Skip to content

Commit

Permalink
fix: clarify filepaths in instructions
Browse files Browse the repository at this point in the history
  • Loading branch information
ishaansehgal99 committed Nov 6, 2023
1 parent 4b017ce commit 609c4a1
Showing 1 changed file with 10 additions and 4 deletions.
14 changes: 10 additions & 4 deletions presets/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,25 +19,31 @@ Each model has its own infrastructure requirements. Kaito controller performs a
### Building the Llama image
1. Select Model Version: Identify the Llama model version to build, such as llama-2-7b or llama-2-7b-chat. Available models include `llama-2-7b, llama-2-13b, llama-2-70b, llama-2-7b-chat, llama-2-13b-chat and llama-2-70b-chat`.

2. Local Preset Path: Point to the local path of the model presets, which are found at the [kaito/presets/llama-2](
2. Clone Kaito Repository: Clone the kaito repository using the git command to your local machine
```
git clone https://github.com/Azure/kaito.git
```
Once cloned, the path to model presets will be a fixed one from the root of the cloned repository - [kaito/presets/llama-2](
https://github.com/Azure/kaito/tree/main/presets/llama-2) or [kaito/presets/llama-2-chat](https://github.com/Azure/kaito/tree/main/presets/llama-2-chat) directories for text and chat models, respectively.

3. Model Weights: Ensure your model weights are organized as /llama/<MODEL-VERSION> for the build process to include them in the Docker image.

4. Build Command:
Execute the Docker build command, replacing placeholders with actual values:
Navigate to the correct folder within the cloned repository to run the Docker build command. The path should be `kaito/docker/presets/llama-2`. Then execute the Docker build command, replacing placeholders with actual values:

```
cd kaito/docker/presets/llama-2
docker build \
--build-arg LLAMA_VERSION=<MODEL-VERSION> \
--build-arg SRC_DIR=<PATH-TO-LLAMA-PRESET> \
-t <YOUR-IMAGE-NAME>:<YOUR-TAG> .
```
For example, to build the llama-2-7b model, the command would look like this:
```
cd kaito/docker/presets/llama-2
docker build \
--build-arg LLAMA_VERSION=llama-2-7b \
--build-arg SRC_DIR=/home/kaito/presets/llama-2 \
--build-arg SRC_DIR=../../../presets/llama-2 \
-t llama-2-7b:latest .
```

Expand All @@ -52,7 +58,7 @@ inference:
presetOptions:
image: <YOUR IMAGE URL>
imagePullSecrets: # Optional
- <IMAGE SECRETS>
- <IMAGE PULL SECRETS>
Replace `<MODEL-VERSION>`, `<YOUR-IMAGE-URL>`, and `<IMAGE-PULL-SECRET>` with your specific details. For a reference implementation, see the example at [kaito_workspace_llama2_7b-chat.yaml](https://github.com/Azure/kaito/blob/main/examples/kaito_workspace_llama2_7b-chat.yaml)


Expand Down

0 comments on commit 609c4a1

Please sign in to comment.