Update README.md

README.md

@@ -40,11 +40,11 @@ widget:
 <Gallery />
 
 # Mayakovsky Style Soviet Constructivist Posters & Cartoons Flux LoRA – Version 2 – by SOON®
-Trained via Ostris' [ai-toolkit](https://replicate.com/ostris/flux-dev-lora-trainer/train) on 50 high-resolution scans of 1910s/1920s posters & artworks by the great Soviet **poet, artist, & Marxist activist Vladimir Mayakovsky**.  <br> 
+Trained via Ostris' [ai-toolkit](https://github.com/ostris/ai-toolkit) on 50 high-resolution scans of 1910s/1920s posters & artworks by the great Soviet **poet, artist, & Marxist activist Vladimir Mayakovsky**.  <br> 
 For this training experiment, we first spent many days rigorously translating the textual elements (slogans, captions, titles, inset poems, speech fragments, etc), with form/signification/rhymes intact, throughout every image subsequently used for training.  <br> 
 These translated textographic elements were, furthermore, re-placed by us into their original visual contexts, using fonts matched up to the sources.  <br> 
 We then manually composed highly detailed paragraph-long captions, wherein we detailed both the graphic and the textual content of each piece, its layout, as well as the most intuitive/intended apprehension of each composition. <br> 
-This is the LoRA checkpoint at step 3000, while the training as a whole went for 5000 steps at a DiT Learning Rate of .00001, batch 1, with the adafactor optimizer, and the text encoders trained alongside the DiT!<br> 
+This is the LoRA checkpoint at step 3000, trained at a DiT Learning Rate of .00001, batch 1, with the adafactor optimizer, and the text encoders trained alongside the DiT. The results are, unfortunately, messier and not as convergent as we'd hoped, even compared to the Flux versions (trained on the same data-set and available here as well). Tactical prompting may perhaps mediate this somewhat. And at least when and where S.D.3.5 breaks during training, it does so in wildly creative and often surprising ways, which sometimes even befit certain art styles.  <br> 
 No synthetic data was used for the training, nor any auto-generated captions! Everything was manually and attentively pre-curated with a deep respect for the sources used.  <br> 
 
 This is a **rank-64/alpha-64 Constructivist Art & Soviet Satirical Cartoon LoRA for Stable Diffusion 3.5 Large** <br>