Add 1024px DreamLite ONNX Android test models

1024/decoder_1024_fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92cc83fbfbe756ffdf9887a62accce8e722cb21d75785dbd4e239d50f40ca25f
+size 4913046

1024/manifest.json

@@ -0,0 +1,87 @@
+{
+  "name": "DreamLite fixed-prompt test",
+  "resolution": 1024,
+  "steps": 4,
+  "seed": 42,
+  "prompt": "A red ceramic teapot on a wooden table, natural window light, product photography",
+  "conditions": "Precomputed text embedding; no live text/image encoder in this test",
+  "schedules": {
+    "1": {
+      "sigmas": [
+        1.0,
+        0.0
+      ],
+      "timesteps": [
+        1000.0
+      ]
+    },
+    "2": {
+      "sigmas": [
+        1.0,
+        0.7595109343528748,
+        0.0
+      ],
+      "timesteps": [
+        1000.0,
+        759.5109252929688
+      ]
+    },
+    "4": {
+      "sigmas": [
+        1.0,
+        0.9045307636260986,
+        0.7595109343528748,
+        0.5128441452980042,
+        0.0
+      ],
+      "timesteps": [
+        1000.0,
+        904.53076171875,
+        759.5109252929688,
+        512.8441162109375
+      ]
+    }
+  },
+  "shapes": {
+    "sample": [
+      1,
+      4,
+      128,
+      256
+    ],
+    "timestep": [
+      1
+    ],
+    "encoder_hidden_states": [
+      1,
+      24,
+      2048
+    ],
+    "encoder_attention_mask": [
+      1,
+      24
+    ],
+    "time_ids": [
+      1,
+      2
+    ]
+  },
+  "files": [
+    {
+      "name": "unet_1024_fp32.onnx",
+      "size": 1562073493,
+      "sha256": "2fdf23642c17ddcc15562dab1a8407f13105b7afedf65af5f96b7b10506e264e"
+    },
+    {
+      "name": "decoder_1024_fp32.onnx",
+      "size": 4913046,
+      "sha256": "92cc83fbfbe756ffdf9887a62accce8e722cb21d75785dbd4e239d50f40ca25f"
+    }
+  ],
+  "repository": "xay2001/dreamlite-android",
+  "revision": "main",
+  "remote_path": "1024",
+  "published": false,
+  "qnn_verified": false,
+  "weights": "FP32 ONNX; HTP requests FP16; NOT paper W8A8"
+}

1024/unet_1024_fp32.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fdf23642c17ddcc15562dab1a8407f13105b7afedf65af5f96b7b10506e264e
+size 1562073493

README.md

@@ -10,7 +10,7 @@ tags:
 - experimental
 ---
 
-# DreamLite Android — 512 px test package
+# DreamLite Android — 512 / 1024 px test packages
 
 Experimental ONNX conversion of [DreamLite-mobile](https://huggingface.co/carlofkl/DreamLite-mobile/tree/diffusers), for the independent DreamLite Lab Android test application. Original implementation: [ByteVisionLab/DreamLite](https://github.com/ByteVisionLab/DreamLite).
 
@@ -18,6 +18,8 @@ Experimental ONNX conversion of [DreamLite-mobile](https://huggingface.co/carlof
 
 `512/unet_512_fp32.onnx` and `512/decoder_512_fp32.onnx` are static-shape FP32 models, exported with ONNX opset 17. `512/manifest.json` lists sizes and SHA-256 checksums. Download individual files; do not rename them.
 
+The `1024/` directory contains the corresponding native 1024px UNet and decoder. APK 0.3 supports resolution selection and separate local storage; each resolution needs approximately 1.57 GB. Changing step count reuses the selected resolution's weights. 1024px inference does not resize a 512px generated image.
+
 The application defaults to **512 × 512, 1 step** and offers 2/4-step comparison. One-step output is experimental and visibly softer in the local sample; four steps is the quality reference. This is not a model trained specifically for one-step inference.
 
 ## Scope and validation
@@ -25,6 +27,7 @@ The application defaults to **512 × 512, 1 step** and offers 2/4-step compariso
 - Fixed, pre-encoded text condition in the test APK: a red ceramic teapot on a wooden table near a window, seed 42. The output image is generated on each run, not cached.
 - The 2B text/image encoder is **not** included. This package does not provide arbitrary text entry or photo editing in the test APK.
 - CPU ONNX numerical checks and a complete 512px, one-step image comparison against the PyTorch baseline were performed. This does not establish general image quality or mobile speed.
+- Complete 1024px, four-step CPU output comparison passed the preset image threshold (PSNR 75.57 dB, pixel MAE 0.001803 on a 0–255 scale). The 1024 UNet's stricter elementwise tolerance test failed (maximum absolute error 0.003335); this record is retained. One/two-step 1024 quality has not been separately evaluated. These remain experimental test models.
 - QNN HTP strict mode requests FP16 execution and disables CPU fallback. **DreamLite phone NPU compatibility and sub-second latency remain unverified.** These are not the paper's W8A8 quantized weights.
 - This 512px experiment does not demonstrate the 1024px, multi-condition project target.