All release notes are available at this page .
| Release Notes V1.2.0 |
Date of release 22 january 2022 |
Features
Layers
| Native LabVIEW | GPU (CUDA) | |
| Add2D | done | done |
| Add3D | done | done |
| Add4D | done | done |
| Additive Attention | done | Next release |
| AlphaDropout2D | done | done |
| AlphaDropout3D | done | done |
| AlphaDropout4D | done | done |
| AlphaDropout5D | done | done |
| Attention | done | Next release |
| Average2D | done | done |
| Average3D | done | done |
| Average4D | done | done |
| AvgPool1D | done | done |
| AvgPool2D | done | done |
| AvgPool3D | done | done |
| BatchNormalization2D | done | done |
| BatchNormalization3D | done | done |
| BatchNormalization4D | done | done |
| BatchNormalization5D | done | done |
| BatchNormalization6D | done | done |
| Bidirectional | done | done |
| Concatenate | done | done |
| Conv1D | done | done |
| Conv1DTranspose | done | done |
| Conv2D | done | done |
| Conv2DTranspose | done | done |
| Conv3D | done | done |
| Conv3DTranspose | done | done |
| ConvLSTM1D | done | Next release |
| ConvLSTM2D | done | Next release |
| ConvLSTM3D | done | Next release |
| Cropping1D | done | done |
| Cropping2D | done | done |
| Cropping3D | done | done |
| Dense | done | done |
| DepthwiseConv2D | done | Next release |
| Dropout2D | done | done |
| Dropout3D | done | done |
| Dropout4D | done | done |
| Dropout5D | done | done |
| Embedding | done | Next release |
| Flatten | done | done |
| GaussianDropout2D | done | Next release |
| GaussianDropout3D | done | Next release |
| GaussianDropout4D | done | Next release |
| GaussianDropout5D | done | Next release |
| GaussianNoise2D | done | Next release |
| GaussianNoise3D | done | Next release |
| GaussianNoise4D | done | Next release |
| GaussianNoise5D | done | Next release |
| GlobalAvgPool1D | done | done |
| GlobalAvgPool2D | done | done |
| GlobalAvgPool3D | done | done |
| GlobalMaxPool1D | done | done |
| GlobalMaxPool2D | done | done |
| GlobalMaxPool3D | done | done |
| GRU | done | done |
| LayerNormalization2D | done | Next release |
| LayerNormalization3D | done | Next release |
| LayerNormalization4D | done | Next release |
| LayerNormalization5D | done | Next release |
| LSTM | done | done |
| MaxPool1D | done | done |
| MaxPool2D | done | done |
| MaxPool3D | done | done |
| MultiHeadAttention | done | done |
| Multiply2D | done | done |
| Multiply3D | done | done |
| Permute3D | done | Next release |
| Reshape | done | done |
| RNN | done | done |
| SeparableConv1D | done | Next release |
| SeparableConv2D | done | Next release |
| SimpleRNN | done | done |
| SpatialDropout1D | done | Next release |
| SpatialDropout2D | done | Next release |
| SpatialDropout3D | done | Next release |
| Substract2D | done | done |
| Substract3D | done | done |
| Substract4D | done | done |
| Substract5D | done | done |
| TimeDistributed | done | Next release |
| UpSampling1D | done | done |
| UpSampling2D | done | done |
| UpSampling3D | done | done |
| ZeroPadding1D | done | done |
| ZeroPadding2D | done | done |
| ZeroPadding3D | done | done |
Enhancement
| Functionality | Information |
| BUG FIXED | |
| Check loss shape | error management was not well wired |
| Add Graph to Graph | concatenation of two graph is now fixed |
| Summary | Switch between successors and predessors |
| GPU | Fixed some memory leak |
| MNIST examples | Corrected bug in dataloading |
| NEW | |
| Automatic dimension layer | Unique Typedef utilisation |
| Weights | New weights data management added (same as data management) |
| Add more detailed documentations | |
| GPU management | New GPU compatibilities detection functionalities add (get) |
| ZeroPadding/Cropping | added with cuda |
| Conv/Pooling Channel Last | Added Conv/ConvTransp/Avg-MaxPool channel_last |
| MNIST CNN example | Added CPU/GPU functionalities |
| HAIBAL 32 | HAIBAL is now compatible LabVIEW 32 |
LabVIEW compatibility versions
LabVIEW 2020 and 2021 are supported with this release.
Version of CUDA API
CUDA 11.7 & Cudnn 8.5.0
To make HAIBAL working properly with Cuda, you must use the platform installer tool available in download section.
