Create config.json
efda8af
verified
-
all_linear_sym
QKV fused and all linear layers sym
-
all_linear_sym_7_5
QKV fused and all linear layers sym, guidance scale 7.5
-
all_linear_sym_7_5_calib12
MI250 QKV fused and all linear layers sym, guidance scale 7.5, calib steps 12
-
all_linear_sym_7_5_noconvout
MI250 QKV fused and all linear layers sym, guidance scale 7.5, no conv_out
-
full_sym
Full symmetric
-
fused_qkv
Fused QKV quant_params.json with zp
-
mi250_all_linear_sym_7_5
MI250 QKV fused and all linear layers sym, guidance scale 7.5
-
mi250_all_sym_8
MI250 QKV fused and all layers sym, guidance scale 8
-
mi250_all_sym_8_step10
MI250 QKV fused and all layers sym, guidance scale 8, 10 step, second try
-
mi250_all_sym_8_step12
MI250 Q
-
mi300_all_sym_8_no4conv_step10_fp32
MI300 FP32, determinism on, g8s10, no4conv
-
mi300_all_sym_8_step10
MI300 QKV fused and all layers sym, guidance scale 8, 10 step
-
mi300_all_sym_8_step10_det
MI300 QKV fused and all layers sym, guidance scale 8, 10 steps, deterministic
-
mi300_all_sym_8_step10_fp32
MI300 FP32, determinism on, g8s10
-
mi300_all_sym_8_step14_fp32
Create config.json
-
nvidia_all_sym_8
NVIDIA All sym, guidance scale 8
-
nvidia_all_sym_8_step10
NVIDIA All sym, guidance scale 8, step 10
-
qkv_sym
QKV fused and sym
-
quant_sdxl
Fix model loading
-
1.63 kB
Updated quant_params
-
1.68 kB
Add config.json from stable-diffusion-xl-base-1.0/unet
-
9.05 kB
Update quant params structure (#2)
-
7.11 GB
Output reference tensors
-
5.14 GB
Updated params.safetensors
-
661 kB
Reference inputs
-
85.1 MB
add missing smoothquant factors
-
86.8 MB
Updated quant_params
-
1.18 kB
Update quant params structure (#2)
-
1.04 kB
Update quant params structure (#2)
-
167 MB
Added vae weights with FP16 fix.
