Add H100 latency chart and prompt, editing and resolution guidance

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/latency_t2i_1k_2k.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -68,6 +68,29 @@ Both adapters are v0.1. Pick one based on whether you need quality or speed.
 
 Each adapter is trained for its own sigma schedule. Load only one at a time.
 
+## Prompts, editing and resolution
+
+The adapters were trained at **1K resolution only**, with a mix of simple and
+upsampled prompts, text-to-image generation, and single- and multi-image editing
+with **up to 3 reference images**. Prompt upsampling is optional; detailed prompts
+usually give better results.
+
+Start at **1024 × 1024** and use at most **3 reference images** for editing.
+Higher resolutions (including 2K) and more reference images may work, but are
+outside the training coverage and quality may vary. The 2K timings below show
+inference speed, not a guarantee of quality at that resolution.
+
+## Speed on a single H100
+
+![Qwen-Image-2.1 and P-Turbo generation latency at 1K and 2K](assets/latency_t2i_1k_2k.png)
+
+Text-to-image with the official Qwen model-card example prompt, BF16, batch size 1,
+on one NVIDIA H100 80GB. Median of 3 requests after one warmup per configuration.
+Includes prompt encoding, denoising and decoding; excludes PNG saving, model
+loading and warmup. Base: 40 steps with KV cache on; P-Turbo: 5 or 8 steps with
+KV cache off, following the respective model-card settings. LoRAs were unmerged;
+no CFG, compilation or CPU offload. These timings do not imply equal image quality.
+
 ## Quickstart
 
 Runs text-to-image and image editing with the same pipeline. Needs a **CUDA GPU**.