Enforcing access restrictions for fine-tuning machine learning models

A method, comprising:

• receiving a request to perform fine-tuning on a trained machine learning model using a specified tuning data set via an interface of a machine learning service;
• obtaining, by the machine learning service, a set of weights of the trained machine learning model, the tuning data set, and tuning instructions for performing the fine-tuning on the trained machine learning model, wherein the trained machine learning model is obtained from a model provider with provider access restrictions on the set of weights, and wherein the tuning data set is obtained from a model consumer with consumer access restrictions on the tuning data set;
• executing, by the machine learning service, the tuning instructions to generate a new set of fine-tuned weights at one or more computing resources of the machine learning service provisioned to perform the fine-tuning such that the set of weights of the trained machine learning are not modified, wherein the fine-tuned weights are used in combination with the set of weights of the trained machine learning model to perform inferences, wherein the executing uses the tuning data set and the set of weights of the trained machine learning model, and wherein the executing comprises:
• enforcing, at the one or more computing resources provisioned to perform the fine-tuning, the provider access restrictions and the consumer access restrictions;
• and
• storing, by the machine learning service, the new set of fine-tuned weights in a storage location that satisfies the consumer access restrictions.

One or more non-transitory, computer-readable storage media, storing program instructions that when executed on or across one or more computing devices cause the one or more computing devices to implement:

• receiving a request to perform fine-tuning on a trained machine learning model using a specified tuning data set via an interface of a machine learning service implemented as part of a provider network;
• obtaining, by the machine learning service, a set of weights of the trained machine learning model, the tuning data set, and tuning instructions for performing the fine-tuning on the trained machine learning model, wherein the trained machine learning model is obtained from a model provider with provider access restrictions on the set of weights, and wherein the tuning data set is obtained from a model consumer with consumer access restrictions on the tuning data set;
• causing, by the machine learning service, the tuning instructions to generate a new set of fine-tuned weights at one or more computing resources of the machine learning service provisioned to perform the fine-tuning such that the set of weights of the trained machine learning are not modified, wherein the fine-tuned weights are used in combination with the set of weights of the trained machine learning model to perform inferences, wherein the generation uses the tuning data set and the set of weights of the trained machine learning model, and wherein in causing the tuning instructions to generate the new set of fine-tuned weights, the program instructions cause the one or more computing devices to implement:
• enforcing, at the one or more computing resources provisioned to perform the fine-tuning, the provider access restrictions and the consumer access restrictions;
• and
• causing, by the machine learning service, the new set of fine-tuned weights to be stored in a storage location that satisfies the consumer access restrictions.