Deploying a llama.cpp Container

When creating an endpoint with a GGUF model, a llama.cpp container is automatically selected, utilizing the latest image built from the master branch of the llama.cpp repository. Upon deployment, a server with an OpenAI-compatible endpoint becomes available.

For a comprehensive list of available endpoints, please refer to the API documentation.

Deployment Steps

To deploy an endpoint with a llama.cpp container, follow these steps:

  1. Create a new endpoint and select a repository containing a GGUF model. The llama.cpp container will be automatically selected.
Select model
  1. Choose the desired GGUF file, noting that memory requirements will vary depending on the selected file. For example, an F16 model requires more memory than a Q4_K_M model.
Select GGUF file
  1. Select your desired hardware configuration.
Select hardware
  1. Optionally, you can customize the container’s configuration settings. For more information, please refer to the Configurations section below.

  2. Click the Create Endpoint button to complete the deployment.

Alternatively, you can follow the video tutorial below for a step-by-step guide on deploying an endpoint with a llama.cpp container:

Configurations

The llama.cpp container offers several configuration options that can be adjusted. After deployed, you can modify these settings by accessing the Settings tab in the endpoint details page.

Basic Configurations

Advanced Configurations

In addition to the basic configurations, you can also modify specific settings by setting environment variables. A list of available environment variables can be found in the API documentation.

Please note that the following environment variables are reserved by the system and cannot be modified:

Troubleshooting

In case the deployment fails, please watch the log output for any error messages.

You can access the logs by clicking on the Logs tab in the endpoint details page. To learn more, refer to the Logs documentation.

Malloc failed: out of memory

If you see this error message in the log:

ggml_backend_cuda_buffer_type_alloc_buffer: allocating 67200.00 MiB on device 0: cuda
Malloc failed: out of memory
llama_kv_cache_init: failed to allocate buffer for kv cache
llama_new_context_with_model: llama_kv_cache_init() failed for self-attention cache
...

That means the selected hardware configuration does not have enough memory to accommodate the selected GGUF model. You can try:

Workload evicted, storage limit exceeded

This error message indicates that the hardware has too little memory to accommodate the selected GGUF model. Try selecting a smaller model or select a larger hardware configuration.

For other issues, please refer to the llama.cpp issues page.

< > Update on GitHub