Commits · amd-shark/sdxl-quant-int8

QKV fused and all linear layers sym

3cee2a6
verified

GiusFra commited on Jul 14

QKV fused and all linear layers sym

cf48f0f
verified

GiusFra commited on Jul 14

QKV fused and sym

6f44cfb
verified

GiusFra commited on Jul 13

QKV fused and sym

b175bf6
verified

GiusFra commited on Jul 13

Fused QKV quant_params.json with zp

7e99883
verified

GiusFra commited on Jul 11

Added vae weights with FP16 fix.

2de7ba8

nickfraser commited on Jul 11

Fused QKV safetensor with zp

0339659
verified

GiusFra commited on Jul 11

Fused QKV safetensor

348012d
verified

GiusFra commited on Jul 10

Fused QKV quant_params.json

a793c5a
verified

GiusFra commited on Jul 10

Fix model loading

7f81513
verified

GiusFra commited on Jul 9

Updates to minimal quantization script. (#1)

72eb84b
verified

GiusFra

nickfraser commited on Jul 9

Update quant params structure (#2)

6b62ce4
verified

nickfraser commited on Jul 3

Reference inputs

17638f5
verified

GiusFra commited on Jul 2

Updated quant_params

fb3aa3b
verified

GiusFra commited on Jul 2

Updated params.safetensors

36c8b73
verified

GiusFra commited on Jul 2

Output reference tensors

6e61570
verified

GiusFra commited on Jul 2

Quantization script

ecec5b7
verified

GiusFra commited on Jun 21

Remove potential overflow / saturation error.

161df88

nickfraser commited on Jun 19

Added comments - highlight possible overflow situation

3f5851c

nickfraser commited on Jun 19

Updated math model to target int8 x int8 kernels.

4024f9d

nickfraser commited on Jun 19

Updated QOp model to fuse SmoothQuant scales with input quantization

dca9b6e

nickfraser commited on Jun 18

Output reference tensors

8e3c05a
verified

GiusFra commited on Jun 14

Add config.json from stable-diffusion-xl-base-1.0/unet

54be8be

Stella Laurenzo commited on Jun 12

Upload params.safetensors with huggingface_hub

1dad0d1
verified

GiusFra commited on Jun 12

add missing smoothquant factors

99e9d19
verified

GiusFra commited on Jun 12

update quant_params with correct shapes

d6a388a
verified

GiusFra commited on Jun 11

Fix: set `keepdim=True`

9ab1060

nickfraser commited on Jun 11

[test] Fixed shapes to match new `quant_param.json`

673c9f2

nickfraser commited on Jun 11

[math_model/test] Added "QOp" implementation and basic tests.

eb5a5f6

nickfraser commited on Jun 11

Upload quant_param.json with huggingface_hub

d67ece3
verified

GiusFra commited on Jun 7

Upload quant_param.json with huggingface_hub

bcd05a6
verified

GiusFra commited on Jun 7

Upload math_model.py with huggingface_hub

049c65f
verified

GiusFra commited on Jun 7

Upload params.safetensors with huggingface_hub

742c3ad
verified

GiusFra commited on Jun 7

Upload params.safetensors with huggingface_hub

76a91d8
verified

GiusFra commited on Jun 7

Upload quant_param.json with huggingface_hub

01fc5a5
verified

GiusFra commited on Jun 7

Upload math_model.py with huggingface_hub

d5dfd96
verified

GiusFra commited on Jun 7

Upload quant_param.json with huggingface_hub

88730c2
verified

GiusFra commited on Jun 6

initial commit

af8fc68
verified

stellaraccident commited on Jun 5

amd-shark
/

sdxl-quant-int8

Commit History

QKV fused and all linear layers sym

3cee2a6
verified

QKV fused and all linear layers sym

cf48f0f
verified

QKV fused and sym

6f44cfb
verified

QKV fused and sym

b175bf6
verified

Fused QKV quant_params.json with zp

7e99883
verified

Added vae weights with FP16 fix.

2de7ba8

Fused QKV safetensor with zp

0339659
verified

Fused QKV safetensor

348012d
verified

Fused QKV quant_params.json

a793c5a
verified

Fix model loading

7f81513
verified

Updates to minimal quantization script. (#1)

72eb84b
verified

Update quant params structure (#2)

6b62ce4
verified

Reference inputs

17638f5
verified

Updated quant_params

fb3aa3b
verified

Updated params.safetensors

36c8b73
verified

Output reference tensors

6e61570
verified

Quantization script

ecec5b7
verified

Remove potential overflow / saturation error.

161df88

Added comments - highlight possible overflow situation

3f5851c

Updated math model to target int8 x int8 kernels.

4024f9d

Updated QOp model to fuse SmoothQuant scales with input quantization

dca9b6e

Output reference tensors

8e3c05a
verified

Add config.json from stable-diffusion-xl-base-1.0/unet

54be8be

Upload params.safetensors with huggingface_hub

1dad0d1
verified

add missing smoothquant factors

99e9d19
verified

update quant_params with correct shapes

d6a388a
verified

Fix: set `keepdim=True`

9ab1060

[test] Fixed shapes to match new `quant_param.json`

673c9f2

[math_model/test] Added "QOp" implementation and basic tests.

eb5a5f6

Upload quant_param.json with huggingface_hub

d67ece3
verified

Upload quant_param.json with huggingface_hub

bcd05a6
verified

Upload math_model.py with huggingface_hub

049c65f
verified

Upload params.safetensors with huggingface_hub

742c3ad
verified

Upload params.safetensors with huggingface_hub

76a91d8
verified

Upload quant_param.json with huggingface_hub

01fc5a5
verified

Upload math_model.py with huggingface_hub

d5dfd96
verified

Upload quant_param.json with huggingface_hub

88730c2
verified

initial commit

af8fc68
verified

Commit History

QKV fused and all linear layers sym 3cee2a6 verified

QKV fused and all linear layers sym cf48f0f verified

QKV fused and sym 6f44cfb verified

QKV fused and sym b175bf6 verified

Fused QKV quant_params.json with zp 7e99883 verified

Added vae weights with FP16 fix. 2de7ba8

Fused QKV safetensor with zp 0339659 verified

Fused QKV safetensor 348012d verified

Fused QKV quant_params.json a793c5a verified

Fix model loading 7f81513 verified

Updates to minimal quantization script. (#1) 72eb84b verified

Update quant params structure (#2) 6b62ce4 verified

Reference inputs 17638f5 verified

Updated quant_params fb3aa3b verified

Updated params.safetensors 36c8b73 verified

Output reference tensors 6e61570 verified

Quantization script ecec5b7 verified

Remove potential overflow / saturation error. 161df88

Added comments - highlight possible overflow situation 3f5851c

Updated math model to target int8 x int8 kernels. 4024f9d

Updated QOp model to fuse SmoothQuant scales with input quantization dca9b6e

Output reference tensors 8e3c05a verified

Add config.json from stable-diffusion-xl-base-1.0/unet 54be8be

Upload params.safetensors with huggingface_hub 1dad0d1 verified

add missing smoothquant factors 99e9d19 verified

update quant_params with correct shapes d6a388a verified

Fix: set `keepdim=True` 9ab1060

[test] Fixed shapes to match new `quant_param.json` 673c9f2

[math_model/test] Added "QOp" implementation and basic tests. eb5a5f6

Upload quant_param.json with huggingface_hub d67ece3 verified

Upload quant_param.json with huggingface_hub bcd05a6 verified

Upload math_model.py with huggingface_hub 049c65f verified

Upload params.safetensors with huggingface_hub 742c3ad verified

Upload params.safetensors with huggingface_hub 76a91d8 verified

Upload quant_param.json with huggingface_hub 01fc5a5 verified

Upload math_model.py with huggingface_hub d5dfd96 verified

Upload quant_param.json with huggingface_hub 88730c2 verified

initial commit af8fc68 verified

QKV fused and all linear layers sym

3cee2a6
verified

QKV fused and all linear layers sym

cf48f0f
verified

QKV fused and sym

6f44cfb
verified

QKV fused and sym

b175bf6
verified

Fused QKV quant_params.json with zp

7e99883
verified

Added vae weights with FP16 fix.

2de7ba8

Fused QKV safetensor with zp

0339659
verified

Fused QKV safetensor

348012d
verified

Fused QKV quant_params.json

a793c5a
verified

Fix model loading

7f81513
verified

Updates to minimal quantization script. (#1)

72eb84b
verified

Update quant params structure (#2)

6b62ce4
verified

Reference inputs

17638f5
verified

Updated quant_params

fb3aa3b
verified

Updated params.safetensors

36c8b73
verified

Output reference tensors

6e61570
verified

Quantization script

ecec5b7
verified

Remove potential overflow / saturation error.

161df88

Added comments - highlight possible overflow situation

3f5851c

Updated math model to target int8 x int8 kernels.

4024f9d

Updated QOp model to fuse SmoothQuant scales with input quantization

dca9b6e

Output reference tensors

8e3c05a
verified

Add config.json from stable-diffusion-xl-base-1.0/unet

54be8be

Upload params.safetensors with huggingface_hub

1dad0d1
verified

add missing smoothquant factors

99e9d19
verified

update quant_params with correct shapes

d6a388a
verified

Fix: set `keepdim=True`

9ab1060

[test] Fixed shapes to match new `quant_param.json`

673c9f2

[math_model/test] Added "QOp" implementation and basic tests.

eb5a5f6

Upload quant_param.json with huggingface_hub

d67ece3
verified

Upload quant_param.json with huggingface_hub

bcd05a6
verified

Upload math_model.py with huggingface_hub

049c65f
verified

Upload params.safetensors with huggingface_hub

742c3ad
verified

Upload params.safetensors with huggingface_hub

76a91d8
verified

Upload quant_param.json with huggingface_hub

01fc5a5
verified

Upload math_model.py with huggingface_hub

d5dfd96
verified

Upload quant_param.json with huggingface_hub

88730c2
verified

initial commit

af8fc68
verified