In one of the recent iterations of the transformers library, they took out the max_model_length warning for larger requests. Previously, I was splitting larger requests on max_model_length (512 chars) prior to sending it to the model server.
I would love to split into more manageable sizes like groups of 20-40 sentences, but wanted to make sure I could ignore the max-model-length now (specifically for the opus-mt models)
In one of the recent iterations of the transformers library, they took out the max_model_length warning for larger requests. Previously, I was splitting larger requests on max_model_length (512 chars) prior to sending it to the model server.
I would love to split into more manageable sizes like groups of 20-40 sentences, but wanted to make sure I could ignore the max-model-length now (specifically for the opus-mt models)