Replace comparison matrices with vertical layout

Add narrow vertical side-by-side WebP q95 matrices for FP8 vs SDNQ and RTX 4090 SDNQ vs official NF4; remove the old wide/square matrix assets after updating the model card references.

.gitattributes

@@ -43,3 +43,5 @@ assets/benchmarks/opensource2.png filter=lfs diff=lfs merge=lfs -text
 assets/samples/collage_landscape.jpg filter=lfs diff=lfs merge=lfs -text
 assets/comparison_matrix.webp filter=lfs diff=lfs merge=lfs -text
 assets/sdnq_vs_nf4_4090_side_by_side.webp filter=lfs diff=lfs merge=lfs -text
+assets/original_vs_sdnq_vertical.webp filter=lfs diff=lfs merge=lfs -text
+assets/sdnq_vs_nf4_4090_vertical.webp filter=lfs diff=lfs merge=lfs -text

README.md

@@ -83,9 +83,9 @@ The FP8 baseline was loaded through the upstream `ideogram4` `Ideogram4Pipeline.
 
 ## Example Matrix
 
-The matrix below contains the 10 original FP8 generations followed by the 10 SDNQ UInt4 generations. It is a square WebP at quality 95.
+The matrix below keeps the original FP8 and SDNQ UInt4 outputs side by side in narrow vertical columns. It is a WebP at quality 95.
 
-![Comparison matrix](assets/comparison_matrix.webp)
+![Original FP8 vs SDNQ UInt4 vertical comparison](assets/original_vs_sdnq_vertical.webp)
 
 ## Prompt Set
 
@@ -105,7 +105,8 @@ The matrix below contains the 10 original FP8 generations followed by the 10 SDN
 ## Files
 
 - `prompts.json`: the 10 structured prompts used for the comparison.
-- `assets/comparison_matrix.webp`: square WebP comparison matrix, quality 95.
+- `assets/original_vs_sdnq_vertical.webp`: vertical side-by-side WebP comparison matrix for original FP8 vs SDNQ UInt4, quality 95.
+- `assets/sdnq_vs_nf4_4090_vertical.webp`: vertical side-by-side WebP comparison matrix for the RTX 4090 SDNQ vs official NF4 follow-up, quality 95.
 - `benchmark/`: raw benchmark JSONL/CSV files and `summary.json`.
 - `quantization_manifest.json`: component-level quantization timings, storage, and VRAM peaks.
 - `ideogram4_sdnq_pipeline.py`: loader helper for the SDNQ custom transformer components.
@@ -119,7 +120,7 @@ Hardware: RunPod NVIDIA GeForce RTX 4090, 24 GB VRAM, single process, concurrenc
 | sdnq | 10.00 | 211.61 | 14124.00 | 14466.00 | 59.65 | 37.05 | 37.57 | 19768.00 | 20521.00 |
 | nf4 | 10.00 | 269.31 | 15370.00 | 15766.00 | 36.57 | 36.31 | 36.77 | 21012.00 | 21801.00 |
 
-![SDNQ vs official NF4 on RTX 4090](assets/sdnq_vs_nf4_4090_side_by_side.webp)
+![SDNQ vs official NF4 on RTX 4090](assets/sdnq_vs_nf4_4090_vertical.webp)
 
 Raw follow-up metrics are in `benchmark/summary_4090_sdnq_vs_nf4.json`, `benchmark/sdnq_4090_metrics.*`, and `benchmark/nf4_4090_metrics.*`. The exact runner used for the follow-up is `benchmark/followup_runner.py`.