omni-viewer-core 0.16.0 → 0.18.0

package/README.md

@@ -16,7 +16,8 @@ models, viewers mount into a DOM element, and everything host-specific
   legacy PPT), Markdown, LaTeX (structure and math preview, not typesetting)
 - **Data & spreadsheets** — Excel, CSV/TSV, JSON, JSONL/NDJSON, YAML, TOML,
   Parquet, Avro, HDF5, MATLAB MAT, NumPy (NPY/NPZ), Safetensors, GGUF, ONNX, TFLite/LiteRT,
-  Keras (.keras and legacy .h5), Protocol Buffers, ReqIF, SQLite
+  Keras (.keras and legacy .h5), Core ML (.mlmodel and .mlpackage),
+  OpenVINO IR (.xml + .bin), PyTorch Export (.pt2), Protocol Buffers, ReqIF, SQLite
 - **Media & graphics** — audio (waveform/spectrogram), video, images,
   Photoshop PSD
 - **Engineering & automotive** — CAN DBC, AUTOSAR ARXML, ASAM A2L, Vector
@@ -172,6 +173,132 @@ contents distinguish a Keras model from any other HDF5 file; hosts that read the
 whole file can refine that with `looksLikeKerasHdf5` from
 `omni-viewer-core/parsers/keras` and mount the Keras viewer instead.
 
+### Core ML models
+
+A `.mlmodel` is a serialized `CoreML.Specification.Model` protobuf, and an
+`.mlpackage` is a bundle whose `Manifest.json` points at one such spec plus the
+weight blobs its ML Program references. One entry point reads both — bytes that
+open with a ZIP header are treated as a packaged bundle:
+
+```ts
+import { mountCoremlViewer } from 'omni-viewer-core/viewers/coreml';
+
+await mountCoremlViewer({ fileName: file.name, data: bytes }, container, ctx);
+```
+
+Weight payloads are never decoded: blob references resolve to a file, an offset,
+and a byte count, and inline weight tensors to their size and quantization, so a
+multi-gigabyte model costs nothing to open. Package members are read in place,
+and JSZip is only consulted for a bundle that was re-zipped with compression.
+
+Two model families carry a graph, and the viewer reads them to different depths:
+
+- An **ML Program** (Core ML 5 and newer) is self-describing — every operation
+  names its own inputs, so attributes are shown exactly as the producer wrote
+  them, with no schema knowledge involved.
+- A **neural network** (the older encoding) stores each layer's parameters in a
+  distinct message per layer type. Layer types resolve for all of them, and
+  attributes are decoded for the common types; weights need no such table, since
+  `WeightParams` has one shape everywhere it appears and is found structurally.
+
+The viewer provides a per-graph computation graph with operation, value, and
+weight inspection, navigation into the blocks a control-flow operation
+references, and searchable operation, input/output, weight, package, and
+model-information panels. Custom layers, pipeline stages, unmapped layer types,
+and weight files that are not present are surfaced as warnings.
+
+Both extensions route by extension, and an extensionless zipped bundle is
+resolved by `probeContainer` from its `Manifest.json` + `Data/com.apple.CoreML`
+layout. A `.mlmodel` carries no leading magic, so hosts that can read a whole
+extensionless file can refine routing with `looksLikeCoremlSpec` from
+`omni-viewer-core/parsers/coreml`.
+
+### OpenVINO IR models
+
+An OpenVINO IR is two files that share a base name. The `.xml` is ordinary XML
+whose root is `<net name version>` holding `<layers>` and `<edges>`; the `.bin`
+has no header, magic, or index of its own — it is a raw concatenation of
+constant payloads that `Const` layers address by `offset` and `size`. The core
+receives the `.xml` as its input and the `.bin` as a sidecar the adapter reads
+from the same directory:
+
+```ts
+import { mountOpenVinoViewer } from 'omni-viewer-core/viewers/openvino';
+
+await mountOpenVinoViewer({ fileName: file.name, data: xmlBytes }, container, ctx, {
+  sidecars: { bin: binBytes } // optional
+});
+```
+
+The topology renders in full without the `.bin`. When it is supplied, every
+constant's byte range is checked against it and its leading values are decoded
+for a preview (all fixed-width IR element types, including f16, bf16, and the
+f8 variants); the rest of the weights are never read, so a multi-gigabyte model
+costs nothing beyond its XML. A `.bin` that is too short for the constants, or
+that has bytes no constant references, is surfaced as a warning — the usual
+signs of a mismatched pair (the unreferenced-bytes signal is only trusted when
+every constant was read).
+
+The viewer provides a computation graph with layer inspection (ports with
+their producers and consumers, `<data>` attributes, `rt_info`, weight ranges,
+and the sub-graphs a `Loop`, `TensorIterator`, or `If` layer carries in its
+body), plus searchable layer, constant, input/output, and model-information
+panels. IR v10 and v11 are the target; the legacy v7-and-earlier layout
+(`<layer precision>` with `<blobs>`) is read on a best-effort basis and
+flagged.
+
+Routing is by content only: `.xml` is shared with every other XML dialect, so
+the descriptor claims no extension and `detectViewer` reaches the viewer
+through its text sample (`looksLikeOpenVinoIr`, also exported from
+`omni-viewer-core/parsers/openvino`). A `.bin` opened on its own has nothing to
+identify it; an adapter that wants to open one should look for the same-named
+`.xml` beside it.
+
+### PyTorch Export packages (.pt2)
+
+A `.pt2` is what `torch.export.save` writes: a ZIP whose members sit under a
+folder named after the file, with the serialized Export IR program in
+`models/<name>.json`, parameters and buffers as raw storage bytes under
+`data/weights/` (indexed by `<name>_weights_config.json`), tensor constants and
+script objects under `data/constants/`, pickled sample inputs, optional
+AOTInductor artifacts under `data/aotinductor/`, and user files under `extra/`.
+
+```ts
+import { mountPt2Viewer } from 'omni-viewer-core/viewers/pt2';
+
+await mountPt2Viewer({ fileName: file.name, data: bytes }, container, ctx);
+```
+
+`parsePt2` reads the archive's central directory and only the JSON members it
+needs; stored weight payloads are viewed in place and sampled for a short value
+preview (every fixed-width torch dtype, including bf16 and the float8
+variants), so a multi-gigabyte package costs nothing beyond its program JSON.
+The viewer shows the ATen graph — placeholders coloured by signature kind
+(user input, parameter, buffer, constant), operator nodes, higher-order
+control-flow nodes with their `cond` / `while_loop` sub-graphs, and outputs —
+with a node inspector (arguments, producers and consumers, module stack,
+metadata, stack trace), plus searchable tables of nodes (sub-graph nodes
+included), weights and constants with their archive status, the graph
+signature (inputs and outputs by kind, buffer mutations, constant inputs), the
+nn.Module hierarchy with preserved call signatures, model information (schema
+and torch versions, opsets, range constraints, guards, operator counts), and
+the archive listing. Symbolic shapes are printed as expressions (`2*s0`)
+rather than the sympy `srepr` the file stores.
+
+A package holding several programs (`package_pt2`) gets a model selector; an
+AOTInductor-only package (`aoti_compile_and_package`) opens on the archive
+tab with its compile metadata. The pre-PT2 `torch.export.save` layout
+(`serialized_exported_program.json` beside torch.save pickles) is read on a
+best-effort basis: the graph is complete, but weights are listed from the
+signature without a preview. Schema 8.x is the target; enums
+written as member names by older versions are accepted.
+
+Routing is by the `.pt2` extension; an extensionless package reaches the viewer
+through `probeContainer`, which recognizes the `archive_format` marker
+(`looksLikePt2Archive` is also exported from `omni-viewer-core/parsers/pt2`).
+A plain `torch.save` checkpoint (`.pt` / `.pth`) is not an Export package and
+is not claimed.
+
 ### Archive host integration
 
 Archive decompression remains adapter-owned: pass an `ArchiveDecoder` to

package/dist/i18n/catalog.de.js

@@ -3,14 +3,18 @@
 export const CATALOG_DE = {
     'audio.title': 'Audio', 'video.title': 'Video', 'dbc.title': 'CAN-Datenbank',
     'audio.play': 'Wiedergabe', 'audio.pause': 'Pause', 'audio.stop': 'Stopp',
-    'audio.loop': 'Schleife', 'audio.clearRegions': 'Bereiche löschen', 'audio.volume': 'Lautstärke',
+    'audio.volume': 'Lautstärke',
+    'audio.loop.region': 'Bereich wiederholen', 'audio.loop.track': 'Titel wiederholen',
     'audio.region.start': 'Bereichsstart (s)', 'audio.region.end': 'Bereichsende (s)', 'audio.region.duration': 'Bereichslänge (s)',
     'audio.zoom': 'Zoom', 'audio.zoomIn': 'Vergrößern', 'audio.zoomOut': 'Verkleinern', 'audio.zoomFit': 'Anpassen',
     'audio.visualization': 'Ansicht', 'audio.vis.waveform': 'Wellenform', 'audio.vis.spectrogram': 'Spektrogramm',
     'audio.vis.both': 'Beide', 'audio.download': 'Herunterladen',
-    'audio.spectrogramScale': 'Skala',
+    'audio.spectrogramScale': 'Spektrogramm-Skala',
     'audio.scale.linear': 'Linear', 'audio.scale.mel': 'Mel', 'audio.scale.bark': 'Bark', 'audio.scale.erb': 'ERB',
+    'audio.scale.logarithmic': 'Logarithmisch',
+    'audio.frequencyRange': 'Frequenzbereich', 'audio.frequency.full': 'Gesamt', 'audio.frequency.limit': 'Bis {hz}',
     'audio.error.save': 'Die Audiodatei konnte nicht gespeichert werden.',
+    'audio.error.spectrogram': 'Das Spektrogramm für diesen Bereich konnte nicht gezeichnet werden.',
     'audio.info.duration': 'Dauer', 'audio.info.sampleRate': 'Abtastrate', 'audio.info.channels': 'Kanäle',
     'audio.info.channelLevels': 'L Spitze {leftPeak}, RMS {leftRms} | R Spitze {rightPeak}, RMS {rightRms}',
     'audio.info.bitDepth': 'Bittiefe', 'audio.info.format': 'Format', 'audio.info.fileSize': 'Dateigröße',
@@ -21,6 +25,7 @@ export const CATALOG_DE = {
     'audio.error.decode': 'Der Browser konnte diesen Audio-Codec nicht dekodieren. Verwenden Sie einen Host mit dem erforderlichen Codec oder transkodieren Sie die Datei.',
     'audio.warning.analysisFailed': 'Die Wellenform-Voranalyse ist fehlgeschlagen — es wird auf die vollständige Browser-Dekodierung zurückgegriffen, die bei einer Datei dieser Größe langsam sein kann.',
     'audio.warning.analysisSkipped': 'Die Wellenform-Voranalyse ist für dieses Format nicht verfügbar — es wird auf die vollständige Browser-Dekodierung zurückgegriffen, die bei einer Datei dieser Größe langsam sein kann.',
+    'audio.warning.reducedFidelity': 'Zur Speicherersparnis mit reduzierter Qualität dekodiert; dem Spektrogramm fehlen die höchsten Frequenzen.',
     'audio.vis.unavailableLarge': 'Das Spektrogramm ist für große Dateien, die aus vorberechneten Spitzenwerten gerendert werden, nicht verfügbar.',
     'video.play': 'Wiedergabe', 'video.pause': 'Pause', 'video.stop': 'Stopp',
     'video.skipBack': '⏪ -{seconds} s', 'video.skipForward': '⏩ +{seconds} s',

package/dist/i18n/catalog.en.js

@@ -6,14 +6,18 @@
 export const CATALOG_EN = {
     'audio.title': 'Audio', 'video.title': 'Video', 'dbc.title': 'CAN database',
     'audio.play': 'Play', 'audio.pause': 'Pause', 'audio.stop': 'Stop',
-    'audio.loop': 'Loop', 'audio.clearRegions': 'Clear regions', 'audio.volume': 'Volume',
+    'audio.volume': 'Volume',
+    'audio.loop.region': 'Loop region', 'audio.loop.track': 'Loop track',
     'audio.region.start': 'Region start (s)', 'audio.region.end': 'Region end (s)', 'audio.region.duration': 'Region length (s)',
     'audio.zoom': 'Zoom', 'audio.zoomIn': 'Zoom in', 'audio.zoomOut': 'Zoom out', 'audio.zoomFit': 'Fit',
     'audio.visualization': 'View', 'audio.vis.waveform': 'Waveform', 'audio.vis.spectrogram': 'Spectrogram',
     'audio.vis.both': 'Both', 'audio.download': 'Download',
-    'audio.spectrogramScale': 'Scale',
+    'audio.spectrogramScale': 'Spectrogram scale',
     'audio.scale.linear': 'Linear', 'audio.scale.mel': 'Mel', 'audio.scale.bark': 'Bark', 'audio.scale.erb': 'ERB',
+    'audio.scale.logarithmic': 'Logarithmic',
+    'audio.frequencyRange': 'Frequency range', 'audio.frequency.full': 'Full', 'audio.frequency.limit': 'Up to {hz}',
     'audio.error.save': 'Saving the audio file failed.',
+    'audio.error.spectrogram': 'The spectrogram could not be rendered for this range.',
     'audio.info.duration': 'Duration', 'audio.info.sampleRate': 'Sample rate', 'audio.info.channels': 'Channels',
     'audio.info.channelLevels': 'L peak {leftPeak}, RMS {leftRms} | R peak {rightPeak}, RMS {rightRms}',
     'audio.info.bitDepth': 'Bit depth', 'audio.info.format': 'Format', 'audio.info.fileSize': 'File size',
@@ -24,6 +28,7 @@ export const CATALOG_EN = {
     'audio.error.decode': 'The browser could not decode this audio codec. Try a host with the required codec or transcode the file.',
     'audio.warning.analysisFailed': 'Waveform pre-analysis failed — falling back to a full browser decode, which may be slow for a file this large.',
     'audio.warning.analysisSkipped': 'Waveform pre-analysis is unavailable for this format — falling back to a full browser decode, which may be slow for a file this large.',
+    'audio.warning.reducedFidelity': 'Decoded at reduced quality to fit this file in memory; the spectrogram loses its highest frequencies.',
     'audio.vis.unavailableLarge': 'Spectrogram is unavailable for large files rendered from precomputed peaks.',
     'video.play': 'Play', 'video.pause': 'Pause', 'video.stop': 'Stop',
     'video.skipBack': '⏪ -{seconds}s', 'video.skipForward': '⏩ +{seconds}s',
@@ -210,6 +215,182 @@ export const CATALOG_EN = {
     'tflite.inspector.inputs': 'Inputs', 'tflite.inspector.outputs': 'Outputs',
     'tflite.inspector.intermediates': 'Intermediates', 'tflite.inspector.options': 'Options',
     'tflite.inspector.customOptions': 'Custom options', 'tflite.inspector.subgraphs': 'Referenced subgraphs',
+    'coreml.title': 'Core ML model', 'coreml.search': 'Search operations, operands, and weights…',
+    'coreml.copyJson': 'Copy JSON', 'coreml.graph': 'Graph', 'coreml.unnamed': 'unnamed',
+    'coreml.operations': 'Operations', 'coreml.io': 'Inputs & Outputs', 'coreml.weights': 'Weights',
+    'coreml.packageContents': 'Package', 'coreml.modelInfo': 'Model Info',
+    'coreml.package': 'Package', 'coreml.updatable': 'Updatable',
+    'coreml.specVersion': 'specification v{version}',
+    'coreml.rows': '{shown} / {total} rows shown', 'coreml.noMatches': 'No rows match the current search.',
+    'coreml.selectNode': 'Select an operation or value to inspect it.',
+    'coreml.graphLimited': 'Graph preview shows {shown} of {total} items.',
+    'coreml.graphEdgesLimited': 'Only the first {count} connections are shown.',
+    'coreml.moreItems': '{count} more…', 'coreml.optional': 'Optional', 'coreml.default': 'default',
+    'coreml.unknown': 'unknown', 'coreml.external': 'from enclosing scope',
+    'coreml.yes': 'Yes', 'coreml.no': 'No', 'coreml.rootItem': 'root model',
+    'coreml.blobPresent': '{count} tensor(s), {referenced} referenced of {size} in the package',
+    'coreml.blobAbsent': '{count} tensor(s), {referenced} referenced — the file is not available here',
+    'coreml.weightFiles': 'Weight files',
+    'coreml.summary.operations': 'Operations', 'coreml.summary.graphs': 'Graphs',
+    'coreml.summary.inputs': 'Inputs', 'coreml.summary.outputs': 'Outputs',
+    'coreml.summary.functions': 'Functions', 'coreml.summary.specVersion': 'Spec version',
+    'coreml.kind.operation': 'Operation', 'coreml.kind.layer': 'Layer', 'coreml.kind.stage': 'Pipeline stage',
+    'coreml.kind.input': 'Input', 'coreml.kind.output': 'Output', 'coreml.kind.state': 'State',
+    'coreml.kind.trainingInput': 'Training input', 'coreml.kind.manifestItem': 'Manifest item',
+    'coreml.kind.archiveMember': 'Archive member',
+    'coreml.storage.blob': 'Weight file', 'coreml.storage.inline': 'Inline',
+    'coreml.column.index': '#', 'coreml.column.name': 'Name', 'coreml.column.type': 'Type',
+    'coreml.column.kind': 'Kind', 'coreml.column.inputs': 'Inputs', 'coreml.column.outputs': 'Outputs',
+    'coreml.column.attributes': 'Attributes', 'coreml.column.weights': 'Weights',
+    'coreml.column.function': 'Function', 'coreml.column.typeShape': 'Type / Shape',
+    'coreml.column.flexibility': 'Flexibility', 'coreml.column.optional': 'Optional',
+    'coreml.column.description': 'Description', 'coreml.column.graph': 'Graph', 'coreml.column.node': 'Operation',
+    'coreml.column.parameter': 'Parameter', 'coreml.column.shape': 'Shape', 'coreml.column.elements': 'Elements',
+    'coreml.column.dataSize': 'Data size', 'coreml.column.storage': 'Storage',
+    'coreml.column.location': 'Location', 'coreml.column.path': 'Path',
+    'coreml.info.modelType': 'Model type', 'coreml.info.specVersion': 'Specification version',
+    'coreml.info.availability': 'OS availability', 'coreml.info.description': 'Description',
+    'coreml.info.author': 'Author', 'coreml.info.license': 'License', 'coreml.info.version': 'Version',
+    'coreml.info.updatable': 'On-device update', 'coreml.info.fileSize': 'File size',
+    'coreml.info.functions': 'Functions', 'coreml.info.defaultFunction': 'Default function',
+    'coreml.info.graphs': 'Graphs', 'coreml.info.classLabels': 'Class labels',
+    'coreml.info.packageVersion': 'Package format', 'coreml.info.modelPath': 'Model path',
+    'coreml.inspector.inputs': 'Inputs', 'coreml.inspector.outputs': 'Outputs',
+    'coreml.inspector.attributes': 'Attributes', 'coreml.inspector.weights': 'Weights',
+    'coreml.inspector.graphs': 'Referenced graphs',
+    'coreml.warning.specVersion': 'The model declares specification version {version}, which this viewer does not have an OS mapping for.',
+    'coreml.warning.noModelType': 'The model declares no model type.',
+    'coreml.warning.emptyGraph': 'The model contains no operations.',
+    'coreml.warning.emptyProgram': 'The ML Program declares no functions.',
+    'coreml.warning.programVersion': 'The ML Program declares version {version}; this viewer targets version 1.',
+    'coreml.warning.emptyPipeline': 'The pipeline contains no stages.',
+    'coreml.warning.pipelineDepth': 'Pipeline nesting deeper than {limit} levels was not read.',
+    'coreml.warning.unknownLayer': 'Layer {name} declares no recognized layer type.',
+    'coreml.warning.customLayers': '{count} custom layer(s) require a matching runtime registration: {names}',
+    'coreml.warning.blobMissing': '{count} weight file(s) the model references are not in this package: {names}',
+    'coreml.warning.blobUnavailable': '{count} weight file(s) live beside this specification and were not read: {names}',
+    'coreml.warning.manifestMissing': 'The package has no Manifest.json; the model was located by its conventional path.',
+    'coreml.warning.manifestInvalid': 'The package Manifest.json could not be read.',
+    'coreml.warning.manifestUnreadable': 'The package Manifest.json could not be extracted, so the package contents are not described.',
+    'coreml.warning.textTruncated': 'Some text fields were longer than this viewer displays and were cut short.',
+    'coreml.warning.archiveEntryLimit': 'Only the first {limit} archive entries were read.',
+    'coreml.warning.serializedModel': 'This is a private serialized model; Apple documents no schema for its contents.',
+    'coreml.warning.linkedModel': 'This model links to a separate model file that this viewer does not read.',
+    'openvino.title': 'OpenVINO IR model', 'openvino.search': 'Search layers, ports, constants, and attributes…',
+    'openvino.copyJson': 'Copy JSON', 'openvino.unnamed': 'unnamed', 'openvino.unknown': 'unknown', 'openvino.scalar': 'scalar',
+    'openvino.irVersion': 'IR v{version}', 'openvino.noBin': 'no .bin supplied', 'openvino.binSize': '.bin {size}',
+    'openvino.graph': 'Graph', 'openvino.layers': 'Layers', 'openvino.constants': 'Constants',
+    'openvino.io': 'Inputs & Outputs', 'openvino.modelInfo': 'Model Info',
+    'openvino.rows': '{shown} / {total} rows shown', 'openvino.noMatches': 'No rows match the current search.',
+    'openvino.selectLayer': 'Select a layer to inspect it.',
+    'openvino.graphLimited': 'Graph preview shows {shown} of {total} layers.',
+    'openvino.graphEdgesLimited': 'Only the first {count} connections are shown.',
+    'openvino.moreItems': '{count} more…',
+    'openvino.kind.input': 'Input', 'openvino.kind.output': 'Output',
+    'openvino.summary.layers': 'Layers', 'openvino.summary.edges': 'Edges', 'openvino.summary.operators': 'Operator types',
+    'openvino.summary.constants': 'Constants', 'openvino.summary.weights': 'Weights', 'openvino.summary.io': 'Inputs / outputs',
+    'openvino.column.id': 'ID', 'openvino.column.name': 'Name', 'openvino.column.type': 'Type', 'openvino.column.version': 'Opset',
+    'openvino.column.inputs': 'Inputs', 'openvino.column.outputs': 'Outputs', 'openvino.column.attributes': 'Attributes',
+    'openvino.column.rtInfo': 'Runtime info', 'openvino.column.kind': 'Kind', 'openvino.column.shape': 'Shape',
+    'openvino.column.elements': 'Elements', 'openvino.column.offset': 'Offset', 'openvino.column.dataSize': 'Data size',
+    'openvino.column.status': 'Status', 'openvino.column.preview': 'Preview', 'openvino.column.tensorNames': 'Tensor names',
+    'openvino.column.typeShape': 'Type / Shape',
+    'openvino.status.unchecked': 'Not checked', 'openvino.status.available': 'In .bin',
+    'openvino.status.out-of-range': 'Beyond .bin', 'openvino.status.invalid': 'Invalid range',
+    'openvino.info.name': 'Network name', 'openvino.info.irVersion': 'IR version', 'openvino.info.opsets': 'Opsets',
+    'openvino.info.xmlSize': 'XML size', 'openvino.info.binSize': '.bin size', 'openvino.info.referencedBytes': 'Referenced weight bytes',
+    'openvino.info.layers': 'Layers', 'openvino.info.edges': 'Edges', 'openvino.info.operator': 'Operator',
+    'openvino.inspector.inputs': 'Inputs', 'openvino.inspector.outputs': 'Outputs', 'openvino.inspector.attributes': 'Attributes',
+    'openvino.inspector.weights': 'Weights', 'openvino.inspector.range': 'offset {offset} · {size}',
+    'openvino.inspector.preview': 'First {count} values', 'openvino.inspector.inputPort': 'input port', 'openvino.inspector.outputPort': 'output port',
+    'openvino.inspector.bodies': 'Sub-graphs', 'openvino.inspector.body': '{kind} · {layers} layers · {edges} edges',
+    'openvino.warning.noVersion': 'The <net> root declares no IR version.',
+    'openvino.warning.legacyVersion': 'IR version {version} predates the opset-based v10 layout; the legacy blobs were read on a best-effort basis.',
+    'openvino.warning.noLayers': 'The network contains no layers.',
+    'openvino.warning.layersLimited': 'Only the first {shown} of {total} layers were read.',
+    'openvino.warning.edgesLimited': 'Only the first {shown} of {total} edges were read.',
+    'openvino.warning.metadataLimited': '{count} metadata entries were not read.',
+    'openvino.warning.bodyLayersLimited': '{count} layer(s) inside control-flow sub-graphs were not read because the layer limit was reached.',
+    'openvino.warning.bodyEdgesLimited': '{count} edge(s) inside control-flow sub-graphs were not read because the edge limit was reached.',
+    'openvino.warning.bodiesLimited': '{count} control-flow sub-graph(s) beyond the per-layer limit were not kept.',
+    'openvino.warning.truncated': 'The XML ends before its root element closes; the topology may be incomplete.',
+    'openvino.warning.danglingEdges': '{count} edge(s) reference a layer that does not exist.',
+    'openvino.warning.invalidConstants': '{count} constant(s) declare an unreadable offset or size.',
+    'openvino.warning.noWeights': 'No .bin was supplied, so constants are shown by range only.',
+    'openvino.warning.weightsOutOfRange': '{count} constant(s) reach beyond the {size} .bin; it may belong to a different model.',
+    'openvino.warning.weightsUnreferenced': '{bytes} of the .bin are not referenced by any constant.',
+    'pt2.title': 'PyTorch Export program (.pt2)', 'pt2.eyebrowLegacy': 'PyTorch Export · legacy archive',
+    'pt2.search': 'Search nodes, values, weights, and modules…', 'pt2.selectModel': 'Model', 'pt2.copyJson': 'Copy JSON',
+    'pt2.schema': 'schema {version}', 'pt2.scalar': 'scalar', 'pt2.rootModule': '(root)', 'pt2.preserved': 'preserved',
+    'pt2.persistent': 'persistent', 'pt2.nonPersistent': 'non-persistent',
+    'pt2.graph': 'Graph', 'pt2.nodes': 'Nodes', 'pt2.weights': 'Weights & Constants', 'pt2.io': 'Inputs & Outputs',
+    'pt2.modules': 'Modules', 'pt2.modelInfo': 'Model Info', 'pt2.archive': 'Archive', 'pt2.extraFiles': 'Extra files', 'pt2.files': 'Files',
+    'pt2.rows': '{shown} / {total} rows shown', 'pt2.noMatches': 'No rows match the current search.',
+    'pt2.selectNode': 'Select a node or value to inspect it.', 'pt2.emptyGraph': 'The graph has no nodes.',
+    'pt2.graphLimited': 'Graph preview shows {shown} of {total} elements.',
+    'pt2.graphEdgesLimited': 'Only the first {count} connections are shown.', 'pt2.moreItems': '{count} more…',
+    'pt2.side.input': 'Input', 'pt2.side.output': 'Output',
+    'pt2.kind.user_input': 'User input', 'pt2.kind.parameter': 'Parameter', 'pt2.kind.buffer': 'Buffer',
+    'pt2.kind.tensor_constant': 'Tensor constant', 'pt2.kind.custom_obj': 'Custom object', 'pt2.kind.opaque_obj': 'Opaque object',
+    'pt2.kind.token': 'Token', 'pt2.kind.constant_input': 'Constant input', 'pt2.kind.user_output': 'User output',
+    'pt2.kind.loss_output': 'Loss output', 'pt2.kind.buffer_mutation': 'Buffer mutation', 'pt2.kind.parameter_mutation': 'Parameter mutation',
+    'pt2.kind.gradient_to_parameter': 'Gradient to parameter', 'pt2.kind.gradient_to_user_input': 'Gradient to user input',
+    'pt2.kind.user_input_mutation': 'User input mutation',
+    'pt2.status.raw': 'Raw bytes', 'pt2.status.pickled': 'Pickled (torch.save)', 'pt2.status.missing': 'Missing from archive',
+    'pt2.status.truncated': 'Shorter than tensor', 'pt2.status.unlisted': 'No payload listed',
+    'pt2.layout.pt2': 'PT2 archive', 'pt2.layout.legacy': 'Legacy torch.export.save layout',
+    'pt2.category.model': 'Program', 'pt2.category.weights': 'Weights', 'pt2.category.constants': 'Constants',
+    'pt2.category.sampleInputs': 'Sample inputs', 'pt2.category.aotinductor': 'AOTInductor', 'pt2.category.extra': 'Extra',
+    'pt2.category.archive': 'Archive marker', 'pt2.category.other': 'Other',
+    'pt2.column.index': '#', 'pt2.column.name': 'Name', 'pt2.column.op': 'Operator', 'pt2.column.module': 'Module',
+    'pt2.column.inputs': 'Inputs', 'pt2.column.outputs': 'Outputs', 'pt2.column.graph': 'Sub-graph', 'pt2.column.kind': 'Kind',
+    'pt2.column.placeholder': 'Placeholder', 'pt2.column.dtype': 'Dtype', 'pt2.column.shape': 'Shape', 'pt2.column.elements': 'Elements',
+    'pt2.column.bytes': 'Tensor bytes', 'pt2.column.path': 'Path', 'pt2.column.fileSize': 'File size', 'pt2.column.status': 'Status',
+    'pt2.column.preview': 'Preview', 'pt2.column.side': 'Side', 'pt2.column.value': 'Value', 'pt2.column.target': 'Target',
+    'pt2.column.typeShape': 'Type / Shape', 'pt2.column.detail': 'Detail', 'pt2.column.class': 'Class', 'pt2.column.nodes': 'Nodes',
+    'pt2.column.totalNodes': 'Nodes incl. children', 'pt2.column.signature': 'Call signature', 'pt2.column.forwardArgs': 'Forward args',
+    'pt2.column.category': 'Category', 'pt2.column.compressedSize': 'Compressed', 'pt2.column.method': 'Method',
+    'pt2.info.model': 'Model', 'pt2.info.torchVersion': 'PyTorch version', 'pt2.info.schemaVersion': 'Schema version',
+    'pt2.info.opsets': 'Opsets', 'pt2.info.verifiers': 'Verifiers', 'pt2.info.nodes': 'Nodes', 'pt2.info.values': 'Values',
+    'pt2.info.parameters': 'Parameters', 'pt2.info.weightBytes': 'Weight bytes', 'pt2.info.sampleInputs': 'Sample inputs',
+    'pt2.info.rangeConstraint': 'Range constraint', 'pt2.info.guard': 'Guard', 'pt2.info.operator': 'Operator',
+    'pt2.info.layout': 'Layout', 'pt2.info.fileSize': 'File size', 'pt2.info.prefix': 'Package folder', 'pt2.info.models': 'Programs',
+    'pt2.info.aotInductor': 'AOTInductor model',
+    'pt2.inspector.inputs': 'Inputs', 'pt2.inspector.outputs': 'Outputs', 'pt2.inspector.subgraphs': 'Sub-graphs',
+    'pt2.inspector.subgraph': '{name} · {nodes} nodes · {inputs} inputs · {outputs} outputs',
+    'pt2.inspector.moduleStack': 'Module stack', 'pt2.inspector.metadata': 'Metadata', 'pt2.inspector.stackTrace': 'Stack trace',
+    'pt2.inspector.value': 'Value', 'pt2.inspector.producer': 'Producer', 'pt2.inspector.consumers': 'Consumers', 'pt2.inspector.payload': 'Payload',
+    'pt2.inspector.fileSize': 'file {size}', 'pt2.inspector.tensorBytes': 'tensor {size}', 'pt2.inspector.preview': 'First {count} values',
+    'pt2.inspector.strides': 'Strides', 'pt2.inspector.storageOffset': 'Storage offset', 'pt2.inspector.device': 'Device',
+    'pt2.inspector.layout': 'Layout', 'pt2.inspector.requiresGrad': 'Requires grad',
+    'pt2.summary.models': 'Programs', 'pt2.summary.nodes': 'Nodes', 'pt2.summary.operators': 'Operator types',
+    'pt2.summary.parameters': 'Parameters', 'pt2.summary.weights': 'Weights', 'pt2.summary.constants': 'Constants',
+    'pt2.summary.io': 'Inputs / outputs', 'pt2.summary.aotInductor': 'AOTInductor models',
+    'pt2.summary.files': 'Archive members', 'pt2.summary.fileSize': 'File size',
+    'pt2.warning.notZip': 'The file is not a ZIP archive, so it cannot be a PT2 package.',
+    'pt2.warning.archiveUnreadable': 'The ZIP central directory could not be read.',
+    'pt2.warning.entryLimit': 'Only the first {limit} archive members were read.',
+    'pt2.warning.legacyLayout': 'This is the pre-PT2 torch.export.save layout: weights are torch.save pickles and are listed from the graph signature only.',
+    'pt2.warning.noArchiveFormat': 'No archive_format marker was found; the package layout was inferred.',
+    'pt2.warning.modelsLimited': 'Only the first {shown} of {total} programs were read.',
+    'pt2.warning.unknownArchiveFormat': 'archive_format is "{value}" rather than "pt2".',
+    'pt2.warning.memberTooLarge': '{name} exceeds the {limit} limit and was not read.',
+    'pt2.warning.memberMethod': '{name} uses ZIP compression method {method}, which is not supported.',
+    'pt2.warning.memberUnreadable': '{name} could not be read from the archive.',
+    'pt2.warning.inflateUnavailable': '{name} is deflated and no inflater (jszip) is available.',
+    'pt2.warning.jsonUnreadable': '{name} is not valid JSON.',
+    'pt2.warning.modelUnreadable': 'The program "{name}" could not be read.',
+    'pt2.warning.schemaVersion': 'Program "{name}" was serialized with schema {version}, newer than the {supported}.x this reader targets; some fields may be missing.',
+    'pt2.warning.nodesLimited': '{count} node(s) were not read because the node limit was reached.',
+    'pt2.warning.valuesLimited': '{count} value(s) were not read because the value limit was reached.',
+    'pt2.warning.payloadsLimited': '{count} weight or constant entries were not read because the limit was reached.',
+    'pt2.warning.danglingRefs': '{count} argument(s) reference a value that no node or input produces.',
+    'pt2.warning.subgraphDepth': '{count} sub-graph(s) nested deeper than {limit} levels were not read.',
+    'pt2.warning.pickledPayloads': 'Program "{name}": {count} tensor(s) are stored as torch.save pickles and are listed without a value preview.',
+    'pt2.warning.payloadMissing': 'Program "{name}": {count} weight file(s) referenced by the payload config are not in the archive.',
+    'pt2.warning.payloadTruncated': 'Program "{name}": {count} weight file(s) are shorter than the tensor they hold.',
+    'pt2.warning.aotiOnly': 'The package holds only AOTInductor compiled artifacts; there is no Export IR graph to show.',
+    'pt2.warning.noModels': 'The package holds no exported program (models/*.json).',
     'keras.title': 'Keras model', 'keras.document.v3': 'Keras v3 archive', 'keras.document.hdf5': 'Keras HDF5 model',
     'keras.version': 'Keras {version}', 'keras.search': 'Search layers, weights, and configuration…',
     'keras.copyJson': 'Copy JSON', 'keras.invalid': 'invalid',
@@ -449,6 +630,7 @@ export const CATALOG_EN = {
     'common.copied': 'Copied to clipboard',
     'common.noClipboard': 'Copy is unavailable on this platform',
     'common.copyTooLarge': 'Too large to copy to the clipboard ({size} bytes). Use file export instead.',
+    'common.copyTooLargeNoExport': 'Too large to copy to the clipboard ({size} bytes).',
     'common.noFileSave': 'Saving files is unavailable on this platform',
     'common.saved': 'Saved {name}',
     'common.savedToOriginal': 'Saved',

package/dist/i18n/catalog.fr.js

@@ -3,14 +3,18 @@
 export const CATALOG_FR = {
     'audio.title': 'Audio', 'video.title': 'Vidéo', 'dbc.title': 'Base de données CAN',
     'audio.play': 'Lecture', 'audio.pause': 'Pause', 'audio.stop': 'Arrêt',
-    'audio.loop': 'Boucle', 'audio.clearRegions': 'Effacer les régions', 'audio.volume': 'Volume',
+    'audio.volume': 'Volume',
+    'audio.loop.region': 'Boucler la région', 'audio.loop.track': 'Boucler la piste',
     'audio.region.start': 'Début de la région (s)', 'audio.region.end': 'Fin de la région (s)', 'audio.region.duration': 'Durée de la région (s)',
     'audio.zoom': 'Zoom', 'audio.zoomIn': 'Zoom avant', 'audio.zoomOut': 'Zoom arrière', 'audio.zoomFit': 'Ajuster',
     'audio.visualization': 'Affichage', 'audio.vis.waveform': 'Forme d’onde', 'audio.vis.spectrogram': 'Spectrogramme',
     'audio.vis.both': 'Les deux', 'audio.download': 'Télécharger',
-    'audio.spectrogramScale': 'Échelle',
+    'audio.spectrogramScale': 'Échelle du spectrogramme',
     'audio.scale.linear': 'Linéaire', 'audio.scale.mel': 'Mel', 'audio.scale.bark': 'Bark', 'audio.scale.erb': 'ERB',
+    'audio.scale.logarithmic': 'Logarithmique',
+    'audio.frequencyRange': 'Plage de fréquences', 'audio.frequency.full': 'Totale', 'audio.frequency.limit': "Jusqu'à {hz}",
     'audio.error.save': 'L’enregistrement du fichier audio a échoué.',
+    'audio.error.spectrogram': 'Le spectrogramme de cette plage n’a pas pu être rendu.',
     'audio.info.duration': 'Durée', 'audio.info.sampleRate': 'Fréquence d’échantillonnage', 'audio.info.channels': 'Canaux',
     'audio.info.channelLevels': 'L crête {leftPeak}, RMS {leftRms} | R crête {rightPeak}, RMS {rightRms}',
     'audio.info.bitDepth': 'Profondeur de bits', 'audio.info.format': 'Format', 'audio.info.fileSize': 'Taille du fichier',
@@ -21,6 +25,7 @@ export const CATALOG_FR = {
     'audio.error.decode': 'Le navigateur n’a pas pu décoder ce codec audio. Essayez un hôte disposant du codec requis ou transcodez le fichier.',
     'audio.warning.analysisFailed': 'L’analyse préalable de la forme d’onde a échoué — retour au décodage complet par le navigateur, qui peut être lent pour un fichier de cette taille.',
     'audio.warning.analysisSkipped': 'L’analyse préalable de la forme d’onde n’est pas disponible pour ce format — retour au décodage complet par le navigateur, qui peut être lent pour un fichier de cette taille.',
+    'audio.warning.reducedFidelity': 'Décodé en qualité réduite pour tenir en mémoire ; le spectrogramme perd ses fréquences les plus hautes.',
     'audio.vis.unavailableLarge': 'Le spectrogramme n’est pas disponible pour les fichiers volumineux rendus à partir de pics précalculés.',
     'video.play': 'Lecture', 'video.pause': 'Pause', 'video.stop': 'Arrêt',
     'video.skipBack': '⏪ -{seconds} s', 'video.skipForward': '⏩ +{seconds} s',

package/dist/i18n/catalog.it.js

@@ -3,14 +3,18 @@
 export const CATALOG_IT = {
     'audio.title': 'Audio', 'video.title': 'Video', 'dbc.title': 'Database CAN',
     'audio.play': 'Riproduci', 'audio.pause': 'Pausa', 'audio.stop': 'Arresta',
-    'audio.loop': 'Ripeti', 'audio.clearRegions': 'Cancella regioni', 'audio.volume': 'Volume',
+    'audio.volume': 'Volume',
+    'audio.loop.region': 'Ripeti regione', 'audio.loop.track': 'Ripeti traccia',
     'audio.region.start': 'Inizio regione (s)', 'audio.region.end': 'Fine regione (s)', 'audio.region.duration': 'Durata regione (s)',
     'audio.zoom': 'Zoom', 'audio.zoomIn': 'Ingrandisci', 'audio.zoomOut': 'Riduci', 'audio.zoomFit': 'Adatta',
     'audio.visualization': 'Vista', 'audio.vis.waveform': 'Forma d’onda', 'audio.vis.spectrogram': 'Spettrogramma',
     'audio.vis.both': 'Entrambi', 'audio.download': 'Scarica',
-    'audio.spectrogramScale': 'Scala',
+    'audio.spectrogramScale': 'Scala dello spettrogramma',
     'audio.scale.linear': 'Lineare', 'audio.scale.mel': 'Mel', 'audio.scale.bark': 'Bark', 'audio.scale.erb': 'ERB',
+    'audio.scale.logarithmic': 'Logaritmica',
+    'audio.frequencyRange': 'Intervallo di frequenza', 'audio.frequency.full': 'Completo', 'audio.frequency.limit': 'Fino a {hz}',
     'audio.error.save': 'Salvataggio del file audio non riuscito.',
+    'audio.error.spectrogram': 'Non è stato possibile disegnare lo spettrogramma per questo intervallo.',
     'audio.info.duration': 'Durata', 'audio.info.sampleRate': 'Frequenza di campionamento', 'audio.info.channels': 'Canali',
     'audio.info.channelLevels': 'L picco {leftPeak}, RMS {leftRms} | R picco {rightPeak}, RMS {rightRms}',
     'audio.info.bitDepth': 'Profondità di bit', 'audio.info.format': 'Formato', 'audio.info.fileSize': 'Dimensione file',
@@ -21,6 +25,7 @@ export const CATALOG_IT = {
     'audio.error.decode': 'Il browser non è riuscito a decodificare questo codec audio. Prova un host con il codec richiesto o transcodifica il file.',
     'audio.warning.analysisFailed': 'L’analisi preliminare della forma d’onda non è riuscita — si torna alla decodifica completa del browser, che può essere lenta per un file di queste dimensioni.',
     'audio.warning.analysisSkipped': 'L’analisi preliminare della forma d’onda non è disponibile per questo formato — si torna alla decodifica completa del browser, che può essere lenta per un file di queste dimensioni.',
+    'audio.warning.reducedFidelity': 'Decodificato a qualità ridotta per rientrare in memoria; lo spettrogramma perde le frequenze più alte.',
     'audio.vis.unavailableLarge': 'Lo spettrogramma non è disponibile per i file di grandi dimensioni renderizzati da picchi precalcolati.',
     'video.play': 'Riproduci', 'video.pause': 'Pausa', 'video.stop': 'Arresta',
     'video.skipBack': '⏪ -{seconds} s', 'video.skipForward': '⏩ +{seconds} s',