Skip to content

NOTSOFAR prep fixes - #21

Open
Lakoc wants to merge 4 commits into
chimechallenge:mainfrom
Lakoc:main
Open

NOTSOFAR prep fixes#21
Lakoc wants to merge 4 commits into
chimechallenge:mainfrom
Lakoc:main

Conversation

@Lakoc

@Lakoc Lakoc commented Sep 23, 2025

Copy link
Copy Markdown

This PR updates the NOTSOFAR-1 dataset generation recipe with the following changes:

  • Replaced azcopy (which caused authentication errors) with huggingface-hub for downloading NOTSOFAR data.
  • Updated Lhotse version dependency.
  • Added generation of single-channel Lhotse manifests.

Intended use

Dataset generation

chime-utils dgen notsofar1 $DATA_DIR/nsf $DATA_DIR/notsofar --part="train,dev,eval"

Lhotse manifest preparation

# Single-channel manifests
chime-utils lhotse-prep notsofar1 \
  -d train_sc,dev_sc,eval_sc \
  --txt-norm none -m sdm \
  $DATA_DIR/notsofar $MANIFESTS_DIR

# Multi-channel manifests
chime-utils lhotse-prep notsofar1 \
  -d train,dev,eval \
  --txt-norm none -m mdm \
  $DATA_DIR/notsofar $MANIFESTS_DIR

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant