22.08 Release candidate 4.

1. 1bbd06b MLBEDSW-6879: TFLG pass-through test crash by Tim Hall · 1 year, 9 months ago 3.5.0 3.5.0.rc4
2. dc7414a MLBEDSW-6844: Exclude resize ops from cascades by Johan Alfvén · 1 year, 9 months ago 3.5.0.rc2
3. 8912f3a MLBEDSW-6769: Fix odd stripe heights for upscaling by erik.andersson@arm.com · 1 year, 9 months ago
4. 255dad7 MLBEDSW-6645: MLCE: Optimize SRAM usage by Johan Alfvén · 1 year, 10 months ago
5. 8e1352a MLBEDSW-6830: MLCE: Fix assert on concat op by Johan Alfvén · 1 year, 9 months ago
6. 21d9031 MLBEDSW-6640: Modify elementwise block size selection by Rickard Bolin · 1 year, 9 months ago
7. dd49a72 Revert reversion of TensorFlow 2.9 update by erik.andersson@arm.com · 1 year, 9 months ago 3.5.0.rc1
8. 885033b MLBEDSW-4157: Add RESIZE_NEAREST_NEIGHBOR support by Tim Hall · 1 year, 10 months ago
9. 47c7636 MLBEDSW-6616: ResizeBilinear align corners is incorrect by Tim Hall · 1 year, 10 months ago
10. e178f38 vela: OFM_SCALE refactor by Tim Hall · 1 year, 10 months ago
11. 783d364 MLBEDSW-6700: Fix compiler assert when fusing tensors by Johan Alfvén · 1 year, 10 months ago
12. a9fba26 MLBEDSW-6710: Revert Tensorflow 2.9 by Johan Alfvén · 1 year, 10 months ago
13. 1156317 MLBEDSW-6703 Add SHAPE operator to supported operators by Fredrik Svedberg · 1 year, 10 months ago
14. 93f492b MLBEDSW-6635: Update to TensorFlow 2.9 by erik.andersson@arm.com · 1 year, 11 months ago
15. d03dc50 MLBEDSW-6496 mlperf_deeplabv3_mnv2_ade20k_int8 fails at verify_output for u65 by Fredrik Svedberg · 1 year, 11 months ago
16. a04f2f7 MLBEDSW-6687 Vela crashes in npu_serialisation.py and tflite_graph_optimiser.py by Fredrik Svedberg · 1 year, 10 months ago
17. c4d35eb MLBEDSW-4856: Removed dead code by oliper01 · 1 year, 11 months ago
18. 6b2a0b4 MLBEDSW-6261: Elementwise cascading by erik.andersson@arm.com · 2 years, 2 months ago
19. 25f48dd MLBEDSW-6314 Static optimisation for quantise OP by Ayaan Masood · 1 year, 11 months ago
20. 4965fae MLBEDSW-6313 Static optimisation for Shape OP by Ayaan Masood · 1 year, 11 months ago
21. 68b8f2f MLBEDSW-6639: Bug fix for evicted FMS in the fast storage allocator by Johan Alfvén · 1 year, 11 months ago
22. 5c30971 MLBEDSW-6347: Improved fast storage allocator by Johan Alfvén · 1 year, 11 months ago
23. 5cc4c76 MLBEDSW-6614 Improve elementwise block size selection by Fredrik Svedberg · 1 year, 11 months ago
24. 7ce6b32 MLBEDSW-4783: Make config handling more user friendly by Rickard Bolin · 2 years ago
25. 6d7a4f0 MLBEDSW-4783: Fix issue with relative paths to config files by Rickard Bolin · 2 years ago
26. 9b8b448 MLBEDSW-6593: Issue with finding some config files by Rickard Bolin · 2 years ago
27. cda4fcb MLBEDSW-6563: networks failing with memory area exceeded in vela by Tim Hall · 2 years ago 3.4.0.rc2
28. 8bc7a65 MLBEDSW-6296: improvement_dram can become NaN by Tim Hall · 2 years ago
29. eb5a4a8 MLBEDSW-6271: Key error when using --verbose-performance option by Rickard Bolin · 2 years ago
30. 0f98de6 MLBEDSW-6384: Updated weight buffering cycle calculation by Johan Alfvén · 2 years ago
31. 1e363b1 MLBEDSW-6430: MLCE: Update to graph has sequential ethos-u ops by Johan Alfvén · 2 years ago
32. 0b20781 MLBEDSW-6430: MLCE: Vela splitting network into two ethos operators by Johan Alfvén · 2 years, 1 month ago
33. 1538dce MLBEDSW-4783: Add config file directory structure by Rickard Bolin · 2 years, 1 month ago
34. c1be087 MLBEDSW-6271: MLCE: Layer wise Utilization info from Vela by Tim Hall · 2 years, 2 months ago
35. 3dae1b6 MLBEDSW-6296: Updated condition for the opt size weight buffering schedule by Johan Alfvén · 2 years ago
36. fd8b500 MLBEDSW-6263: Use separate tensors for double buffering by Rickard Bolin · 2 years ago
37. 6f4cb03 MLBEDSW-6296: Regression caused by bigger weight buffering size by Johan Alfvén · 2 years ago 3.4.0.rc1
38. e51a05c MLBEDSW-6454: Enable ReLu with negative alpha value by Johan Alfvén · 2 years ago
39. 114baba MLBEDSW-6452: Add byte offset in command stream by Tim Hall · 2 years ago
40. e80038a Revert "MLBEDSW-6312: Find block config improvement" by Tim Hall · 2 years ago
41. ab677b3 MLBEDSW-6500: Address offset out of range by Johan Alfvén · 2 years ago
42. b5df773 Revert "MLBEDSW-6263: Use separate tensors for double buffering" by Tim Hall · 2 years ago
43. 95b07c1 MLBEDSW-6425: Update to TensorFlow 2.8 (bugfix) by Rickard Bolin · 2 years, 1 month ago
44. d66f801 MLBEDSW-6425: Update to TensorFlow 2.8 by Rickard Bolin · 2 years, 1 month ago
45. a2ec5aa MLBEDSW-5384 FC layers run on NPU if underlying shape is 2D by Ayaan Masood · 2 years, 1 month ago
46. f9267da MLBEDSW-6407: Vela fails with TypeError in npu_performance by Tim Hall · 2 years, 1 month ago
47. 814d01f MLBEDSW-6371: Output diff caused by operator clone bug by Rickard Bolin · 2 years, 1 month ago
48. cce7f2d MLBEDSW-6339 Performance drop on wav2letter by Johan Alfvén · 2 years, 1 month ago
49. a3fe665 vela: Minor refactor by Tim Hall · 2 years, 2 months ago dev/mlbedsw-6271
50. 68df8a1 vela: Added debug info to external API by Tim Hall · 2 years, 2 months ago
51. d857507 Update version of Black to 22.3.0 by Jonas Ohlsson · 2 years, 2 months ago
52. cc5f4de MLBEDSW-6263: Use separate tensors for double buffering by Louis Verhaard · 2 years, 3 months ago
53. a19b467 MLBEDSW-6249: HillClimb improved stuck avoidance by Louis Verhaard · 2 years, 3 months ago
54. 37ba98c MLBEDSW-6098: Order check in cascade builder by Louis Verhaard · 2 years, 2 months ago
55. 3c5cfe9 MLBEDSW-6298: MLCE: Unable to find a valid block config by Tim Hall · 2 years, 2 months ago
56. 845e232 MLBEDSW-3367 Add mypy to pre-commit by Jonas Ohlsson · 2 years, 3 months ago
57. d2b5510 MLBEDSW-6312: Find block config improvement by Louis Verhaard · 2 years, 2 months ago
58. 43d2758 MLBEDSW-5332: Bug fix optimise_strided_conv by Louis Verhaard · 2 years, 2 months ago
59. 5c8f1e5 MLBEDSW-6245: Bug fix fast storage allocator by Louis Verhaard · 2 years, 3 months ago
60. 77b448f Fix bug storing encoded NPU weight UUIDs by Jonas Ohlsson · 2 years, 2 months ago
61. 1ddb2ed Vela: Fix diff in mean op by Diqing Zhong · 2 years, 2 months ago
62. f8e353b Updated elementwise cycle calculation by Johan Alfvén · 2 years, 3 months ago
63. 25e700c MLBEDSW-3367 Update pre-commit flake8 version by Jonas Ohlsson · 2 years, 2 months ago
64. 0ab2edc MLBEDSW-6247: MLCE: Issue when running a model with Padding by Tim Hall · 2 years, 3 months ago
65. d5cbef3 MLBEDSW-5873 Fixed divide by zero warning in memory transfer efficiency by Ayaan Masood · 2 years, 3 months ago
66. b801dda MLBEDSW-5880 Fixed Vela verbose weight flag by Ayaan Masood · 2 years, 3 months ago
67. 5e0ae55 MLBEDSW-6148: Reduce SRAM usage for elementwise op by Johan Alfvén · 2 years, 3 months ago
68. 0bb7ad1 MLBEDSW-5554: Constraints for single-axis mean operations on NPU by James Peet · 2 years, 3 months ago
69. 1b9218e MLBEDSW-6180: Protect overwrite of subgraph output by Johan Alfvén · 2 years, 3 months ago 3.3.0.rc1
70. 211165a MLBEDSW-5582: MLCE: memory corruption with zero concat by Johan Alfvén · 2 years, 3 months ago
71. de6cb64 MLBEDSW-5839: Port of improved spilling behaviour by erik.andersson@arm.com · 2 years, 3 months ago
72. 8d57aaa MLBEDSW-6148: Allow overwrite of subgraph input by Johan Alfvén · 2 years, 3 months ago
73. e546def MLBEDSW-3623: Diff on semantic_segmentation by Rickard Bolin · 2 years, 4 months ago
74. 628928d MLBEDSW-6060: Revert patch for MLBEDSW-5582 by Johan Alfvén · 2 years, 4 months ago
75. e91b531 MLBEDSW-6018: Fix double buffering on dual core by Louis Verhaard · 2 years, 4 months ago
76. 849ff81 MLBEDSW-5582: MLCE: memory corruption with zero concat by Tim Hall · 2 years, 5 months ago
77. 2de898a MLBEDSW-4870: Update to TensorFlow 2.7 by Rickard Bolin · 2 years, 5 months ago
78. 36da8d3 Optimize tensor allocation verification by Johan Alfvén · 2 years, 4 months ago
79. 1c08afa MLBEDSW-5534: Enet_640_640_int8 output diff by Rickard Bolin · 2 years, 4 months ago
80. bdb1d6e Set product in driver action config struct by Jonny Svärd · 2 years, 4 months ago
81. 3bbbed6 MLBEDSW-4704: Crash when loading empty constant tensors by erik.andersson@arm.com · 2 years, 5 months ago
82. 91c5a14 MLBEDSW-5740: Fix assert when setting address on identical LUT tensors by Johan Alfvén · 2 years, 5 months ago
83. 51a8dce MLBEDSW-5844: Inconsistent calculation of read shapes by Tim Hall · 2 years, 5 months ago
84. 3df5b96 MLBEDSW-5834: split shape is None when split offset is not by Tim Hall · 2 years, 5 months ago
85. 3751aa4 MLBEDSW-5629: MLCE: Model falling when creating explicit_padding by Tim Hall · 2 years, 5 months ago
86. 7d7cb67 MLBEDSW-5554: Place MEAN op exceeding max height with axis==1 on CPU by Rickard Bolin · 2 years, 5 months ago
87. e3dd2f3 MLBEDSW-5717 Fix for sigmoid int16 by Patrik Gustavsson · 2 years, 5 months ago
88. eed30ea MLBEDSW-5417: Update release notes & supported ops by Dwight Lidman · 2 years, 6 months ago 3.2.0 3.2.0.rc3
89. 837c31c MLBEDSW-5507: Fix vela summary for passes by Tim Hall · 2 years, 6 months ago
90. 3584a9c MLBEDSW-3602: Output mismatch on some mobilenet_v1 int8 and int16 by Tim Hall · 2 years, 6 months ago
91. e3d18b0 TOSA: Add ifm ofm elem size into raw output by Diqing Zhong · 2 years, 6 months ago 3.2.0.rc2
92. 399c4a2 MLBEDSW-5383 npu_find_block_configs() differs between v2.1.1 and v3.1.0 by James Ward · 2 years, 7 months ago 3.2.0.rc1
93. e6607c5 MLBEDSW-5496: Revert patch for MLBEDSW-4704 by Dwight Lidman · 2 years, 6 months ago
94. 9338978 MLBEDSW-5410 Clarify "Subgraph IO Summary" output by James Ward · 2 years, 7 months ago
95. cc8569f MLBEDSW-5209 Vela: output diff depthwise with non-zero zero points by Fredrik Svedberg · 2 years, 7 months ago
96. b1a9a92 MLBEDSW-5437: MLCE: Vela ignores back2back Activations in model by Tim Hall · 2 years, 7 months ago
97. ba2555e MLBEDSW-4925: Fix resize bilinear attribute check by erik.andersson@arm.com · 2 years, 7 months ago
98. 0af0d38 MLBEDSW-4704: Vela crashes when serializing subset of Inception v3 network by Alex Matthews · 2 years, 9 months ago
99. 3e13434 MLBEDSW-5440 MLCE: Vela crashing due to mismatch in padding shape in OFM by James Ward · 2 years, 7 months ago
100. 0ae2848 MLBEDSW-5450 MLCE: Vela to handle skip Tensor by Fredrik Svedberg · 2 years, 7 months ago