TMVA_SOFIE_PyTorch_HiggsModel.py File Reference

Functions

 TMVA_SOFIE_PyTorch_HiggsModel.GenerateCode (modelFile="model.onnx")    TMVA_SOFIE_PyTorch_HiggsModel.PrepareData ()    TMVA_SOFIE_PyTorch_HiggsModel.TrainModel (x_train, y_train, x_check, name)      TMVA_SOFIE_PyTorch_HiggsModel.modelName = GenerateCode(modelFile)  Step 2 : Parse model and generate inference code with SOFIE.
   TMVA_SOFIE_PyTorch_HiggsModel.session = sofie.Session()    TMVA_SOFIE_PyTorch_HiggsModel.sofie = getattr(ROOT, "TMVA_SOFIE_" + modelName)  Step 3 : Compile the generated C++ model code.
  str TMVA_SOFIE_PyTorch_HiggsModel.TRAIN_SCRIPT    TMVA_SOFIE_PyTorch_HiggsModel.x_check = x_test[:10]    TMVA_SOFIE_PyTorch_HiggsModel.x_test  Step 1 : Create and train the model, export it to ONNX.
   TMVA_SOFIE_PyTorch_HiggsModel.x_train  Step 1 : Create and train the model, export it to ONNX.
   TMVA_SOFIE_PyTorch_HiggsModel.y = session.infer(x_check[i])    TMVA_SOFIE_PyTorch_HiggsModel.y_test  Step 1 : Create and train the model, export it to ONNX.
   TMVA_SOFIE_PyTorch_HiggsModel.y_train  Step 1 : Create and train the model, export it to ONNX.
   TMVA_SOFIE_PyTorch_HiggsModel.ytorch  

Detailed Description

This macro trains a simple deep neural network on the Higgs dataset with PyTorch, exports the model to ONNX and runs the SOFIE parser on it to generate and compile C++ inference code.

The trained model is saved as HiggsModel.onnx and is used as input by other SOFIE tutorials (e.g. TMVA_SOFIE_RDataFrame.C), so this macro needs to be run before them.

The PyTorch export and ROOT's SOFIE parser are both linked against protobuf, but usually against different versions, so loading them in the same process leads to a symbol clash. We therefore run the PyTorch training and ONNX export in a separate Python process and only use ROOT before and afterwards.

size of data 10000
Sequential(
  (0): Linear(in_features=7, out_features=64, bias=True)
  (1): ReLU()
  (2): Linear(in_features=64, out_features=64, bias=True)
  (3): ReLU()
  (4): Linear(in_features=64, out_features=1, bias=True)
  (5): Sigmoid()
)
Epoch 1/5 - average loss: 0.6746
Epoch 2/5 - average loss: 0.6556
Epoch 3/5 - average loss: 0.6445
Epoch 4/5 - average loss: 0.6362
Epoch 5/5 - average loss: 0.6314
calling torch.onnx.export with parameters {'input_names': ['input'], 'output_names': ['output'], 'external_data': False, 'dynamo': True}
[torch.onnx] Obtain model graph for `Sequential([...]` with `torch.export.export(..., strict=False)`...
[torch.onnx] Obtain model graph for `Sequential([...]` with `torch.export.export(..., strict=False)`... ✅
[torch.onnx] Run decompositions...
[torch.onnx] Run decompositions... ✅
[torch.onnx] Translate the graph into ONNX...
[torch.onnx] Translate the graph into ONNX... ✅
[torch.onnx] Optimize the ONNX graph...
[torch.onnx] Optimize the ONNX graph... ✅
model exported to ONNX as HiggsModel.onnx
input to model is  [1.3551283  1.0198661  0.98278755 0.5504138  1.2055093  0.91609305
 0.93084896] 
    -> output using SOFIE =  0.3522462546825409  using PyTorch =  0.35224625
input to model is  [1.0965776 0.9103265 1.9756684 1.3508093 1.3468878 1.4005579 1.1609015] 
    -> output using SOFIE =  0.3683587312698364  using PyTorch =  0.36835867
input to model is  [0.846992  0.9408182 0.98906   1.6148995 1.038698  1.2381754 1.0323234] 
    -> output using SOFIE =  0.602892279624939  using PyTorch =  0.6028923
input to model is  [1.897264   1.234499   0.98704207 0.708829   0.7279103  0.9053675
 0.76521254] 
    -> output using SOFIE =  0.5798262357711792  using PyTorch =  0.57982624
input to model is  [0.791873   0.9792347  0.9924756  0.9159218  1.1000326  0.94064647
 0.79085195] 
    -> output using SOFIE =  0.5650977492332458  using PyTorch =  0.56509775
input to model is  [0.9692043  0.6372814  0.9850732  0.9201175  0.72131383 0.8001433
 0.7160924 ] 
    -> output using SOFIE =  0.5997770428657532  using PyTorch =  0.5997771
input to model is  [1.5037444 1.1279533 0.9814414 1.5327642 0.7886151 1.1838427 1.0383142] 
    -> output using SOFIE =  0.5189254283905029  using PyTorch =  0.5189254
input to model is  [1.2042431  1.0750061  1.5724212  1.1590953  1.367509   1.1043229
 0.99204683] 
    -> output using SOFIE =  0.4127229154109955  using PyTorch =  0.41272292
input to model is  [1.012179   0.76250947 0.9957243  0.48331824 0.4295301  0.55478483
 0.7100585 ] 
    -> output using SOFIE =  0.3701574504375458  using PyTorch =  0.37015742
input to model is  [0.8616951  1.1908345  0.99018383 1.2523934  1.1448306  1.0246366
 0.9923519 ] 
    -> output using SOFIE =  0.5680633187294006  using PyTorch =  0.5680633
OK

Definition in file TMVA_SOFIE_PyTorch_HiggsModel.py.