From eed30eacb543cc1da8adc4bddce555787bd96a73 Mon Sep 17 00:00:00 2001 From: Dwight Lidman Date: Thu, 25 Nov 2021 17:34:42 +0100 Subject: MLBEDSW-5417: Update release notes & supported ops This commit updates the release notes for Vela version 3.2.0. It also updates the SUPPORTED_OPS.md file with new constraints. Updated the API version as a result of the bug fix commit 399c4a2d77df791e5d988c51d7fb1824ac4f266f. Updated Vela version in setup.py. Signed-off-by: Dwight Lidman Change-Id: I181e89f639a1da6013e8511ebe2d7e4f81242916 --- RELEASES.md | 38 ++++++++++++++++++++++++++++++++++++-- 1 file changed, 36 insertions(+), 2 deletions(-) (limited to 'RELEASES.md') diff --git a/RELEASES.md b/RELEASES.md index 76743221..6011d046 100644 --- a/RELEASES.md +++ b/RELEASES.md @@ -5,6 +5,39 @@ 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 3.2.0 - 26/11/2021 + +**Main feature changes:** + +* Bug fixes +* Documentation of community bug reporting via Phabricator on ML Platform +* New operator support: EXPAND_DIMS +* Extended experimental TOSA support + +**Interface changes:** + +* External API v1.2 + * Fixed bug in `npu_find_block_configs()` with Conv1D optimisation + +* Verbose output + * Clarified Subgraph IO Summary + +* Summary output + * Removed incorrect passes information + * Fixed and reformatted the reporting of CPU operators + +**Reported defect fixes:** + +* Clarified section on access cycles in PERFORMANCE.md (MLCE-654) +* Crash due to mismatch in padding shape in OFM (MLCE-652) +* Bug where back to back activations were ignored (MLCE-650) +* Output mismatch for unsupported RESHAPE with 16-bit 5D tensors (MLCE-630) +* Bug where tensors with "skip connections" were incorrectly handled (MLCE-621) +* Crash due to the same tensor being used for both inputs to operator (MLCE-620) +* Missing constraint on batch size for MEAN operator (MLCE-619) +* Bug where third-party custom ops were not passed through correctly (MLCE-602) +* Crash due to mismatch in tensor indices for EXP operator (MLCE-599) + ## Release 3.1.0 - 30/08/2021 **Main feature changes:** @@ -24,7 +57,7 @@ fixed. The version numbering adheres to the * Bug with IFM box depth for convolutions fused with SPLIT (MLCE-490) * Bug with missing attribute from integer type (MLCE-534) -* Bug with incorrect options in TensorFlow Lite mapping (MLCE-427) +* Bug with incorrect options in TensorFlow Lite mapping (MLCE-427) ## Release 3.0.0 - 28/05/2021 @@ -120,7 +153,8 @@ configurations * `--verbose-config`, `--cache-bias-scale-tensor`, `--memory-mode` * `--cpu-tensor-alignment` * Addition of External APIs: - * Command Stream generation, Driver Payload creation, and Finding a Block Config + * Command Stream generation, Driver Payload creation, and Finding a Block +Config **Reported defect fixes:** -- cgit v1.2.1