lihongjie
commited on
Commit
·
39ad93d
1
Parent(s):
41b3743
update
Browse files
README.md
CHANGED
|
@@ -42,24 +42,28 @@ For those who are interested in model conversion, you can try to export axmodel
|
|
| 42 |
|
| 43 |
Download all files from this repository to the device
|
| 44 |
|
| 45 |
-
### 1.
|
| 46 |
|
| 47 |
-
####
|
| 48 |
|
| 49 |
-
####
|
| 50 |
|
| 51 |
**Running HTTP Tokenizer Server** and **Processing Prompt Speech** require these Python packages. If you run these two step on a PC, install them on the PC.
|
| 52 |
```
|
| 53 |
pip3 install -r scripts/requirements.txt
|
| 54 |
```
|
| 55 |
|
| 56 |
-
|
| 57 |
```
|
| 58 |
cd scripts
|
| 59 |
python cosyvoice2_tokenizer.py --host {your host} --port {your port}
|
| 60 |
```
|
| 61 |
|
| 62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
1) Moidfy the HTTP host in `run_ax650.sh`.
|
| 64 |
|
| 65 |
2) Run `run_ax650.sh`
|
|
@@ -127,17 +131,42 @@ Output Speech:
|
|
| 127 |
[output.wav](asset/output.wav)
|
| 128 |
|
| 129 |
|
| 130 |
-
####
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
If you want to replicate a specific sound, do this step.
|
| 132 |
You can use audio in asset/ .
|
| 133 |
|
| 134 |
-
|
| 135 |
```
|
| 136 |
pip3 install modelscope
|
| 137 |
modelscope download --model pengzhendong/wetext --local_dir pengzhendong/wetext
|
| 138 |
```
|
| 139 |
|
| 140 |
-
|
| 141 |
Example:
|
| 142 |
```
|
| 143 |
python3 scripts/process_prompt.py --prompt_text asset/zh_man1.txt --prompt_speech asset/zh_man1.wav --output zh_man1
|
|
|
|
| 42 |
|
| 43 |
Download all files from this repository to the device
|
| 44 |
|
| 45 |
+
### 1. PrePare
|
| 46 |
|
| 47 |
+
#### 1.1 Copy this project to AX650 Board
|
| 48 |
|
| 49 |
+
#### 1.2 Prepare Dependencies
|
| 50 |
|
| 51 |
**Running HTTP Tokenizer Server** and **Processing Prompt Speech** require these Python packages. If you run these two step on a PC, install them on the PC.
|
| 52 |
```
|
| 53 |
pip3 install -r scripts/requirements.txt
|
| 54 |
```
|
| 55 |
|
| 56 |
+
### 2. Start HTTP Tokenizer Server
|
| 57 |
```
|
| 58 |
cd scripts
|
| 59 |
python cosyvoice2_tokenizer.py --host {your host} --port {your port}
|
| 60 |
```
|
| 61 |
|
| 62 |
+
|
| 63 |
+
### 3. Run on Axera Device
|
| 64 |
+
There are 2 kinds of device, AX650 Board and AXCL aarch64 Board.
|
| 65 |
+
|
| 66 |
+
#### 3.1 Run on AX650 Board
|
| 67 |
1) Moidfy the HTTP host in `run_ax650.sh`.
|
| 68 |
|
| 69 |
2) Run `run_ax650.sh`
|
|
|
|
| 131 |
[output.wav](asset/output.wav)
|
| 132 |
|
| 133 |
|
| 134 |
+
#### Or run on AX650 Board with Gradio GUI
|
| 135 |
+
1) Start server
|
| 136 |
+
```
|
| 137 |
+
bash run_api_ax650.sh
|
| 138 |
+
```
|
| 139 |
+
2) Start Gradio GUI
|
| 140 |
+
```
|
| 141 |
+
python scripts/gradio_demo.py
|
| 142 |
+
```
|
| 143 |
+
|
| 144 |
+
#### 3.2 Run on AXCL aarch64 Board
|
| 145 |
+
```
|
| 146 |
+
bash run_axcl_aarch64.sh
|
| 147 |
+
```
|
| 148 |
+
#### Or run on AXCL aarch64 Board with Gradio GUI
|
| 149 |
+
1) Start server
|
| 150 |
+
```
|
| 151 |
+
bash run_api_axcl_aarch64.sh
|
| 152 |
+
```
|
| 153 |
+
2) Start Gradio GUI
|
| 154 |
+
```
|
| 155 |
+
python scripts/gradio_demo.py
|
| 156 |
+
```
|
| 157 |
+
|
| 158 |
+
|
| 159 |
+
### Optional. Process Prompt Speech
|
| 160 |
If you want to replicate a specific sound, do this step.
|
| 161 |
You can use audio in asset/ .
|
| 162 |
|
| 163 |
+
#### (1). Downlaod wetext
|
| 164 |
```
|
| 165 |
pip3 install modelscope
|
| 166 |
modelscope download --model pengzhendong/wetext --local_dir pengzhendong/wetext
|
| 167 |
```
|
| 168 |
|
| 169 |
+
#### (2). Process Prompt Speech
|
| 170 |
Example:
|
| 171 |
```
|
| 172 |
python3 scripts/process_prompt.py --prompt_text asset/zh_man1.txt --prompt_speech asset/zh_man1.wav --output zh_man1
|