Upload PiEa checkpoint

.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
+eval_comparison_grid.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,80 @@
 ---
 license: apache-2.0
+language:
+- en
+pipeline_tag: image-to-image
+library_name: pytorch
+tags:
+- image-to-image
+- diffusion
+- pixel-diffusion-decoder
+- super-resolution
+- denoising
+- pytorch
+- safetensors
+- pie
 ---
+
+# PiEa: Pixel Diffusion Decoder
+
+`PiEa` is a compact image-to-image Pixel Diffusion Decoder. It is trained to reconstruct clean pixels from noisy pixels while conditioning on a degraded version of the same image. The model is designed for decoder-style image restoration, image refinement, and pixel-space reconstruction experiments.
+
+PiEa is not a text-to-image generator. It is a decoder component: it expects image-like conditioning and produces image-like denoising predictions.
+
+## Model details
+
+| Property | Value |
+| --- | --- |
+| Model name | `PiEa` |
+| Developer | PiEa-ai |
+| Model type | Pixel diffusion decoder |
+| Input type | `PiE` |
+| Architecture | Compact U-Net diffusion decoder |
+| Parameters | 1,010,776,675 |
+| Resolution | 512 x 512 |
+| Base channels | 464 |
+| Channel multipliers | [1, 2, 4, 6] |
+| Dataset used | `['huggan/wikiart', 'huggan/smithsonian_butterflies_subset']` |
+| Optimizer | AdamW |
+| Precision | bfloat16 autocast |
+
+## Throughput
+
+| Metric | Value |
+| --- | --- |
+| Training steps | 9,141 |
+| Image tokens processed | 2,396,258,304 |
+| Image tokens/sec | 1331229.59 |
+| Target image tokens/sec | 250,000 |
+| Final loss | 0.023287 |
+
+Image tokens are counted as processed spatial pixels: `batch_size * height * width` per optimization step.
+
+## Usage
+
+This checkpoint is saved as raw PyTorch/safetensors artifacts. Load the model with the architecture definition used in the training script or port the weights into a compatible pixel decoder implementation.
+
+```python
+from safetensors.torch import load_file
+
+state_dict = load_file("model.safetensors")
+```
+
+## Limitations
+
+- PiEa is an experimental pixel decoder, not a full image generation system.
+- It is trained on image reconstruction and denoising, not prompt following.
+- Quality depends on the surrounding pipeline and conditioning signal.
+- The small checkpoint is optimized for throughput and experimentation rather than maximum fidelity.
+- Evaluate on your target image distribution before use.
+
+## Citation
+
+```bibtex
+@misc{piea2026pixeldecoder,
+  title={PiEa: Pixel Diffusion Decoder},
+  author={PiEa-ai},
+  year={2026},
+  url={https://huggingface.co/PiEa-ai/PiEa}
+}
+```

config.json

@@ -0,0 +1,24 @@
+{
+  "architectures": [
+    "PiEaPixelDiffusionDecoder"
+  ],
+  "model_type": "piea_pixel_diffusion_decoder",
+  "model_name": "PiEa",
+  "manufacturer": "PiEa-ai",
+  "input_type": "PiE",
+  "pipeline_tag": "image-to-image",
+  "image_size": 512,
+  "in_channels": 6,
+  "out_channels": 3,
+  "base_channels": 464,
+  "channel_mults": [
+    1,
+    2,
+    4,
+    6
+  ],
+  "prediction_type": "epsilon",
+  "conditioning": "degraded_image",
+  "torch_dtype": "bfloat16",
+  "parameter_count": 1010776675
+}

evaluation_report.json

@@ -0,0 +1,15 @@
+{
+  "model_name": "PiEa",
+  "model_dir": "/artifacts/20260528-163638/PiEa",
+  "dataset": "huggan/wikiart",
+  "num_samples": 8,
+  "image_size": 512,
+  "sigma": 0.35,
+  "mse": 0.0034244786365889013,
+  "l1": 0.04037976358085871,
+  "psnr": 30.674655300944988,
+  "condition_mse": 0.006136163972162952,
+  "condition_psnr": 28.14163034902346,
+  "comparison_grid": "/artifacts/20260528-163638/PiEa/eval_comparison_grid.png",
+  "grid_columns": "clean | degraded_condition | noisy_input | piea_reconstruction | absolute_difference"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a209eb8e1f20db3026bcc57e67a92bd90ade410133ca7e6969e175749fecc791
+size 4043116796

training_report.json

@@ -0,0 +1,54 @@
+{
+  "model_name": "PiEa",
+  "repo_id": "PiEa-ai/PiEa",
+  "dataset": [
+    "huggan/wikiart",
+    "huggan/smithsonian_butterflies_subset"
+  ],
+  "run_id": "20260528-163638",
+  "model_dir": "/artifacts/20260528-163638/PiEa",
+  "parameter_count": 1010776675,
+  "train_minutes": 30,
+  "steps": 9141,
+  "batch_size": 1,
+  "image_size": 512,
+  "base_channels": 464,
+  "channel_mults": [
+    1,
+    2,
+    4,
+    6
+  ],
+  "compile_model": false,
+  "resume_latest": true,
+  "source_checkpoint": "/artifacts/20260528-163440/PiEa",
+  "image_tokens": 2396258304,
+  "image_tokens_per_second": 1331229.5910597388,
+  "target_tokens_per_second": 250000,
+  "final_loss": 0.023287300020456314,
+  "cached_images": 1024,
+  "config": {
+    "architectures": [
+      "PiEaPixelDiffusionDecoder"
+    ],
+    "model_type": "piea_pixel_diffusion_decoder",
+    "model_name": "PiEa",
+    "manufacturer": "PiEa-ai",
+    "input_type": "PiE",
+    "pipeline_tag": "image-to-image",
+    "image_size": 512,
+    "in_channels": 6,
+    "out_channels": 3,
+    "base_channels": 464,
+    "channel_mults": [
+      1,
+      2,
+      4,
+      6
+    ],
+    "prediction_type": "epsilon",
+    "conditioning": "degraded_image",
+    "torch_dtype": "bfloat16",
+    "parameter_count": 1010776675
+  }
+}