I’m working with an LLM model and would like to export it using torch.export.export to obtain the PyTorch IR. As I understand, exporting an LLM typically involves two separate graphs: prefill and decode. I’ve already successfully exported the decode graph. import torch, transformers from torch.exp…

How to torch.export tinyllama's prefill model

shuiqiang July 18, 2025, 5:59am 4

Thanks, strict = False ,It not work.

Topic		Replies	Views
LLama 3.1 torch.compile & static cache 🤗Transformers	2	338	December 9, 2024
How does the ONNX exporter work for GenerationModel with `past_key_value`? 🤗Optimum	9	2439	February 17, 2023
AOTInductor with Llama-3.2-3B-Instruct Intermediate	0	100	November 14, 2024
Inference workflow in compile mode using transformers.pipeline() 🤗Transformers	0	33	August 26, 2024
How to export mistralai/Mistral-7B-v0.1 to Tflite to use in TensorFlow Autocomplete? 🤗Optimum	1	650	June 28, 2024