From 9b4e9cc8a0311e5243d69b73ed073e7ea441982e Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期三, 27 三月 2024 16:05:29 +0800
Subject: [PATCH] train update
---
funasr/models/conformer/template.yaml | 14 +++++++-------
1 files changed, 7 insertions(+), 7 deletions(-)
diff --git a/funasr/models/conformer/template.yaml b/funasr/models/conformer/template.yaml
index 4cbeca4..5a8687e 100644
--- a/funasr/models/conformer/template.yaml
+++ b/funasr/models/conformer/template.yaml
@@ -6,8 +6,7 @@
# tables.print()
# network architecture
-#model: funasr.models.paraformer.model:Paraformer
-model: Transformer
+model: Conformer
model_conf:
ctc_weight: 0.3
lsm_weight: 0.1 # label smoothing option
@@ -16,14 +15,14 @@
# encoder
encoder: ConformerEncoder
encoder_conf:
- output_size: 256 # dimension of attention
+ output_size: 256
attention_heads: 4
- linear_units: 2048 # the number of units of position-wise feed forward
- num_blocks: 12 # the number of encoder blocks
+ linear_units: 2048
+ num_blocks: 12
dropout_rate: 0.1
positional_dropout_rate: 0.1
attention_dropout_rate: 0.0
- input_layer: conv2d # encoder architecture type
+ input_layer: conv2d
normalize_before: true
pos_enc_layer_type: rel_pos
selfattention_layer_type: rel_selfattn
@@ -52,6 +51,7 @@
n_mels: 80
frame_length: 25
frame_shift: 10
+ dither: 0.0
lfr_m: 1
lfr_n: 1
@@ -95,7 +95,7 @@
dataset: AudioDataset
dataset_conf:
index_ds: IndexDSJsonl
- batch_sampler: DynamicBatchLocalShuffleSampler
+ batch_sampler: BatchSampler
batch_type: example # example or length
batch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;
max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,
--
Gitblit v1.9.1