DaisyChainAI
/

DaisyChain-Train

distributed-training

Model card Files Files and versions

DaisyChain-Train / daisychain /verified

908 kB

Ctrl+K

Ctrl+K

1 contributor

History: 3 commits

Quazim0t0's picture

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

99509b7 verified about 16 hours ago

weights
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September about 16 hours ago
__init__.py

690 Bytes
Old-hardware training through emulated GPU logic 3 months ago
backends.py

3.28 kB
verified: bound GEMM memory, certify the multiply LUT, test the shipped paths about 1 month ago
common.py

2.76 kB
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September about 16 hours ago
instrument.py

2.2 kB
verified: bound GEMM memory, certify the multiply LUT, test the shipped paths about 1 month ago
kernel.py

1.56 kB
verified: bound GEMM memory, certify the multiply LUT, test the shipped paths about 1 month ago
lut.py

4.01 kB
verified: bound GEMM memory, certify the multiply LUT, test the shipped paths about 1 month ago
mul8.py

5.98 kB
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September about 16 hours ago
mul8_tied.py

4.07 kB
Old-hardware training through emulated GPU logic 3 months ago
ops.py

7.57 kB
Refuse NaN/Inf at every int8 quantize; requant rounds half-up (retrained); bounded units; README for September about 16 hours ago
qat.py

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