This commit is contained in:
JinZr 2023-08-24 13:19:48 +08:00
parent 7feaa6185d
commit 3672d23631

View File

@ -97,7 +97,7 @@ def compute_fbank_switchboard(
prefix = dir_name prefix = dir_name
suffix = "jsonl.gz" suffix = "jsonl.gz"
manifests = { manifests = {
"eval2000": "data/manifests/eval2000/eval2000_cuts_all.jsonl.gz", "eval2000": "data/manifests/eval2000/eval2000_cuts_all_trimmed.jsonl.gz",
} }
assert manifests is not None assert manifests is not None
@ -111,12 +111,7 @@ def compute_fbank_switchboard(
logging.info(f"{prefix} already exists - skipping.") logging.info(f"{prefix} already exists - skipping.")
return return
logging.info(f"Processing {prefix}") logging.info(f"Processing {prefix}")
cut_set = ( cut_set = CutSet.from_file(manifests[prefix]).resample(16000)
CutSet.from_file(manifests[prefix])
.resample(16000)
.to_eager()
.filter(lambda c: c.duration > 0.5)
)
cut_set = cut_set.compute_and_store_features( cut_set = cut_set.compute_and_store_features(
extractor=extractor, extractor=extractor,