Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Here are the TODO lists. Help from the community is appreciated.
@pkufool please have a look which part(s) you are interested in.
kaldifeat_core
fromk2_torch
. We can make the binaries depend onkaldifeat_core
directly.k2_torch
in https://github.com/k2-fsa/k2/blob/c3a7404e0d9da4d8d99f3f15c54d80b321c1b960/cmake/k2Config.cmake.in#L48torch_api.h
HLG.pt
. The returned type can bestd::shared_ptr<k2::FsaClass>
. We can define an aliasFsaClassPtr
for it, like RaggedShapePtrk2::CtcTopo()
. It should also return a value of typeFsaClassPtr
.- log_softmax_out: a 3-D tensor of shape (N, T, C)
- log_softmax_out_lens, a 1-D tensor of shape (N,)
- FsaClassPtr, can be either a CtcTopo or an HLG
and it returns
std::vector<std::vector<int32_t>>
- https://github.com/k2-fsa/k2/blob/v2.0-pre/k2/torch/bin/ctc_decode.cu
- https://github.com/k2-fsa/k2/blob/v2.0-pre/k2/torch/bin/hlg_decode.cu
We can use https://github.com/wenet-e2e/wenet/releases/download/v2.0.1/chs.tar.gz and https://github.com/wenet-e2e/wenet/releases/download/v2.0.1/en.tar.gz for testing.
I suggest that we also add scripts to show how to build an HLG graph for models from wenet. We can put the scripts
and models on huggingface.