Upload Stage 1 and Stage 2 checkpoints

README.md

@@ -0,0 +1,39 @@
+---
+library_name: pytorch
+license: other
+tags:
+- super-resolution
+- latent-diffusion
+- pytorch
+- rocm
+- research
+---
+
+# sr-diffusion
+
+Research checkpoint storage for the `sr-diffusion` project.
+
+GitHub: https://github.com/BitIntx/sr-diffusion
+
+This project trains a vision-only x4 latent diffusion super-resolution pipeline
+from scratch. It does not use a pretrained text-to-image diffusion model.
+
+Current artifacts are study/research checkpoints. They are not a production SR
+model and are not intended for commercial use. Dataset license constraints
+should be reviewed before redistributing derived weights publicly.
+
+## Artifacts
+
+| Path | Source | SHA256 |
+| --- | --- | --- |
+| `checkpoints/stage1_autoencoder_best_eval_recon.pt` | `best_eval_recon.pt` | `6f01ec6d1ded` |
+| `configs/autoencoder_photo10k.yaml` | `autoencoder_photo10k.yaml` | `b1743ae89efc` |
+| `checkpoints/stage2_latent_pretrain_step_0004000.pt` | `step_0004000.pt` | `0bf569e909d5` |
+| `configs/latent_pretrain_photo10k.yaml` | `latent_pretrain_photo10k.yaml` | `85175fa83f41` |
+| `metrics/stage2_step_0004000_metrics.json` | `step_0004000_metrics.json` | `189c98b4c35c` |
+
+## Stages
+
+- Stage 1: factor-4 VAE / Autoencoder over 512px HR crops.
+- Stage 2: deterministic LR-to-HR-latent pretraining with the Stage 1 VAE frozen.
+- Stage 3: conditional latent diffusion U-Net, planned.