2.92 MB

Ctrl+K

1 contributor

History: 84 commits

Quazim0t0

Rewrite model card in first person; ASCII hyphens only; keep all facts

346b731 verified 12 days ago

config
Old-hardware training through emulated GPU logic 3 months ago
daisychain
Reductions refuse what they used to accept: GPU row-max non-finite guard, averageGrads length/NaN refusal, int32 accumulator bound, capacity-weighted cluster loss, finite reduced-gradient check, empty CUDA_VISIBLE_DEVICES 13 days ago
docker
Old-hardware training through emulated GPU logic 3 months ago
docs
Old-hardware training through emulated GPU logic 3 months ago
examples
Old-hardware training through emulated GPU logic 3 months ago
scripts
Old-hardware training through emulated GPU logic 3 months ago
web
Logic pass: CPU attention mirrors get the same int32 bound as the GPU kernels (peers refuse identically); gpuMlpChain returns pooled buffers when the row-max guard throws 13 days ago
.gitattributes

1.52 kB
initial commit 3 months ago
.gitignore

119 Bytes
Old-hardware training through emulated GPU logic 3 months ago
LICENSE

1.09 kB
Old-hardware training through emulated GPU logic 3 months ago
README.md

29.6 kB
Rewrite model card in first person; ASCII hyphens only; keep all facts 12 days ago
export_luts_web.py

3.06 kB
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September 16 days ago
pyproject.toml

1.06 kB
Update URLs for DaisyChain-Train 3 months ago
requirements.txt

35 Bytes
Old-hardware training through emulated GPU logic 3 months ago
test_cluster_reductions.py

4.09 kB
Reductions refuse what they used to accept: GPU row-max non-finite guard, averageGrads length/NaN refusal, int32 accumulator bound, capacity-weighted cluster loss, finite reduced-gradient check, empty CUDA_VISIBLE_DEVICES 13 days ago
test_verified_units.py

8.35 kB
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September 16 days ago