-
Notifications
You must be signed in to change notification settings - Fork 334
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
a0e40fa
commit 1760771
Showing
4 changed files
with
166 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Binary file modified
BIN
-10.3 KB
(85%)
doc/python/file_format_converter/file_format_converter_workflow.png
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
22 changes: 22 additions & 0 deletions
22
doc/python/file_format_converter/onnx/neural_network_console_example_coverage.rst
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,22 @@ | ||
| Neural network console examples support status. | ||
| =============================================== | ||
|
|
||
| ============================ ======== ======================================== | ||
| Network Status memo | ||
| ============================ ======== ======================================== | ||
| 01_logistic_regression OK | ||
| 02_binary_cnn OK | ||
| 06_auto_encoder NG Converter problem | ||
| 10_deep_mlp OK | ||
| 11_deconvolution NG DeConvolution does not support yet. | ||
| 12_residual_learning NG Converter problem | ||
| LSTM_auto_encoder NG Split does not support yet. | ||
| LeNet OK | ||
| bidirectional_elman_net NG Split does not support yet. | ||
| elman_net NG Split does not support yet. | ||
| elman_net_with_attention NG Split does not support yet. | ||
| gated_recurrent_unit(GRU) NG Split does not support yet. | ||
| long_short_term_memory(LSTM) NG Split does not support yet. | ||
| mnist_dcgan_with_label NG Split does not support yet. | ||
| stacked_GRU NG Split does not support yet. | ||
| ============================ ======== ======================================== |
141 changes: 141 additions & 0 deletions
141
doc/python/file_format_converter/onnx/operator_coverage.rst
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,141 @@ | ||
| Operator support status | ||
| ======================= | ||
|
|
||
| This is a status list of [ONNX operators](https://github.com/onnx/onnx/blob/master/docs/Operators.md) | ||
| that indicates if each operator can be converted to NNP. | ||
|
|
||
| - Not started The ONNX operator hasn't been checked if it can be converted to NNabla. | ||
| - Not implemented The ONNX operator has been checked if it can be converted to NNabla, but the implementation has not started. | ||
| - OK The ONNX operator can map to a NNabla operator. | ||
| - Not finished The solution is not perfect/finished, for example, the operator can map to a combination of NNabla operators. | ||
| - Not in NNabla Hard to find a solution with existing NNabla operators. | ||
|
|
||
| ======================================== =============== ================================================= | ||
| Operator Status Description | ||
| ======================================== =============== ================================================= | ||
| Abs OK | ||
| Add Not finished broadcast will be converted to a BroadcastTo | ||
| And Not finished broadcast will be converted to a BroadcastTo | ||
| ArgMax Not in NNabla Operator does not exist in NNabla | ||
| ArgMin Not in NNabla Operator does not exist in NNabla | ||
| AveragePool Not finished autopad not supported. pads must have same | ||
| value for begin and end. | ||
| BatchNormalization Not finished is_test=false not supported (only inference) | ||
| Cast Not in NNabla Operator does not exist in NNabla(No type | ||
| information is exposed in NNP) | ||
| Ceil Not implemented Should map to Ceil | ||
| Clip Not finished Converted to Identity, MaximumScalar, | ||
| MinimumScalar, or both depending on the attribute | ||
| Concat OK | ||
| Constant Not finished Converted to an input parameter | ||
| Conv Not finished auto_pad not supported. pads must have same value | ||
| for begin and end. | ||
| ConvTranspose Not implemented Should map to Deconvolution? | ||
| DepthToSpace Not in NNabla Operator does not exist in NNabla | ||
| Div Not finished broadcast will be converted to a BroadcastTo | ||
| Dropout Not finished mask output will be removed since NNabla does | ||
| not produce mask output. | ||
| Elu OK | ||
| Equal Not finished broadcast will be converted to a BroadcastTo. | ||
| Input data type will all be converted to int64 | ||
| since NNP does not have type information | ||
| Exp OK | ||
| Flatten Not in NNabla Operator does not exist in NNabla | ||
| Floor Not implemented Should map to Floor | ||
| GRU Not in NNabla Operator does not exist in NNabla | ||
| Gather Not in NNabla Operator does not exist in NNabla | ||
| Gemm Not finished alpha and beta is not supported. | ||
| Input A and B must be two dimensional, | ||
| and input C must be one dimensional. | ||
| transA, transB will be converted to | ||
| a separate transpose operator | ||
| GlobalAveragePool OK | ||
| GlobalLpPool Not in NNabla Operator does not exist in NNabla | ||
| GlobalMaxPool Not in NNabla Operator does not exist in NNabla | ||
| Greater Not finished broadcast will be converted to a BroadcastTo | ||
| HardSigmoid Not implemented Should be able to map to | ||
| MulScalar+AddScalar+MinimumScalar+ReLU | ||
| Hardmax Not in NNabla Operator does not exist in NNabla | ||
| Identity OK | ||
| InstanceNormalization Not in NNabla Operator does not exist in NNabla | ||
| LRN Not finished Converted to | ||
| PowScalar+Tranpose+SumPooling+Transpose+MulScalar+AddScalar+PowScalar. | ||
| Currently only odd size is allowed. | ||
| LSTM Not in NNabla Operator does not exist in NNabla | ||
| LeakyRelu OK | ||
| Less Not finished broadcast will be converted to a BroadcastTo | ||
| Log OK | ||
| LogSoftmax Not finished Converted to Exp+Sum+Log+Sub2. | ||
| Only works on input shape like N*C*1*1 | ||
| LpNormalization Not in NNabla Operator does not exist in NNabla | ||
| LpPool Not in NNabla Operator does not exist in NNabla | ||
| MatMul OK | ||
| Max Not finished Only input of two tensors is currently supported | ||
| MaxPool Not finished auto_pad is not supported. | ||
| pads must have same value for begin and end. | ||
| MaxRoiPool Not in NNabla Operator does not exist in NNabla | ||
| Mean Not in NNabla Operator does not exist in NNabla | ||
| Min Not finished Only input of two tensors is currently supported | ||
| Mul Not finished broadcast will be converted to a BroadcastTo | ||
| Neg Not finished Converted to MulScalar | ||
| Not OK | ||
| Or Not finished broadcast will be converted to a BroadcastTo | ||
| PRelu OK | ||
| Pad Not finished For NNP to ONNX conversion, input buffer's | ||
| dimension is assumed to be 4D if the shape cannot be determined. | ||
| Pow Not finished broadcast will be converted to a BroadcastTo | ||
| RNN Not in NNabla Operator does not exist in NNabla | ||
| RandomNormal Not implemented Should be able to map to Randn | ||
| RandomNormalLike Not in NNabla Operator does not exist in NNabla | ||
| RandomUniform Not implemented Should be able to map to Rand | ||
| RandomUniformLike Not in NNabla Operator does not exist in NNabla | ||
| Reciprocal Not finished Converted to RDivScalar | ||
| ReduceL1 Not in NNabla Operator does not exist in NNabla | ||
| ReduceL2 Not in NNabla Operator does not exist in NNabla | ||
| ReduceLogSum Not in NNabla Operator does not exist in NNabla | ||
| ReduceLogSumExp Not in NNabla Operator does not exist in NNabla | ||
| ReduceMax OK | ||
| ReduceMean OK | ||
| ReduceMin OK | ||
| ReduceProd OK | ||
| ReduceSum OK | ||
| ReduceSumSquare Not in NNabla Operator does not exit in NNabla | ||
| Relu OK | ||
| Reshape Not finished implementing | ||
| Selu OK | ||
| Sigmoid OK | ||
| Size Not in NNabla Operator does not exist in NNabla | ||
| Slice Not in NNabla Operator does not exist in NNabla | ||
| Softmax Not finished Only works on input shape like N*C*1*1 | ||
| Softplus Not finished Converted to Exp + AddScalar + Log | ||
| Softsign Not finished Converted to Abs + AddScalar + Div2 | ||
| SpaceToDepth Not in NNabla Operator does not exist in NNabla | ||
| Split Not in NNabla Operator does not exist in NNabla | ||
| Sqrt Not in NNabla Operator does not exist in NNabla | ||
| Squeeze Not in NNabla Operator does not exist in NNabla | ||
| Sub Not finished broadcast will be converted to a BroadcastTo | ||
| Sum Not finished Supporting two inputs only | ||
| Tanh OK | ||
| Tile Not in NNabla Operator does not exist in NNabla | ||
| TopK Not in NNabla Operator does not exist in NNabla | ||
| Transpose OK | ||
| Unsqueeze Not in NNabla Operator does not exist in NNabla | ||
| Xor Not finished broadcast will be converted to a BroadcastTo | ||
| experimental ATen Not started | ||
| experimental Affine Not started | ||
| experimental ConstantFill Not started | ||
| experimental Crop Not started | ||
| experimental FC Not started | ||
| experimental GRUUnit Not started | ||
| experimental GivenTensorFill Not started | ||
| experimental If Not started | ||
| experimental ImageScaler Not started | ||
| experimental Loop Not started | ||
| experimental LoopIndexTensor Not started | ||
| experimental MeanVarianceNormalization Not started | ||
| experimental ParametricSoftplus Not started | ||
| experimental Scale Not started | ||
| experimental ScaledTanh Not started | ||
| experimental ThresholdedRelu Not started | ||
| experimental Upsample Not started | ||
| ======================================== =============== ================================================= |