From fd27111b17fe57163cd82bfe52fac2ba1aec1013 Mon Sep 17 00:00:00 2001 From: Tim Hall Date: Wed, 17 May 2023 13:19:12 +0100 Subject: MLBEDSW-7494: Update release notes - Added release information - Minor changes to SUPPORTED_OPS.md including version info Change-Id: I91fae4c40c6c1f25b874268b18d077a9babd4875 Signed-off-by: Tim Hall --- RELEASES.md | 27 +++++++++++++++++++++++++++ SUPPORTED_OPS.md | 7 +++++-- ethosu/vela/tflite_supported_operators.py | 3 +-- ethosu/vela/vela.py | 5 ++--- 4 files changed, 35 insertions(+), 7 deletions(-) diff --git a/RELEASES.md b/RELEASES.md index b73374d2..8c6dd467 100644 --- a/RELEASES.md +++ b/RELEASES.md @@ -22,6 +22,33 @@ main feature changes, interface changes and reported defects that have been fixed. The version numbering adheres to the [semantic versioning](https://semver.org/) scheme. +## Release 3.8.0 - 17/05/2023 + +**Main feature changes:** + +* New operator support: ARG_MAX, UNIDIRECTIONAL_SEQUENCE_LSTM, EXP +* Improved CONV_2D striding support +* Upgrade TensorFlow Lite support to version 2.11 +* Changed MEAN operator implementation to match changes in TensorFlow Lite reference kernels +* Fixed int8 and int16 fusing of PAD and AVERAGE_POOL_2D +* Fixed issues with pass-through support for: + * Resource variables + * Fused activation attribute and missing attributes +* Improved documentation of verbose CLI options + +**Interface changes:** + +* Addition of CLI options: + * `--verbose-progress` + +**Reported defect fixes:** + +* Compilation progress feedback (MLCE-1009) +* Reduced compilation time for large networks (MLCE-1009) +* Fixed handling of invalid RESHAPE operators (MLCE-997) +* Fixed crash when passing-through semantically incorrect operators with missing attributes (MLCE-1027) + + ## Release 3.7.0 - 16/02/2023 **Main feature changes:** diff --git a/SUPPORTED_OPS.md b/SUPPORTED_OPS.md index 4c2a098a..0c3e1e03 100644 --- a/SUPPORTED_OPS.md +++ b/SUPPORTED_OPS.md @@ -15,10 +15,11 @@ WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. --> + # Supported Ops This file was automatically generated by Vela using the `--supported-ops-report` parameter. -Vela version: `3.7.1.dev17+geeff1bcf` +Vela version: `3.8.0` This file complies with [**Gitiles Markdown syntax**](https://github.com/google/gitiles/blob/master/Documentation/markdown.md) @@ -299,7 +300,9 @@ This is a list of constraints that the RESIZE_BILINEAR operator must satisfy in W and H scaling must be equal and OFM W and H must be 2x/4x/8x IFM W and H, if align_corners is False - The size tensor must match the output tensor shape - Both align_corners and half_pixel_centers can't be True -- Half_pixel_centers for resize bilinear requires that OFM W and H is 2x IFM W and H +- For half_pixel_centers the width and height of the IFM and OFM must match one of the following criteria: + IFM W and H are both 1 + OFM W and H is 2x IFM W and H ### TFLite RESIZE_NEAREST_NEIGHBOR Constraints diff --git a/ethosu/vela/tflite_supported_operators.py b/ethosu/vela/tflite_supported_operators.py index 396fbc23..b47104d7 100644 --- a/ethosu/vela/tflite_supported_operators.py +++ b/ethosu/vela/tflite_supported_operators.py @@ -708,8 +708,7 @@ class TFLiteSupportedOperators: @staticmethod def constraint_resizebi_half_pixel_centers_dims(op): - """Half_pixel_centers for resize bilinear requires that the width - and height of the IFM and OFM must match one of the following criteria: + """For half_pixel_centers the width and height of the IFM and OFM must match one of the following criteria: IFM W and H are both 1 OFM W and H is 2x IFM W and H""" half_pixel_centers = op.attrs.get("half_pixel_centers", False) diff --git a/ethosu/vela/vela.py b/ethosu/vela/vela.py index 0e5e6eba..078f4a41 100644 --- a/ethosu/vela/vela.py +++ b/ethosu/vela/vela.py @@ -19,7 +19,6 @@ # # Provides command line interface, options parsing, and network loading. Before calling the compiler driver. import argparse -import datetime import glob import os import sys @@ -179,8 +178,7 @@ def print_subgraph_io_summary(nng): def generate_license(): lines = [ "", + "", ] return lines -- cgit v1.2.1