GigaAM v3 e2e RNNT ONNX

Prepared runtime artifacts derived from upstream ai-sage/GigaAM-v3 revision e2e_rnnt.

This repository contains exported ONNX runtime artifacts for the current GigaAM v3_e2e_rnnt line, plus INT8 variants and a converted tokenizer JSON. The source preparation flow uses a minimal vendored extraction of the upstream RNNT model code to avoid depending on the full upstream runtime stack during conversion and publication.

Repository ID: cnonim/gigaam-v3-e2e-rnnt-onnx

Included files:

  • onnx/v3_e2e_rnnt_encoder.onnx
  • onnx/v3_e2e_rnnt_decoder.onnx
  • onnx/v3_e2e_rnnt_encoder_int8.onnx
  • onnx/v3_e2e_rnnt_decoder_int8.onnx
  • onnx/v3_e2e_rnnt_tokenizer.json

Preparation notes:

  • based on upstream ai-sage/GigaAM-v3 model artifacts
  • exported to ONNX encoder/decoder runtime files
  • includes INT8-quantized encoder and decoder variants
  • includes tokenizer.json converted from the upstream SentencePiece tokenizer
  • uses a reduced vendored conversion path focused on the RNNT export surface

This repository is limited to the current v3_e2e_rnnt line. Future model lines are intentionally out of scope for this publication surface.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for cnonim/gigaam-v3-e2e-rnnt-onnx

Quantized
(4)
this model