Karthik Ganesan

Labels for CMU MOSEI raw dataset

3

Hi, Thank you for the great work. I am trying to build a feature extraction pipeline using ESPNet and CMU-MOSEI , I am unable to find the labels to the...

How to load a pre-trained and encoder and decoder ?

1

Hi I would like to load pre-trained conformer RNN-T . I would like to load both the pre-trained encoder and decoder . https://github.com/siddhu001/espnet/blob/936917515557de1dd8519d67c9b455670b166536/egs2/slurp_transcript/asr1/run.sh#L22 In the above I see we can...

Add wav2vec2 adapter to s3prl frontend

2

This PR adds adapters to wav2vec2 model by adapting s3prl.py An updated common voice recipe PR will be created with a sample config

Add Native onnx export support to encoders and CTC

8

This PR updates make_pad_mask and related functions to make ESPnet transformer/conformer encoder and CTC ONNX export friendly . This particular change would reduce maintenance efforts of espnet_onnx package to be...

Add unified S3PRL adapter interface

5

This PR adds 1. a unified adapter interface that should work with all the S3PRL front-ends that are build on top wav2vec2 like Transformerencoder . 2. Dependency related external library...

vllm +zero2 hangs

32

Team, thank you so much for this wonderful toolkit! we are trying to test the vllm setting with mistralai/Mistral-7B-Instruct-v0.2 model with zero2  ray job submit --address="http://127.0.0.1:8265" \ --runtime-env-json='{"working_dir": "/openrlhf",...

**🚀 Feature Description** Realtime voice conversion to build accent translation Want streaming support in both encoder and vocoder **Solution** A streaming phone recognizer / encoder word bound boundary prediction ,...

Support for HF models

2

Hi, do we support huggingface models? If any internal internal tests were done with llama and the insights can be shared that will be great