text_encoder/z_image_fp8-quantization.json

{
 "format": "FP8",
 "scale": "per-row",
 "per_tensor_relative_error": 0.016641,
 "per_tensor_weights_lost": 0.031543,
 "matrices_measured": 253
}