cyaaronk

Updated code. Features implemented: 1. batch processing 2. Automatic unpadding 3. String argument to specify module output to save. E.g. --output_modules=[encoder.encoders.*] 4. Use pytorch module forward hooks 5. Can save...

No problem :) > On 20 Oct 2022, at 1:10 AM, Shinji Watanabe ***@***.***> wrote: > >  > Thanks! > We're now busy with the ICASSP deadline. > Also,...

@simpleoier : Thanks for the comment! my current implementation can only precompute feats for asr. However, if we were to precompte other feats like speaker emb, I think we may...

As for an example to precompute features, please see the bash script I posted earlier Best regards, Aaron > On 30 Oct 2022, at 12:21 AM, Chin Yuen Aaron Aaron...

@ftshijt I am interested in contributing to your repo. Please invite me so we can collaborate :) Best regards, Aaron > On 6 Nov 2022, at 2:41 PM, Jiatong ***@***.***>...

Dear Sir @BenjaminBossan , may I know where can we find the results in the first figure? Which file saves the metrics?

@BenjaminBossan Thank you for the hard work and effort! I wonder if some initial results can be released first as I am testing some methods already but I am not...