Add training details

README.md

@@ -6,13 +6,15 @@ base_model:
 pipeline_tag: text-to-image
 tags:
 - '360'
-- '360°'
-- '360-degree'
-- '360-image'
+- 360°
+- 360-degree
+- 360-image
 - equirectangular
 - equirectangular-projection
 - image-generation
 - text-to-image
+datasets:
+- CaptionEmporium/pexels-568k-internvl2
 ---
 
 # Qwen 360 Diffusion
@@ -31,6 +33,19 @@ Based on extensive testing, the model's capabilities vastly exceed all other cur
 
 The model is designed to be capable of producing equirectangular images that can be used for non-VR purposes such as general imagery, photography, artwork, architecture, portraiture, and many other concepts. 
 
+### Training Details
+
+Training was first performed using nf4 for 8 epochs.
+- `qwen-360-diffusion-int4-bf16-v1.safetensors` was trained for 7 epochs.
+- `qwen-360-diffusion-int4-bf16-v1-b.safetensors` was trained for 8 epochs.
+
+Training then continued at int8 for another 4 epochs.
+- `qwen-360-diffusion-int8-bf16-v1.safetensors` was trained for a total of 12 epochs.
+
+Each 360 degree training image was randomly rotated horizontally 3 times for data augmentation (original + 3 rotations).
+
+For regularization, 64k images were randomly selected from the [pexels-568k-internvl2](https://huggingface.co/datasets/CaptionEmporium/pexels-568k-internvl2) dataset.
+
 ---