README: add note for landscapes-epoch31

README.md

@@ -25,7 +25,7 @@ language:
 
 ## Model description
 
-Trained on 336 illustrations from _Nouveau dictionnaire encyclopédique universel illustré,_ A.K.A. the Trousset encyclopedia.
+Trained on illustrations from _Nouveau dictionnaire encyclopédique universel illustré,_ A.K.A. the Trousset encyclopedia.
 
 FLUX.1 [dev] has decent performance with “woodcut illustration” already,
 but sometimes it uses mid-tones instead of proper black & white shading.
@@ -46,6 +46,13 @@ They also include “1880s, 19th century”; I haven't tested prompting those.
 
 ## Download model
 
+- [trousset-landscapes.epoch31](./trousset-landscapes.epoch31.safetensors?download=true):
+  From a more specific subset of 97 images with a consistent aesthetic;
+  cityscapes and landscapes and a few other scenes with a similar mostly-full-frame style.
+  Appears to be effective at encouraging finer lines for shading compared to the base model's rendering of _woodcut illustration._
+- [woodcut-illustration.epoch8](./woodcut-illustration.epoch8.safetensors?download=true):
+  First published attempt. Sourced from 336 images, trained for 8 epochs (2688 steps). Some effect but not great.
+
 Weights for this model are available in Safetensors format.
 
 [Download](/keturn/woodcut-illustrations-Trousset-LoRA/tree/main) them in the Files & versions tab.
@@ -71,7 +78,8 @@ Some perhaps relevant settings:
     I read somewhere that alpha should be set to the same, and raising it from the default 1.0 seems to have helped.</dd>
 
   <dt>--network_args "train_blocks=single" "train_single_block_indices=18-37" "verbose=True"</dt>
-  <dd>In hopes that focusing on the higher half of the blocks would help prioritize small-scale detail over large structure.</dd>
+  <dd>In hopes that focusing on the higher half of the blocks would help prioritize small-scale detail over large structure.
+  (adjusted to `9-37` in the `trousset-landscapes.epoch31` training.)</dd>
 
   <dt>--timestep_sampling shift   --discrete_flow_shift 1.0</dt>
   <dd>fluxgym's default shift is 3.1. Unverified hunch that we want to bring this back down to focus on lower timesteps for small-scale stuff.</dd>
@@ -84,7 +92,7 @@ To accomodate 12 GB VRAM budget, used bfloat16 precision with the u-net at float
 I can think of lots of things one _could_ do to perhaps improve the result, though I don't have a good sense for which are most effective.
 
 - More dataset curation. The images were _relatively_ consistent, but the botanical illustrations are better than the mammals,
-  the engineering illustrations are different than the cityscapes, etc.
+  the engineering illustrations are different than the cityscapes, etc. I did some of this in the form of:
   - [`trousset-bugs-and-botany-sources.txt`](./trousset-bugs-and-botany-sources.txt): mostly flowers, close-up detail illustrations with no backgrounds.
   - [`trousset-landscape-sources.txt`](./trousset-landscape-sources.txt): cityscapes and landscapes and a few other scenes with a similar mostly-full-frame style.
 - Dataset augmentation. (I haven't yet turned to any of the tricks like mirroring the images, but I'm not sure if more is better at this point.)