Slyne Deng

potential memory leak

4

Hi~ I use valgrind to test the below line: ``` import kaldifeat ``` The log is here: https://1drv.ms/t/s!AhtoTbISXXlSgSLtqDWjo8RP_ouA?e=Sn1cd0 Log summary: ``` ==20510== LEAK SUMMARY: ==20510== definitely lost: 0 bytes in...

patch quantization

4

Update the code to be compatible with latest nemo. With batchsize=1, now the WER/CER is normal. But still, the exported onnx model doesn't contain seq_length, which affects models with ```conv_mask:True```...

IterableDataset with batch samples of different size

3

Hi, I want to train a large dataset with accelerate. However, my dataset has different shape (same batch, different sequence length) of samples. This [line](https://github.com/huggingface/accelerate/blob/b2956acbe925845bf202224797334cfe4da18396/src/accelerate/data_loader.py#L356) requires the big batch for...

How to built ctc_topo with blankid != 0

1

Hi experts, I was wondering if the build_ctc_topo interface can be used for blankid is not equal to 0? What's the correct way to build the ctc_topo in this situation?...

adding inference_mode() causes "Inference tensors do not track version counter"

2

Hi K2 experts, I'm trying to add the k2 fast beam decoder to my acoustic model. It works great with torch.inference_mode() set to false. But it raises the below error...

fast_beam_search_LG outputs empty for sentence with oov words

13

Hi developers, I'm not sure if this issue is from my side or not. I use k2 + fast_beam_search and can get good result. However, after adding the LG.pt to...

mixtral evaluation PR

3

Please refer to the previous PR: https://github.com/NVIDIA/NeMo/pull/8988

Update actions.py

2

fix minor bug "print will be never reached"

Small fix for Mixtral as judge eval pipeline

2

# What does this PR do ? This PR is a fix for https://github.com/NVIDIA/NeMo/pull/8634. **Collection**: No collection would be affected. We can run the script independently. And nemo dependency is...

[DRAFT] Add lita inference

1

# What does this PR do ? Notice this is only a draft. * Convert LITA ckpt to nemo model * Add LITA FAST & SLOW tokens * Add LITA...