INNER CODE UNIT · Python

model_inputs

FoundationVision/Liquid · evaluation/app.py:224

            model_inputs = vqllm.prepare_inputs_for_generation(input_ids, **model_kwargs)

            if i > 0 and guidance_scale>1:
                outputs = vqllm(
                    **model_inputs,
                    return_dict=True,
                    output_attentions=False,
                    output_hidden_states=False,
                )
            else:
                outputs = vqllm(
                    **model_inputs,
                    return_dict=True,
                    output_attentions=False,
                    output_hidden_states=False,
                )

            next_token_logits = outputs.logits[:, -1:, :]

View source record →