DeepSpeed
adding 8bit dequantization kernel for asym fine-grained block quantization in zero-inference
#4450
Merged
Merged
adding 8bit dequantization kernel for asym fine-grained block quantization in zero-inference #4450
tjruwase
merged
15 commits into
master
from
styoun/zero-inf-8bit-q
kernels added for asym fine-grained block quantization with 8bits
a9328b9f
formatting
347dea02
clean up the code
a3734881
Merge branch 'master' into styoun/zero-inf-8bit-q
561cd16f
tjruwase
commented
on 2023-10-05
rename quantize_int4.cu to quantize_intX.cu
e0ad1673
Merge branch 'styoun/zero-inf-8bit-q' of https://github.com/microsoft…
90858e7c
rename test_int4_quantization.py to test_intX_quantization.py
2d341405
"rename test_int4_quantization.py to test_intX_quantization.py"
b9896cd3
rename
359c3e58
tjruwase
commented
on 2023-10-05
tjruwase
commented
on 2023-10-05
fix after the pr comments
1ea4bf45
increased coverage of QuantLinear test
0b216378
Merge branch 'master' into styoun/zero-inf-8bit-q
571ab839
formatting
68c258c3
Merge branch 'styoun/zero-inf-8bit-q' of https://github.com/microsoft…
7c7310dc
Merge branch 'master' into styoun/zero-inf-8bit-q
c9704348
tjruwase
approved these changes
on 2023-10-10
Login to write a write a comment.
Login via GitHub