From 8a0930d682fe3206e0b41c694fc03d7d10c7eed2 Mon Sep 17 00:00:00 2001
From: shixian.shi <shixian.shi@alibaba-inc.com>
Date: 星期二, 10 十月 2023 11:35:42 +0800
Subject: [PATCH] paraformer-speaker inference pipeline
---
egs_modelscope/asr/TEMPLATE/README.md | 21 +++++++++++++--------
1 files changed, 13 insertions(+), 8 deletions(-)
diff --git a/egs_modelscope/asr/TEMPLATE/README.md b/egs_modelscope/asr/TEMPLATE/README.md
index cf0ba84..4cf6a7e 100644
--- a/egs_modelscope/asr/TEMPLATE/README.md
+++ b/egs_modelscope/asr/TEMPLATE/README.md
@@ -1,3 +1,5 @@
+([绠�浣撲腑鏂嘳(./README_zh.md)|English)
+
# Speech Recognition
> **Note**:
@@ -25,15 +27,18 @@
inference_pipeline = pipeline(
task=Tasks.auto_speech_recognition,
model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online',
- model_revision='v1.0.6',
+ model_revision='v1.0.7',
update_model=False,
mode='paraformer_streaming'
)
import soundfile
speech, sample_rate = soundfile.read("example/asr_example.wav")
-chunk_size = [5, 10, 5] #[5, 10, 5] 600ms, [8, 8, 4] 480ms
-param_dict = {"cache": dict(), "is_final": False, "chunk_size": chunk_size}
+chunk_size = [0, 10, 5] #[0, 10, 5] 600ms, [0, 8, 4] 480ms
+encoder_chunk_look_back = 4 #number of chunks to lookback for encoder self-attention
+decoder_chunk_look_back = 1 #number of encoder chunks to lookback for decoder cross-attention
+param_dict = {"cache": dict(), "is_final": False, "chunk_size": chunk_size,
+ "encoder_chunk_look_back": encoder_chunk_look_back, "decoder_chunk_look_back": decoder_chunk_look_back}
chunk_stride = chunk_size[1] * 960 # 600ms銆�480ms
# first chunk, 600ms
speech_chunk = speech[0:chunk_stride]
@@ -53,7 +58,7 @@
inference_pipeline = pipeline(
task=Tasks.auto_speech_recognition,
model='damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online',
- model_revision='v1.0.6',
+ model_revision='v1.0.7',
update_model=False,
mode="paraformer_fake_streaming"
)
@@ -230,10 +235,10 @@
- `batch_bins`: batch size. For dataset_type is `small`, `batch_bins` indicates the feature frames. For dataset_type is `large`, `batch_bins` indicates the duration in ms
- `max_epoch`: number of training epoch
- `lr`: learning rate
- - `init_param`: init model path, load modelscope model initialization by default. For example: ["checkpoint/20epoch.pb"]
- - `freeze_param`: Freeze model parameters. For example锛歔"encoder"]
- - `ignore_init_mismatch`: Ignore size mismatch when loading pre-trained model
- - `use_lora`: Fine-tuning model use lora, more detail please refer to [LORA](https://arxiv.org/pdf/2106.09685.pdf)
+ - `init_param`: `[]`(Default), init model path, load modelscope model initialization by default. For example: ["checkpoint/20epoch.pb"]
+ - `freeze_param`: `[]`(Default), Freeze model parameters. For example锛歔"encoder"]
+ - `ignore_init_mismatch`: `True`(Default), Ignore size mismatch when loading pre-trained model
+ - `use_lora`: `False`(Default), Fine-tuning model use lora, more detail please refer to [LORA](https://arxiv.org/pdf/2106.09685.pdf)
- Training data formats锛�
```sh
--
Gitblit v1.9.1