From d147c2901569771e88f4c27ccf9d5eb1e96ef803 Mon Sep 17 00:00:00 2001 From: dohe0342 Date: Thu, 26 Jan 2023 14:40:45 +0900 Subject: [PATCH] from local --- .../.train_adapter.py.swp | Bin 90112 -> 90112 bytes .../train_adapter.py | 3 ++- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train_adapter.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train_adapter.py.swp index eee1565b6b95a9a793ce5b9e9624adbb059a3e65..15e53574e1ba1e09179227d179b825eb8a75dba4 100644 GIT binary patch delta 159 zcmZoTz}j$tRW!*U%+puFQqO<^2m}}yf`u+6U)m^oi-mlK0C2RIxR6LM*##FKtx)8kwR%nNqlCCLT29PoJmv80sz$$Eaw0K diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py index 0e17a6c10..9c5bd69a6 100755 --- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py +++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train_adapter.py @@ -835,7 +835,7 @@ def compute_loss( #texts = batch["supervisions"]["text"] texts = [] - for utt_id in + for utt_id in token_ids = sp.encode(texts, out_type=int) y = k2.RaggedTensor(token_ids).to(device) @@ -1595,6 +1595,7 @@ def run_adapter(rank, world_size, args, wb=None): pl[text[0]] = ' '.join(text[1:]) pl_texts = pl + def remove_short_and_long_utt(c: Cut): return 1.0 <= c.duration <= 20.0