-
SpeechLM2: S2S->S2T data reader, excessive padding fixes (#14124)
发布于
2025-08-02 01:29:58 +00:00 - S2S data to audio-input/text-output multimodal conversation converter
Signed-off-by: Piotr Żelasko petezor@gmail.com
- WIP: make multimodal conversation data readable from NeMo ASR tar files
Signed-off-by: Piotr Żelasko petezor@gmail.com
- fix
Signed-off-by: Piotr Żelasko petezor@gmail.com
- loosen duration check, auto-propagate audio locator tag
Signed-off-by: Piotr Żelasko petezor@gmail.com
- fix edge case in tar conversation
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
- Apply isort and black reformatting
Signed-off-by: pzelasko pzelasko@users.noreply.github.com
- Parsing nested icfgs; fixes in parsing multimodal conversation
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
- Remove excessive padding in SALM audio placeholder replacement step
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
- skipme support for nemo multimodal conversation
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Apply isort and black reformatting
Signed-off-by: pzelasko pzelasko@users.noreply.github.com
- add a test to validate padding removal in placeholder replacement fn
Signed-off-by: Piotr Żelasko petezor@gmail.com
- fix
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Fix measure_total_length handling for bucketing constraints
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Proper dtype conversion in HF export and extra system prompt support for multi-turn conversation parsers
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Fix for reading corrupted audio files
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
- Remove prompt format fn override no longer needed
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
- Fix text-only-mode padding for SALM
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Fix CI
Signed-off-by: Piotr Żelasko petezor@gmail.com
- Potential fix for code scanning alert no. 15468: Except block handles 'BaseException'
Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
Signed-off-by: Piotr Żelasko petezor@gmail.com- Fixes
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
Signed-off-by: Piotr Żelasko petezor@gmail.com
Signed-off-by: Piotr Żelasko pzelasko@nvidia.com
Signed-off-by: pzelasko pzelasko@users.noreply.github.com
Co-authored-by: pzelasko pzelasko@users.noreply.github.com
Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>下载附件