Skip to content

docs: correct contributor prerequisites and checkpoint scope #356

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
LauraGPT wants to merge 2 commits into
base: main
Choose a base branch
from
Open

docs: correct contributor prerequisites and checkpoint scope #356

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a73e0f6
docs: correct contributor prerequisites and checkpoint scope
LauraGPT Sep 22, 2026
db9d5ec
docs: include FastAPI CLI prerequisite in contributor setup
LauraGPT Sep 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 17 additions & 5 deletions CONTRIBUTING.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,9 @@ Thank you for your interest in contributing to SenseVoice! This guide will help

### Prerequisites

- Python 3.8 or higher
- Python 3.12 is recommended for this guide. The current `torch>=2.12.1` requirement
needs Python 3.10 or newer; Python 3.8 and 3.9 cannot satisfy the current
dependency set.
- Git
- (Optional) CUDA-compatible GPU for faster inference

Expand All @@ -29,7 +31,7 @@ source venv/bin/activate # On Windows: venv\Scripts\activate
3. **Install dependencies:**

```bash
pip install -r requirements.txt
python -m pip install -r requirements.txt
```

4. **Verify the installation:**
Expand All @@ -43,16 +45,21 @@ python -c "from funasr import AutoModel; print('Installation successful')"
If you don't have a GPU, you can run SenseVoice on CPU by setting the device to `"cpu"`:

```python
from funasr import AutoModel

model = AutoModel(
model="iic/SenseVoiceSmall",
trust_remote_code=True,
device="cpu",
)
```

For the FastAPI server:
For the FastAPI server, also install the CLI dependencies. The base `fastapi`
package in `requirements.txt` does not include the `fastapi` command's optional
dependencies:

```bash
python -m pip install "fastapi[standard]"
export SENSEVOICE_DEVICE=cpu
fastapi run --port 50000
```
Expand Down Expand Up @@ -127,13 +134,18 @@ SenseVoice/

## Understanding the Model

SenseVoice is a non-autoregressive encoder-only model that outputs:
The released SenseVoiceSmall checkpoint is a non-autoregressive encoder-only model that outputs:

- **Speech transcription** (ASR) across 50+ languages
- **Speech transcription** (ASR) for Mandarin, Cantonese, English, Japanese, and Korean
- **Emotion labels**: `HAPPY`, `SAD`, `ANGRY`, `NEUTRAL`, `FEARFUL`, `DISGUSTED`, `SURPRISED`
- **Audio event labels**: `BGM`, `Speech`, `Applause`, `Laughter`, `Cry`, `Sneeze`, `Breath`, `Cough`
- **Language identification** for Mandarin, English, Cantonese, Japanese, and Korean

The broader SenseVoice research reports support for more than 50 languages; that is
not the language support of the released SenseVoiceSmall checkpoint. See the
[README](./README.md) for the distinction between research scope and released
capabilities.

The model uses a SANM (Self-Attention with Normalized Memory) encoder architecture with CTC decoding. Emotion and event labels are predicted from the first 4 encoder output tokens, while the remaining tokens produce the transcription.

## Docker
Expand Down
Loading