From 8934efe719467d23c34f0f23cc0da7230f2879e0 Mon Sep 17 00:00:00 2001 From: Hongji Wang Date: Tue, 20 Aug 2024 12:52:06 +0800 Subject: [PATCH] [docs] update docs/pretrained.md (#349) --- docs/pretrained.md | 2 ++ examples/voxceleb/v2/run_wavlm.sh | 2 +- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/docs/pretrained.md b/docs/pretrained.md index cac7bb21..7fdc3dd8 100644 --- a/docs/pretrained.md +++ b/docs/pretrained.md @@ -39,6 +39,8 @@ in [the voxconverse recipe](https://github.com/wenet-e2e/wespeaker/tree/master/e ## Model List +The model with suffix **LM** means that it is further fine-tuned using large-margin fine-tuning, which could perform better on long audios, e.g. >3s. + ### modelscope | Datasets | Languages | Checkpoint (pt) | Runtime Model (onnx) | diff --git a/examples/voxceleb/v2/run_wavlm.sh b/examples/voxceleb/v2/run_wavlm.sh index 9494ea16..4eb35a4e 100755 --- a/examples/voxceleb/v2/run_wavlm.sh +++ b/examples/voxceleb/v2/run_wavlm.sh @@ -12,7 +12,7 @@ data_type="shard" # shard/raw config=conf/ecapa_tdnn_WavLM_frozen.yaml exp_dir=exp/ECAPA_TDNN_GLOB_c512-ASTP-emb192-WavLM_large_frozen_num_frms150-aug0.6-spTrue-saFalse-ArcMargin_intertopk_subcenter-SGD-epoch150 -gpus="[0,1,2,3]" #,4,5,6,7]" +gpus="[0,1,2,3,4,5,6,7]" num_avg=10 checkpoint=