AVS Based Approach
Compute Request: The Coordinator initiates the process by sending a compute request. This request is the demand for computational resources to perform a specific task, such as running an inference model.
Retrieves fine-tuned Models: Upon receiving the compute request, each Operator (e.g., Operator [H100] and Operator [A100]) retrieves the necessary fine-tuned models from the Model Storage. These could be pre-trained machine learning models that have been further refined for specific tasks or datasets. The model names (e.g., Model A [Llama 3.1], Model B [Claude]) suggest versions or types of models available for use.
Inference Output: The Operators then perform the computation required for the request, which includes executing the selected models to generate an output. This output is the result of the inference process applied to the provided data.
Inference Output: The inference output from the Operators is collected by the Coordinator. This received data is the processed result that was requested initially.
Checks Operator Registration: Before or after the inference output is generated, the Domain coordinator checks the registration of the operators. This step implies a validation mechanism, possibly linked to security, credibility, or capability of the operators.
Last updated