Feat (script): Added pre-generated `captions.tsv`

Feat (script): Added initial reproduction README.

Feat (script): Added option to validate on MLPerf validation set & to load a pre-quantized checkpoint.

Fix (script): Updated dependencies.

Feat (script/env): Added conda environment YAMLs

Fix (script/env): added pandas requirement.

Reorg (minimal): moved minimal script to sub-directory.

Fix (script): Quantize output of QuantLora layers.

Fix (script): print output directory.

Feat (script): Added calibration size argument.

Fix (minimal): Move to CPU before export

Quantization script

Remove potential overflow / saturation error.

Added comments - highlight possible overflow situation

Updated math model to target int8 x int8 kernels.

Updated QOp model to fuse SmoothQuant scales with input quantization

Output reference tensors

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

Upload params.safetensors with huggingface_hub

add missing smoothquant factors

update quant_params with correct shapes

Fix: set `keepdim=True`

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

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

Upload quant_param.json with huggingface_hub

Upload quant_param.json with huggingface_hub

Upload math_model.py with huggingface_hub

Upload params.safetensors with huggingface_hub

Upload params.safetensors with huggingface_hub

Upload quant_param.json with huggingface_hub

Upload math_model.py with huggingface_hub

Upload quant_param.json with huggingface_hub

initial commit