Gigaam v3 Rnnt Coreml
This is a Core ML conversion of v3rnnt from ai-sage/GigaAM-v3. It contains separate encoder, predictor, and joiner ML Programs.
Model source
- Kind
- Other
- File size
- 1.37–842 MB
- Base model
- ai-sage/GigaAM-v3
- Version
- vRevision 6a8eb79
- Updated
- Oct 31, 2026
- Downloads
- 8
- Source
- Hugging Face
Source description
GigaAM v3_rnnt Core ML staged artifact
This is a Core ML conversion of v3_rnnt from ai-sage/GigaAM-v3. It contains separate encoder, predictor, and joiner ML Programs.
Provenance
repository: https://github.com/salute-developers/GigaAM
revision: 559d88d6b72541412743929f633a6ae7c9950b85
variant: v3_rnnt
- Conversion precision:
float32 - Upstream license:
MIT
Runtime contract
The artifact uses fixed batch-one Core ML inputs for Russian audio of at most 25 seconds. See model-config.json and conversion-manifest.json for the exact tensor and checksum contracts.
This staged artifact must pass its parity validation before publication.
Sources
1 source
Sources
1 sourceVerified Oct 31
Download files
Repository files
File inventory only; completeness is unverified.
- conversion-manifest.json
- encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
- encoder.mlpackage/Manifest.json
- joiner.mlpackage/Data/com.apple.CoreML/weights/weight.bin
- joiner.mlpackage/Manifest.json
- model-config.json
- predictor.mlpackage/Data/com.apple.CoreML/weights/weight.bin
- predictor.mlpackage/Manifest.json
- tokenizer.json
- validation-report.json
Model artifacts
3 artifacts
Model artifacts
3 artifactsencoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin
bin · 842 MB · SHA-256 36c0c9c11f4e…940e · Hugging Face
Download filejoiner.mlpackage/Data/com.apple.CoreML/weights/weight.bin
bin · 1.37 MB · SHA-256 86aa5a2ee3c0…d873 · Hugging Face
Download filepredictor.mlpackage/Data/com.apple.CoreML/weights/weight.bin
bin · 3.17 MB · SHA-256 dac169601fdf…1dbe · Hugging Face
Download fileSource excerpts
3 excerpts
Source excerpts
3 excerpts--- license: mit language: - ru pipeline_tag: automatic-speech-recognition library_name: coreml base_model: ai-sage/GigaAM-v3 tags: - coreml - gigaam - speech-recognition - macos - rnnt - v3_rnnt --- # GigaAM v3_rnnt Core ML staged artifact This is a Core ML conversion of `v3_rnnt` from [ai-sage/GigaAM-v3](https://huggingface.co/ai-sage/GigaAM-v3). It contains separate encoder, predictor, and joiner ML Programs. ## Provenance ```yaml repository: https://github.com/salute-developers/GigaAM revision: 559d88d6b72541412743929f633a6ae7c9950b85 variant: v3_rnnt ``` - Conversion precision: `float32` - Upstream license: `MIT` ## Runtime contract The artifact uses fixed batch-one Core ML inputs for Russian audio of at most 25 seconds. See `model-config.json` and `conversion-manifest.json` for the exact tensor and checksum contracts. This staged artifact must pass its parity validation before publication.
Source context: 8 downloads · 0 likes · Pipeline automatic-speech-recognition · Library coreml · Repo vadimsuhanov/gigaam-v3-rnnt-coreml
Source context: 6 downloads · 0 likes · Pipeline automatic-speech-recognition · Library coreml · Repo vadimsuhanov/gigaam-v3-rnnt-coreml
