Prepare v1.0.1 (IBM#14)

### Description

- Prepare version 1.0.1.

#### Features

_N/A_

#### Fixes

- Fix incorrect description for biases and hidden_weights in doc.
- Fix incorrect description for convolution 2D calculations.
- Fix memory leak from sample by add missing ztensor cleanup.
- Fix divide by zero scenario during RNN operations when pre-check is enabled.
- Fix invalid pointer inside testDriver_zdnn_gru_dual_layers.c
- Fix description and returned status for ZDNN_INVALID_CONCAT_INFO

Signed-off-by: Nicholas Marion <[email protected]>

Prepare v1.0.0 (IBM#10)

### Description

- Bump version to v1.0.0
- Update `common_rnn.c` to sync with intended v0.4.0 version.
- Minor Samples fix

#### Features

No new features were added.

#### Fixes

Resolved issue with building tests IBM#9. Bug was caused because `common_rnn.c` was not updated along with other files for v0.4.0.

Sync to 0.4.0 (IBM#5)

Sync latest changes.

- RNN Layer 2 Support
- RNN Bi-directional Output Transformation Support
- Removal of non-inclusive IT Terms
- RNN Pre-check Support
- Optimized zTensor reserved space
- Transform optimizations
- Minor bug fixes

Signed-off-by: Nicholas Marion [email protected]

Prepare for v0.3.1 (IBM#1)

1.  Currently a target attribute is used to build only the functions required for initialization without -march=z14. This however leads to problems when building with fortify source. The function wrappers for the fortify source feature are marked always_inline but GCC would have to reject inlining into functions with a target attribute.

2. With that commit the init routines are moved to a separate source file: zdnn_init.c which is built with a different set of options.

3. Add CONTRIBUTION.md file

4. Minor typo in API doc.

Signed-off-by: Andreas Krebbel [email protected]
Signed-off-by: Nicholas Marion [email protected]