https://github.com/saladtechnologies/whisper-large
An inference server running the Whisper Large model
https://github.com/saladtechnologies/whisper-large
Last synced: 8 months ago
JSON representation
An inference server running the Whisper Large model
- Host: GitHub
- URL: https://github.com/saladtechnologies/whisper-large
- Owner: SaladTechnologies
- License: mit
- Created: 2023-08-22T18:27:22.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2023-08-23T15:21:46.000Z (about 2 years ago)
- Last Synced: 2025-02-16T16:59:00.480Z (8 months ago)
- Language: Python
- Size: 269 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# whisper-large
An inference server running the Whisper Large model
## Build the container
First, you will need to download the whisper large model from the [whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) model page. You can do this by running the following command:
```bash
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/openai/whisper-large-v2 ./model
# remove pointers to the weights
rm ./model/flax_model.msgpack
rm ./model/pytorch_model.bin
rm ./model/tf_model.h5
```
Now, download the actual pytorch weights with wget:
```bash
wget https://huggingface.co/openai/whisper-large-v2/resolve/main/pytorch_model.bin -P ./model
```
Now, you can build the container:
```bash
docker build -t saladtechnologies/whisper-large:latest .
```
## Run the container
To run the container, make sure you mount your GPU, and expose port 8888:
```bash
docker run \
--gpus all \
-p 1111:1111 \
-e HOST="0.0.0.0" \
saladtechnologies/whisper-large:latest
```
## Use The Container
```bash
curl -X POST \
'http://localhost:1111/generate/' \
--header 'Content-Type: application/octet-stream' \
--data-binary '@/home/shawn/code/SaladTechnologies/whisper-large/Recording.wav'
```