Spaces:
Running
on
T4
Running
on
T4
File size: 2,048 Bytes
2485dd8 b30004d 2485dd8 070b677 bb98523 2485dd8 8373c4c 2485dd8 c397098 2485dd8 c397098 5067fec c397098 2485dd8 c397098 2485dd8 c397098 8672687 2485dd8 4c67f94 2485dd8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 |
---
title: Seamless Streaming
emoji: 📞
colorFrom: blue
colorTo: yellow
sdk: docker
pinned: false
suggested_hardware: t4-small
models:
- facebook/seamless-streaming
---
# Seamless Streaming demo
## Running on HF spaces
You can simply duplicate the space to run it.
## Running locally
### Install backend seamless_server dependencies
> [!NOTE]
> Please note: we *do not* recommend running the model on CPU. CPU inference will be slow and introduce noticable delays in the simultaneous translation.
> [!NOTE]
> The example below is for PyTorch stable (2.1.1) and variant cu118.
> Check [here](https://pytorch.org/get-started/locally/) to find the torch/torchaudio command for your variant.
> Check [here](https://github.com/facebookresearch/fairseq2#variants) to find the fairseq2 command for your variant.
If running for the first time, create conda environment and install the desired torch version. Then install the rest of the requirements:
```
cd seamless_server
conda create --yes --name smlss_server python=3.8 libsndfile==1.0.31
conda activate smlss_server
conda install --yes pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
pip install fairseq2 --pre --extra-index-url https://fair.pkg.atmeta.com/fairseq2/whl/nightly/pt2.1.1/cu118
pip install -r requirements.txt
```
### Install frontend streaming-react-app dependencies
```
conda install -c conda-forge nodejs
cd streaming-react-app
npm install --global yarn
yarn
yarn build # this will create the dist/ folder
```
### Running the server
The server can be run locally with uvicorn below.
Run the server in dev mode:
```
cd seamless_server
uvicorn app_pubsub:app --reload --host localhost
```
Run the server in prod mode:
```
cd seamless_server
uvicorn app_pubsub:app --host 0.0.0.0
```
To enable additional logging from uvicorn pass `--log-level debug` or `--log-level trace`.
### Debuging
If you enable "Server Debug Flag" when starting streaming from the client, this enables extensive debug logging and it saves audio files in /debug folder. |