Flan-t5 trained with HF+FSDP saves corrupted weights for shared embeddings, Use this function to make sure it can be correctly loaded.

Checks the token count for each message in your list This is not part of the OpenAI API spec.

Creates a completion for the chat message

Creates embeddings for the text

Disable the redundant torch default initialization to accelerate model creation.

Compute the DPO loss for a batch of policy and reference model log probabilities. Args: policy_chosen_logps: Log probabilities of

Overriding built-in evaluation loop to store metrics for each batch. Prediction/evaluation loop, shared by `Trainer.evaluate()` and `

Uses the conv template of the base model