https://github.com/replicate/cli
CLI for Replicate
https://github.com/replicate/cli
Last synced: 3 months ago
JSON representation
CLI for Replicate
- Host: GitHub
- URL: https://github.com/replicate/cli
- Owner: replicate
- License: apache-2.0
- Created: 2023-08-11T12:07:07.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2024-09-16T18:01:48.000Z (over 1 year ago)
- Last Synced: 2025-02-28T09:18:32.576Z (11 months ago)
- Language: Go
- Homepage: https://replicate.com
- Size: 840 KB
- Stars: 75
- Watchers: 18
- Forks: 11
- Open Issues: 9
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Replicate CLI
## Install
If you're using macOS, you can install the Replicate CLI using Homebrew:
```console
brew tap replicate/tap
brew install replicate
```
Or you can build from source and install it with these commands
(requires Go 1.20 or later):
```console
make
sudo make install
```
## Upgrade
If you previously installed the CLI with Homebrew,
you can upgrade to the latest version by running the following command:
```console
brew upgrade replicate
```
## Usage
Grab your API token from [replicate.com/account](https://replicate.com/account)
and set the `REPLICATE_API_TOKEN` environment variable.
```console
$ export REPLICATE_API_TOKEN=
```
---
```console
Usage:
replicate [command]
Core commands:
hardware Interact with hardware
model Interact with models
prediction Interact with predictions
scaffold Create a new local development environment from a prediction
training Interact with trainings
Alias commands:
run Alias for "prediction create"
stream Alias for "prediction create --stream"
train Alias for "training create"
Additional Commands:
completion Generate the autocompletion script for the specified shell
help Help about any command
Flags:
-h, --help help for replicate
-v, --version version for replicate
Use "replicate [command] --help" for more information about a command.```
```
---
### Create a prediction
Generate an image with [SDXL].
```console
$ replicate run stability-ai/sdxl \
prompt="a studio photo of a rainbow colored corgi"
Prediction created: https://replicate.com/p/jpgp263bdekvxileu2ppsy46v4
```
### Stream prediction output
Run [LLaMA 2] and stream output tokens to your terminal.
```console
$ replicate run meta/llama-2-70b-chat --stream \
prompt="Tell me a joke about llamas"
Sure, here's a joke about llamas for you:
Why did the llama refuse to play poker?
Because he always got fleeced!
```
### Create a local development environment from a prediction
Create a Node.js or Python project from a prediction.
```console
$ replicate scaffold https://replicate.com/p/jpgp263bdekvxileu2ppsy46v4 --template=node
Cloning starter repo and installing dependencies...
Cloning into 'jpgp263bdekvxileu2ppsy46v4'...
Writing new index.js...
Running example prediction...
[
'https://replicate.delivery/pbxt/P79eJmjeJsql40QpRbWVDtGJSoTtLTdJ494kpQexSDhYGy0jA/out-0.png'
]
Done!
```
### Chain multiple predictions
Generate an image with [SDXL] and upscale that image with [ESRGAN].
```console
$ replicate run stability-ai/sdxl \
prompt="a studio photo of a rainbow colored corgi" | \
replicate run nightmareai/real-esrgan --web \
image={{.output[0]}}
# opens prediction in browser (https://replicate.com/p/jpgp263bdekvxileu2ppsy46v4)
```
### Create a model
Create a new model on Replicate.
```console
$ replicate model create yourname/model --private --hardware gpu-a40-small
```
To list available hardware types:
```console
$ replicate hardware list
```
After creating your model, you can [fine-tune an existing model](https://replicate.com/docs/fine-tuning) or [build and push a custom model using Cog](https://replicate.com/docs/guides/push-a-model).
### Fine-tune a model
Fine-tune [SDXL] with your own images:
```console
$ replicate train --destination mattt/sdxl-dreambooth --web \
stability-ai/sdxl \
input_images=@path/to/pictures.zip \
use_face_detection_instead=true
# opens the training in browser
```
> [!NOTE]
> Use the `@` prefix to upload a file from your local filesystem.
> It works like curl's `--data-binary` option.
For more information,
see [our blog post about fine-tuning with SDXL](https://replicate.com/blog/fine-tune-sdxl).
### View a model's inputs and outputs
Get the schema for [SunoAI Bark]
```console
$ replicate model schema suno-ai/bark
Inputs:
- prompt: Input prompt (type: string)
- history_prompt: history choice for audio cloning, choose from the list (type: )
- custom_history_prompt: Provide your own .npz file with history choice for audio cloning, this will override the previous history_prompt setting (type: string)
- text_temp: generation temperature (1.0 more diverse, 0.0 more conservative) (type: number)
- waveform_temp: generation temperature (1.0 more diverse, 0.0 more conservative) (type: number)
- output_full: return full generation as a .npz file to be used as a history prompt (type: boolean)
Output:
- type: object
```
[api]: https://replicate.com/docs/reference/http
[LLaMA 2]: https://replicate.com/replicate/llama-2-70b-chat
[SDXL]: https://replicate.com/stability-ai/sdxl
[ESRGAN]: https://replicate.com/nightmareai/real-esrgan
[SunoAI Bark]: https://replicate.com/suno-ai/bark