npm - oidn-web - Versions diffs - 0.4.0 → 0.5.0 - Mend

oidn-web 0.4.0 → 0.5.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (106) hide show

package/CHANGELOG.md +45 -0
package/README.md +83 -19
package/dist/oidn.js +3239 -2642
package/dist/oidn.umd.cjs +512 -296
package/lib/UNet.d.ts +54 -20
package/lib/UNet.js +194 -118
package/lib/UNet.js.map +1 -1
package/lib/backend.d.ts +1 -8
package/lib/backend.js +1 -9
package/lib/backend.js.map +1 -1
package/lib/finalRgbShader.d.ts +13 -0
package/lib/finalRgbShader.js +160 -0
package/lib/finalRgbShader.js.map +1 -0
package/lib/graphOptimizer.js +1 -2
package/lib/graphOptimizer.js.map +1 -1
package/lib/hdrTransfer.d.ts +14 -0
package/lib/hdrTransfer.js +61 -0
package/lib/hdrTransfer.js.map +1 -0
package/lib/main.d.ts +16 -7
package/lib/main.js +4 -5
package/lib/main.js.map +1 -1
package/lib/nativeUNet.d.ts +39 -3
package/lib/nativeUNet.js +449 -120
package/lib/nativeUNet.js.map +1 -1
package/lib/process.d.ts +5 -11
package/lib/process.js +35 -49
package/lib/process.js.map +1 -1
package/lib/tileScheduler.d.ts +32 -4
package/lib/tileScheduler.js +133 -20
package/lib/tileScheduler.js.map +1 -1
package/package.json +9 -2
package/src/UNet.ts +287 -158
package/src/backend.ts +1 -14
package/src/finalRgbShader.ts +186 -0
package/src/graphOptimizer.ts +1 -2
package/src/hdrTransfer.ts +88 -0
package/src/main.ts +28 -13
package/src/nativeUNet.ts +515 -116
package/src/process.ts +43 -70
package/src/tileScheduler.ts +216 -24
package/benchmarks/compare.mjs +0 -651
package/benchmarks/leak.mjs +0 -255
package/benchmarks/results/before-spatial.json +0 -391
package/benchmarks/results/before-spatial.md +0 -47
package/benchmarks/results/int8-scan.json +0 -2007
package/benchmarks/results/int8-scan.md +0 -160
package/benchmarks/results/int8-w8a8-scan.json +0 -2007
package/benchmarks/results/int8-w8a8-scan.md +0 -160
package/benchmarks/results/int8-weight-channel.json +0 -1413
package/benchmarks/results/int8-weight-channel.md +0 -118
package/benchmarks/results/int8-weight-only.json +0 -1437
package/benchmarks/results/int8-weight-only.md +0 -118
package/benchmarks/results/kernel-webnn-final.json +0 -1115
package/benchmarks/results/kernel-webnn-final.md +0 -104
package/benchmarks/results/latest-optimized.json +0 -375
package/benchmarks/results/latest-optimized.md +0 -47
package/benchmarks/results/latest.json +0 -391
package/benchmarks/results/latest.md +0 -47
package/benchmarks/results/profile-baseline.json +0 -331
package/benchmarks/results/profile-baseline.md +0 -12
package/benchmarks/results/profile-conv2x.json +0 -331
package/benchmarks/results/profile-conv2x.md +0 -12
package/benchmarks/results/profile-fast-init.json +0 -385
package/benchmarks/results/profile-fast-init.md +0 -47
package/benchmarks/results/profile-fp16-fma.json +0 -369
package/benchmarks/results/profile-fp16-fma.md +0 -47
package/benchmarks/results/profile-fp16-tiled-decoder.json +0 -369
package/benchmarks/results/profile-fp16-tiled-decoder.md +0 -47
package/benchmarks/results/profile-fp16-tiled-encoder.json +0 -369
package/benchmarks/results/profile-fp16-tiled-encoder.md +0 -47
package/benchmarks/results/profile-fp16-unfused-pool.json +0 -385
package/benchmarks/results/profile-fp16-unfused-pool.md +0 -47
package/benchmarks/results/profile-input-major.json +0 -347
package/benchmarks/results/profile-input-major.md +0 -12
package/benchmarks/results/profile-k16.json +0 -347
package/benchmarks/results/profile-k16.md +0 -12
package/benchmarks/results/profile-k4.json +0 -347
package/benchmarks/results/profile-k4.md +0 -12
package/benchmarks/results/profile-pool-reuse.json +0 -331
package/benchmarks/results/profile-pool-reuse.md +0 -12
package/benchmarks/results/profile-precompiled.json +0 -385
package/benchmarks/results/profile-precompiled.md +0 -47
package/benchmarks/results/profile-static-channels.json +0 -385
package/benchmarks/results/profile-static-channels.md +0 -47
package/benchmarks/results/profile-static-io.json +0 -385
package/benchmarks/results/profile-static-io.md +0 -47
package/benchmarks/results/profile-tiled-conv.json +0 -331
package/benchmarks/results/profile-tiled-conv.md +0 -12
package/benchmarks/results/profile-tiled-decoder.json +0 -347
package/benchmarks/results/profile-tiled-decoder.md +0 -12
package/benchmarks/results/profile-tiled-matmul.json +0 -331
package/benchmarks/results/profile-tiled-matmul.md +0 -12
package/benchmarks/results/profile-unfused-decoder.json +0 -379
package/benchmarks/results/profile-unfused-decoder.md +0 -12
package/benchmarks/results/profile-unfused-pool.json +0 -347
package/benchmarks/results/profile-unfused-pool.md +0 -12
package/benchmarks/results/spatial-auto.json +0 -575
package/benchmarks/results/spatial-auto.md +0 -61
package/benchmarks/results/subgroup-smoke.json +0 -1094
package/benchmarks/results/subgroup-smoke.md +0 -104
package/benchmarks/results/webnn-smoke.json +0 -739
package/benchmarks/results/webnn-smoke.md +0 -76
package/scripts/inspect-model.mjs +0 -64
package/tests/modelSpec.test.mjs +0 -128
package/tests/resourceLifecycle.test.mjs +0 -383
package/tests/tileScheduler.test.mjs +0 -90

package/benchmarks/results/latest.json DELETED Viewed

@@ -1,391 +0,0 @@
-{
-  "generatedAt": "2026-08-20T04:14:37.676Z",
-  "currentCommit": "5d9c1c8771e2a03c90bfadafec4b5dd28fada446",
-  "baselineCommit": "136c00c34aa5cf4561b4137034dbd643177f0a5f",
-  "settings": {
-    "width": 512,
-    "height": 512,
-    "tileSize": 512,
-    "warmup": 1,
-    "runs": 5,
-    "model": "rt_hdr_calb_cnrm_large.tza",
-    "gpuQueueCompletionIncluded": true
-  },
-  "browser": "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome",
-  "adapter": {
-    "vendor": "apple",
-    "architecture": "metal-3",
-    "device": "",
-    "description": ""
-  },
-  "results": [
-    {
-      "label": "TFJS (136c00c)",
-      "bundle": "baseline",
-      "baseline": true,
-      "initializationMs": 9.100000001490116,
-      "timesMs": [
-        34.80000000447035,
-        34.79999999701977,
-        34.600000001490116,
-        34.5,
-        34.600000001490116
-      ],
-      "supportsFP16": true,
-      "adapter": {
-        "vendor": "apple",
-        "architecture": "metal-3",
-        "device": "",
-        "description": ""
-      },
-      "summary": {
-        "minMs": 34.5,
-        "medianMs": 34.600000001490116,
-        "meanMs": 34.66000000089407,
-        "p95Ms": 34.80000000447035,
-        "maxMs": 34.80000000447035
-      },
-      "messages": []
-    },
-    {
-      "label": "WGSL FP32",
-      "bundle": "current",
-      "baseline": false,
-      "precision": "fp32",
-      "initializationMs": 36.19999999552965,
-      "timesMs": [
-        35.70000000298023,
-        35.79999999701977,
-        35.600000001490116,
-        35.79999999701977,
-        35.600000001490116
-      ],
-      "executionProfile": {
-        "totalMs": 35.260974,
-        "layers": [
-          {
-            "id": "input-pack",
-            "durationMs": 0.050663
-          },
-          {
-            "id": "enc_conv1a",
-            "durationMs": 0.839491
-          },
-          {
-            "id": "enc_conv1b",
-            "durationMs": 3.697358
-          },
-          {
-            "id": "pool1",
-            "durationMs": 0.410641
-          },
-          {
-            "id": "enc_conv2a",
-            "durationMs": 1.41958
-          },
-          {
-            "id": "enc_conv2b",
-            "durationMs": 2.268612
-          },
-          {
-            "id": "pool2",
-            "durationMs": 0.054497
-          },
-          {
-            "id": "enc_conv3a",
-            "durationMs": 0.769328
-          },
-          {
-            "id": "enc_conv3b",
-            "durationMs": 0.986439
-          },
-          {
-            "id": "pool3",
-            "durationMs": 0.018041
-          },
-          {
-            "id": "enc_conv4a",
-            "durationMs": 0.400018
-          },
-          {
-            "id": "enc_conv4b",
-            "durationMs": 0.611212
-          },
-          {
-            "id": "pool4",
-            "durationMs": 0.008375
-          },
-          {
-            "id": "enc_conv5a",
-            "durationMs": 0.258609
-          },
-          {
-            "id": "enc_conv5b",
-            "durationMs": 0.336772
-          },
-          {
-            "id": "dec_conv4a",
-            "durationMs": 1.438454
-          },
-          {
-            "id": "dec_conv4b",
-            "durationMs": 0.612338
-          },
-          {
-            "id": "dec_conv3a",
-            "durationMs": 2.351731
-          },
-          {
-            "id": "dec_conv3b",
-            "durationMs": 0.987107
-          },
-          {
-            "id": "dec_conv2a",
-            "durationMs": 4.435979
-          },
-          {
-            "id": "dec_conv2b",
-            "durationMs": 2.204491
-          },
-          {
-            "id": "dec_conv1a",
-            "durationMs": 6.840541
-          },
-          {
-            "id": "dec_conv1b",
-            "durationMs": 3.763853
-          },
-          {
-            "id": "dec_conv1c",
-            "durationMs": 0.496844
-          }
-        ]
-      },
-      "runtimeInfo": {
-        "configuredEngine": "wgsl",
-        "gpuEngine": "wgsl",
-        "precision": "fp32",
-        "model": "oidn-unet-large-v1",
-        "modelFamily": "oidn-unet-large",
-        "inputChannels": 9,
-        "dynamicTile": {
-          "enabled": false,
-          "currentTileSize": 512,
-          "minTileSize": 256,
-          "maxTileSize": 512,
-          "targetTileTimeMs": 16
-        },
-        "lastExecution": {
-          "width": 512,
-          "height": 512,
-          "tileWidth": 512,
-          "tileHeight": 512,
-          "tileCount": 1,
-          "durationMs": 37.100000001490116,
-          "tileTimeMs": {
-            "min": 37,
-            "median": 37,
-            "mean": 37,
-            "max": 37
-          }
-        }
-      },
-      "supportsFP16": true,
-      "adapter": {
-        "vendor": "apple",
-        "architecture": "metal-3",
-        "device": "",
-        "description": ""
-      },
-      "summary": {
-        "minMs": 35.600000001490116,
-        "medianMs": 35.70000000298023,
-        "meanMs": 35.7,
-        "p95Ms": 35.79999999701977,
-        "maxMs": 35.79999999701977
-      },
-      "messages": [],
-      "validation": {
-        "sampleCount": 6144,
-        "meanAbsoluteError": 0.0000011096847932397698,
-        "rootMeanSquaredError": 0.0000018071588418628667,
-        "maxAbsoluteError": 0.000013589859008789062,
-        "passed": true,
-        "thresholds": {
-          "maxMeanError": 0.0001,
-          "maxAbsoluteError": 0.001
-        }
-      }
-    },
-    {
-      "label": "WGSL FP16",
-      "bundle": "current",
-      "baseline": false,
-      "precision": "fp16",
-      "initializationMs": 27.799999997019768,
-      "timesMs": [
-        27.600000001490116,
-        27.5,
-        27.69999999552965,
-        27.600000001490116,
-        27.700000002980232
-      ],
-      "executionProfile": {
-        "totalMs": 27.002728,
-        "layers": [
-          {
-            "id": "input-pack",
-            "durationMs": 0.044205
-          },
-          {
-            "id": "enc_conv1a",
-            "durationMs": 0.617922
-          },
-          {
-            "id": "enc_conv1b",
-            "durationMs": 3.10056
-          },
-          {
-            "id": "pool1",
-            "durationMs": 0.348062
-          },
-          {
-            "id": "enc_conv2a",
-            "durationMs": 1.193719
-          },
-          {
-            "id": "enc_conv2b",
-            "durationMs": 2.061333
-          },
-          {
-            "id": "pool2",
-            "durationMs": 0.049996
-          },
-          {
-            "id": "enc_conv3a",
-            "durationMs": 0.678334
-          },
-          {
-            "id": "enc_conv3b",
-            "durationMs": 0.836574
-          },
-          {
-            "id": "pool3",
-            "durationMs": 0.017041
-          },
-          {
-            "id": "enc_conv4a",
-            "durationMs": 0.340646
-          },
-          {
-            "id": "enc_conv4b",
-            "durationMs": 0.528885
-          },
-          {
-            "id": "pool4",
-            "durationMs": 0.011041
-          },
-          {
-            "id": "enc_conv5a",
-            "durationMs": 0.182155
-          },
-          {
-            "id": "enc_conv5b",
-            "durationMs": 0.233986
-          },
-          {
-            "id": "dec_conv4a",
-            "durationMs": 0.946983
-          },
-          {
-            "id": "dec_conv4b",
-            "durationMs": 0.528885
-          },
-          {
-            "id": "dec_conv3a",
-            "durationMs": 1.704563
-          },
-          {
-            "id": "dec_conv3b",
-            "durationMs": 0.839656
-          },
-          {
-            "id": "dec_conv2a",
-            "durationMs": 3.035857
-          },
-          {
-            "id": "dec_conv2b",
-            "durationMs": 1.97488
-          },
-          {
-            "id": "dec_conv1a",
-            "durationMs": 4.425147
-          },
-          {
-            "id": "dec_conv1b",
-            "durationMs": 3.108393
-          },
-          {
-            "id": "dec_conv1c",
-            "durationMs": 0.193905
-          }
-        ]
-      },
-      "runtimeInfo": {
-        "configuredEngine": "wgsl",
-        "gpuEngine": "wgsl",
-        "precision": "fp16",
-        "model": "oidn-unet-large-v1",
-        "modelFamily": "oidn-unet-large",
-        "inputChannels": 9,
-        "dynamicTile": {
-          "enabled": false,
-          "currentTileSize": 512,
-          "minTileSize": 256,
-          "maxTileSize": 512,
-          "targetTileTimeMs": 16
-        },
-        "lastExecution": {
-          "width": 512,
-          "height": 512,
-          "tileWidth": 512,
-          "tileHeight": 512,
-          "tileCount": 1,
-          "durationMs": 28.599999994039536,
-          "tileTimeMs": {
-            "min": 28.599999994039536,
-            "median": 28.599999994039536,
-            "mean": 28.599999994039536,
-            "max": 28.599999994039536
-          }
-        }
-      },
-      "supportsFP16": true,
-      "adapter": {
-        "vendor": "apple",
-        "architecture": "metal-3",
-        "device": "",
-        "description": ""
-      },
-      "summary": {
-        "minMs": 27.5,
-        "medianMs": 27.600000001490116,
-        "meanMs": 27.620000000298024,
-        "p95Ms": 27.700000002980232,
-        "maxMs": 27.700000002980232
-      },
-      "messages": [],
-      "validation": {
-        "sampleCount": 6144,
-        "meanAbsoluteError": 0.0010210334021394374,
-        "rootMeanSquaredError": 0.001609680379854656,
-        "maxAbsoluteError": 0.012395620346069336,
-        "passed": true,
-        "thresholds": {
-          "maxMeanError": 0.005,
-          "maxAbsoluteError": 0.05
-        }
-      }
-    }
-  ]
-}

package/benchmarks/results/latest.md DELETED Viewed

@@ -1,47 +0,0 @@
-# oidn-web benchmark
-- Current: `5d9c1c8771e2a03c90bfadafec4b5dd28fada446`
-- TFJS baseline: `136c00c34aa5cf4561b4137034dbd643177f0a5f`
-- Input: 512x512, fixed tile 512, 5 runs after 1 warmup(s)
-- Adapter: apple
-| Runtime | Init ms | Median ms | P95 ms | Speedup | Precision |
-| --- | ---: | ---: | ---: | ---: | --- |
-| TFJS (136c00c) | 9.10 | 34.60 | 34.80 | 1.00x | fp32 |
-| WGSL FP32 | 36.20 | 35.70 | 35.80 | 0.97x | fp32 |
-| WGSL FP16 | 27.80 | 27.60 | 27.70 | 1.25x | fp16 |
-## Output validation
-Compared against sampled TFJS FP32 output.
-| Runtime | Status | MAE | RMSE | Max error | Samples |
-| --- | --- | ---: | ---: | ---: | ---: |
-| WGSL FP32 | pass | 1.11e-6 | 1.81e-6 | 1.36e-5 | 6144 |
-| WGSL FP16 | pass | 1.02e-3 | 1.61e-3 | 1.24e-2 | 6144 |
-## GPU hot layers
-### WGSL FP32
-Profiled GPU total: 35.26 ms
-| Node | GPU ms |
-| --- | ---: |
-| dec_conv1a | 6.84 |
-| dec_conv2a | 4.44 |
-| dec_conv1b | 3.76 |
-| enc_conv1b | 3.70 |
-| dec_conv3a | 2.35 |
-### WGSL FP16
-Profiled GPU total: 27.00 ms
-| Node | GPU ms |
-| --- | ---: |
-| dec_conv1a | 4.43 |
-| dec_conv1b | 3.11 |
-| enc_conv1b | 3.10 |
-| dec_conv2a | 3.04 |
-| enc_conv2b | 2.06 |