1.25.2

• DepthToSpace
  • Support for DepthToSpace (CRD mode) transformation of dynamic tensors with multiple undefined dimensions.
  • ONNX: fp32.onnx.zip
  • TFLite: fp32_float32_dynamic.tflite.zip
  • INPUT: float32 [batch_size,3,height,width]
  • OUTPUT: float32 [batch_size,3,height,width]

ONNX	TFLite

• Failed to convert Real-ESRGAN x4-general (NCHW) #563

What's Changed

• Support for DepthToSpace transformation of dynamic tensors with multiple undefined dimensions by @PINTO0309 in #668

Full Changelog: 1.25.1...1.25.2

1.25.1

• Add -inimc / --input_names_to_interrupt_model_conversion option.
  By specifying ONNX input or output names, only the middle part of the model can be converted. This is useful when you want to see what output is obtained in what part of the model after conversion, or when debugging the model conversion operation itself.

  For example, take a model with multiple inputs and multiple outputs as shown in the figure below to try a partial transformation.

  

  • To convert by specifying only the input name to start the conversion

    wget https://github.com/PINTO0309/onnx2tf/releases/download/1.25.0/cf_fus.onnx
    onnx2tf -i cf_fus.onnx -inimc 448 -coion
    

    

  • To convert by specifying only the output name to end the conversion

    wget https://github.com/PINTO0309/onnx2tf/releases/download/1.25.0/cf_fus.onnx
    onnx2tf -i cf_fus.onnx -onimc dep_sec -coion
    

    

  • To perform a conversion by specifying the input name to start the conversion and the output name to end the conversion

    wget https://github.com/PINTO0309/onnx2tf/releases/download/1.25.0/cf_fus.onnx
    onnx2tf -i cf_fus.onnx -inimc 448 -onimc velocity -coion
    

    

What's Changed

• Add -inimc option by @PINTO0309 in #667

Full Changelog: 1.25.0...1.25.1

1.25.0

• tensorflow==2.17.0
• onnx==1.16.1
• onnxruntime==1.18.1
  • CUDA EP 12.5
  • TensorRT EP 10.2.0

What's Changed

• Add type hints by @PINTO0309 in #665
• tf==2.17.0, onnx==1.16.1, onnxruntime==1.18.1 by @PINTO0309 in #666

Full Changelog: 1.24.1...1.25.0

1.24.1

• Optimization of four consecutive arithmetic operations using scalar values ​​following Gemm and MatMul.

  If no change occurs in the shape of the input tensor and the output tensor after the operation, skip the broadcast regardless of whether the previous operation was Gemm or not. It should be effective no matter how many arithmetic operations on scalar values ​​are performed in succession.

  • repro_new_onnx_model_v2_2.onnx repro_new_onnx_model_v2_2.onnx.zip

    • Without -b option

      ONNX	TFLite

    • With -b 1 option

      ONNX	TFLite

      

  • ViT-B-16__openai_partial.onnx ViT-B-16__openai_partial_cut.onnx.zip

    ONNX	TFLite

• pre_explicit_broadcast should not expend scalar tensor #573

What's Changed

• Optimization of four consecutive arithmetic operations using scalar values ​​following Gemm and MatMul by @PINTO0309 in #664

Full Changelog: 1.24.0...1.24.1

1.24.0

• Significantly upgraded Flatbuffer schema_v3 v2.11.0 -> v2.17.0
• diff (old vs new)
  • diff_output.txt
• StableHLO
• FullyConnected - quantized_bias_type

What's Changed

• Significantly upgraded Flatbuffer schema_v3 v2.11.0 -> v2.17.0 by @PINTO0309 in #663

Full Changelog: 1.23.3...1.24.0

1.23.3

1. Disable arm64/aarch64 containers
2. Fix GroupConvolution conversion bug on -osd and -oiqt.

• Error when without -dgc: onnx2tf.py -i shufflenet-9.onnx -oiqt #657

• [CI] GitHub Actions fails to build ARM64/aarch64 docker image #653

• https://github.com/PINTO0309/onnx2tf/actions/runs/9729858105/job/26852247647

• https://github.com/PINTO0309/onnx2tf/actions/runs/9771330921/job/26973945295

What's Changed

• Disable arm64/aarch64 containers and fix GroupConvolution conversion bug on -osd and -oiqt by @PINTO0309 in #661

Full Changelog: 1.23.2...1.23.3

1.23.2

• Temporary deletion arm64

What's Changed

• Temporary deletion arm64 by @PINTO0309 in #660

Full Changelog: 1.23.1...1.23.2

1.23.1

• Fix arm64 test

What's Changed

• Fix arm64 by @PINTO0309 in #659

Full Changelog: 1.23.0...1.23.1

1.23.0

• Generation of saved_model with GroupConvolution and Full Integer Quantization without -dgc, -osd.
• Improved stability of Full Integer Quantization.
• e.g. YOLOv9 with GroupConvolution + Dynamic Input [1, None, None, 3]
• yolov9_n_wholebody_with_wheelchair_post_0100_1x3x384x672_full_integer_quant.tflite.zip
  
  
  

What's Changed

• Generation of saved_model with GroupConvolution and Full Integer Quantization without --dgc by @PINTO0309 in #655

Full Changelog: 1.22.6...1.23.0

1.22.6

• Support for INT8 auto-calibration of models with undefined dimensions in the input OP.
  

What's Changed

• Support for INT8 auto-calibration of models with undefined dimensions in the input OP by @PINTO0309 in #652

Full Changelog: 1.22.5...1.22.6