Web CI #114: Commit 579dd7d pushed by hwang2409
3m 16s main
main
3m 16s
fix: remove flush_grad_shadows call (method removed in revert) CI #113: Commit 63c3572 pushed by hwang2409
3m 18s main
main
3m 18s
revert: restore exact WSL2 working state (commit ad7ed82) CI #112: Commit a177eeb pushed by hwang2409
3m 8s main
main
3m 8s
debug: disable cuDNN conv2d, re-enable cuDNN BN — isolating conv back… CI #111: Commit 5cb3ca0 pushed by hwang2409
3m 17s main
main
3m 17s
fix: restore exact working config from WSL2 (commit ad7ed82) CI #110: Commit 40b2be7 pushed by hwang2409
3m 7s main
main
3m 7s