Fast Neural Style 4 styles LiteRT fp16 (fully-GPU, Pixel 8a corr 0.9999, 9ms)

.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
+samples/sample.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,51 @@
+---
+license: bsd-3-clause
+library_name: LiteRT
+pipeline_tag: image-to-image
+tags: [litert, tflite, on-device, android, gpu, style-transfer, neural-style, fast-neural-style]
+base_model: pytorch/examples
+---
+
+# Fast Neural Style Transfer — LiteRT (on-device, fully-GPU, 4 styles)
+
+Fast neural **style transfer** ([PyTorch examples](https://github.com/pytorch/examples/tree/main/fast_neural_style)
+`TransformerNet`, Johnson et al.), converted to **LiteRT** and running **fully on the `CompiledModel` GPU**
+(ML Drift) on Android. Applies an artistic style to a photo — **4 styles** (candy / mosaic / rain_princess /
+udnie), each a **3.5 MB** fp16 graph.
+
+![Fast Neural Style — content + candy / mosaic / rain / udnie (on-device LiteRT GPU)](samples/sample.png)
+
+## On-device (Pixel 8a, Tensor G3 — verified)
+
+| | |
+|---|---|
+| nodes on GPU | **350 / 350** LITERT_CL (full residency) |
+| inference | **~9 ms** (256×256) |
+| size | 3.5 MB per style (fp16) |
+| accuracy | device-vs-PyTorch corr **0.9998–0.9999** (all 4 styles) |
+
+```
+image[1,3,256,256] (RGB 0-255) →[GPU: TransformerNet]→ stylized[1,3,256,256] (RGB 0-255)
+```
+
+## How it converts (litert-torch) — three numerically-exact re-authorings
+
+1. **`ReflectionPad2d` → zero-pad** (`GATHER_ND` → `PAD`; border-only difference).
+2. **Large conv activations → conv-weight scaling.** The conv outputs reach ≈ |5000|, where the Mali delegate's
+   fp16 conv accumulation loses precision → garbage (device corr 0.34 at full residency — *residency ≠
+   correctness*). Each conv is followed by an `InstanceNorm` (which is **scale-invariant**), so scaling those
+   conv weights down so the output is ≈ |10| is **exact** (IN output unchanged) and keeps the fp16 accumulation
+   precise → corr 1.0.
+3. **`InstanceNorm` → SafeInstanceNorm** (down-scaled-domain spatial reduction, fp16-safe; SafeLayerNorm class).
+
+Upsample is `interpolate(nearest)` (no transposed conv → no ZeroStuff). Result: banned ops NONE, ≤4D,
+tflite-vs-torch corr **1.0**, device-vs-torch corr **0.9999**.
+
+## Preprocessing
+
+Center-crop to square, resize to 256×256, RGB **0–255** (no normalization), NCHW. Output is 0–255 RGB (clamp).
+
+## License
+
+[BSD-3-Clause](https://github.com/pytorch/examples/blob/main/LICENSE). Upstream:
+[pytorch/examples](https://github.com/pytorch/examples/tree/main/fast_neural_style).

style_candy_fp16.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b09d87e4f15ad6ba5f61c6599614bee8db097e09c751a4cf2f15abb5e5b543f9
+size 3482944

style_mosaic_fp16.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc6fe40fd4da73b2b9ad585dba20264f2b863ed587d33dc7ab6bd493c8af5911
+size 3482944

style_rain_princess_fp16.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0939d4e16ff78dc67f8b91ff6fc8cfea22a4724e022ef067bbb8d0fd8ca87cd8
+size 3482944

style_udnie_fp16.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:516f6bacbe9487b4a6d47bd5d8ae8b41634ea480f53f20069e5165e397fa6615
+size 3482944