COMMITS
/ backends/arm/tosa_quant_utils.py May 16, 2025
P
Arm backend: Fix TOSA 1.0 node visitor for sum (#10908)
Per Åstrand committed
May 6, 2025
O
Arm backend: Rename const-tensors for TOSA 1.0
Oscar Andersson committed
May 2, 2025
O
Arm backend: Update rescale to handle more dtypes
Oscar Andersson committed
S
Arm backend: Update Rescale and affected nodes to support TOSA 1.0 (#10656)
SaoirseARM committed
April 24, 2025
D
Arm backend: update quantizer/__init__
Digant Desai committed
April 23, 2025
D
Revert "Arm backend: Populate __init__.py for quantizer and Arm root" (#10395)
Digant Desai committed
April 22, 2025
O
Arm backend: Populate __init__.py for quantizer and Arm root (#10351)
Oscar Andersson committed
April 10, 2025
D
Arm backend: Tosa tools update (#9451)
Digant Desai committed
March 27, 2025
O
Arm backend: Add missing __init__.py to passes (#9710)
Oscar Andersson committed
March 25, 2025
E
Arm backend: Add ComputeConstantOpsAOT pass (#9504)
Erik Lundell committed
March 3, 2025
P
Arm backend: Allow lists as input for rescales
Per Åstrand committed
February 26, 2025
P
Arm backend: Annotate types (#8723)
per held committed
January 28, 2025
P
Lintrunner: Enable mypy testing on backends/arm (#7776)
per held committed
January 20, 2025
P
[Arm] Cleanup unused functions (#7725)
Per Åstrand committed
January 17, 2025
O
Revert "Remove unused functions for quantization handling" (#7724)
Oscar Andersson committed
P
Remove unused functions for quantization handling (#7700)
Per Åstrand committed
January 7, 2025
O
Arm backend: qdq folding support for remaining operators (#7340)
Oscar Andersson committed
December 20, 2024
P
Revert "Revert "Add full operator to fold dq/q handling" (#7351)" (#7362)
Per Åstrand committed
December 18, 2024
D
Revert "Add full operator to fold dq/q handling" (#7351)
Digant Desai committed
December 12, 2024
P
Add helper functions for Q/DQ folding pass
Per Åstrand committed
October 29, 2024
P
Add functions for usage with DQ/Q folding pass
Per Åstrand committed
December 4, 2024
S
Arm backend: Explicitly convert quantized value to int64
Sebastian Larsson committed
November 11, 2024
E
Search graph for quantization parameters (#6690)
Erik Lundell committed
November 5, 2024
D
Revert "Search graph for quantization nodes (#6452)" (#6649)
Digant Desai committed
November 4, 2024
E
Search graph for quantization nodes (#6452)
Erik Lundell committed
October 21, 2024
O
Arm backend: Add layer_norm decomposition (#6288)
Oscar Andersson committed
September 11, 2024
G
Add missing Pyre mode headers] [batch:21/424] [shard:6/N]
generatedunixname89002005307016 committed
September 5, 2024
H
Buckify backends/arm for meta internal use.
hsharma35 committed
August 14, 2024
S
Use single rounding as default for TOSA lowering
SaoirseARM committed
July 24, 2024
E
Add sub operator for Arm backend (#4074)
Erik Lundell committed
July 23, 2024
E
Add sigmoid operator to Arm backend (#4114)
Erik Lundell committed
E
Add full op for Arm backend (#4073)
Erik Lundell committed
July 11, 2024
P
Quantization types (#4094)
Per Åstrand committed
May 21, 2024
E
Refactor arm quantization (#3616)
Erik Lundell committed
April 24, 2024
F
Fix for TOSA BI clamp ops (#3092)
Fredrik Knutsson committed
March 22, 2024
F
Add convolution unit test for Arm backend (#2427)
Fredrik Knutsson committed
March 20, 2024
J
Temporary fix of different memory formats between Pytorch and TOSA (#2371)
Jerry Ge committed
January 9, 2024
J
Reformat function names to match pep 8 convention (#1503)
Jerry Ge committed
December 8, 2023
J
Replace Linear lowering of using Matmul with Conv2d (#1336)
Jerry Ge committed
November 2, 2023
J
Refactor TOSA backend preprocess to use visitor pattern (#1099)
Jerry Ge committed
October 9, 2023
T
Add initial lowering of aten.convolution to tosa.conv2d support (#615)
TatWai Chong committed
October 4, 2023
T
Add `buildRescale` helper function with scale width and double rounding option (#567)
TatWai Chong committed
October 3, 2023
J
Add quantized linear layer lowering (#549)
Jerry-Ge committed
September 29, 2023
J
First commit of quantized lowering of Add (#495)
Jerry Ge committed