RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.

[kunxiongzhu]

System Info

GPU: NVIDIA A6000
python 3.10

Environment/Platform

• Website/web-app
• Browser extension
• Server-side (e.g., Node.js, Deno, Bun)
• Desktop app (e.g., Electron)
• Other (e.g., VSCode extension)

Description

I’m not sure why I can successfully export the Llama2 model from Hugging Face to ONNX, but the DeepSeek model fails, even though both are 7B models with similar sizes. I’m using the same command for both exports, yet Llama2 works without any issues while DeepSeek doesn’t. Could you help me understand why this might be happening?

Here is my command:
llama2: (worked well)
python -m scripts.convert --model_id /mnt/data/ehdd1/home/models/hf/Llama-2-7b-chat-hf/Llama-2-7b-chat-hf/ --output_parent_dir /mnt/data/ehdd1/home/models/onnx/Llama-2-7b-chat-hf/fp16/ --skip_onnxslim --quantize False --task text-generation

deepseek: (failed)
python -m scripts.convert --model_id /mnt/data/ehdd1/home/models/hf/deepseek-llm-7b-chat/deepseek-llm-7b-chat/ --output_parent_dir /mnt/data/ehdd1/home/models/onnx/deepseek-llm-7b-chat/fp16/ --skip_onnxslim --quantize False --task text-generation

bug:
Traceback (most recent call last):
File "", line 198, in _run_module_as_main
File "", line 88, in _run_code
File "/home/myid/kz96891/Code/non-determinism/framework/transformers.js/scripts/convert.py", line 455, in
main()
File "/home/myid/kz96891/Code/non-determinism/framework/transformers.js/scripts/convert.py", line 342, in main
main_export(**export_kwargs)
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/optimum/exporters/onnx/main.py", line 373, in main_export
onnx_export_from_model(
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/optimum/exporters/onnx/convert.py", line 1199, in onnx_export_from_model
_, onnx_outputs = export_models(
^^^^^^^^^^^^^^
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/optimum/exporters/onnx/convert.py", line 786, in export_models
export(
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/optimum/exporters/onnx/convert.py", line 892, in export
export_output = export_pytorch(
^^^^^^^^^^^^^^^
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/optimum/exporters/onnx/convert.py", line 585, in export_pytorch
onnx_export(
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/torch/onnx/init.py", line 375, in export
export(
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/torch/onnx/utils.py", line 503, in export
_export(
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/torch/onnx/utils.py", line 1565, in _export
graph, params_dict, torch_out = _model_to_graph(
^^^^^^^^^^^^^^^^
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/torch/onnx/utils.py", line 1118, in _model_to_graph
graph = _optimize_graph(
^^^^^^^^^^^^^^^^
File "/home/myid/kz96891/anaconda3/envs/python3.10/lib/python3.12/site-packages/torch/onnx/utils.py", line 664, in _optimize_graph
_C._jit_pass_onnx_graph_shape_type_inference(
RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library. Therefore the output file must be a file path, so that the ONNX external data can be written to the same directory. Please specify the output file name.

Plz, is there any way to solve this? I don't know what's the factor fr.

Reproduction

python -m scripts.convert --model_id /mnt/data/ehdd1/home/models/hf/deepseek-llm-7b-chat/deepseek-llm-7b-chat/ --output_parent_dir /mnt/data/ehdd1/home/models/onnx/deepseek-llm-7b-chat/fp16/ --skip_onnxslim --quantize False --task text-generation

[NSTiwari]

I'm getting the same error while converting PaliGemma 2. Any updates?