TOSA Reference Model 0.80 Release
Conformance test repository is now deprecated, conformance test
generation scripts are in the reference model repository
Deprecated TensorFlow Lite based conformance tests
Framework tests still generate TFL tests
Add support for new operators
DIM operator for shape inference
New unsigned attributes for RESCALE input and output
Improve TOSA flatbuffer compatibility checks
Fixes for test and reference model cleanup
Enable reading/writing native data types for numpy files
Various other cleanup, see the git revision history for complete details
41542cf Remove draft tag for 0.80 release by Eric Kunze · 9 months ago v0.80.0 84b00fd Support new RESCALE attributes by Eric Kunze · 9 months ago fb73fc1 Update TosaVersion compatibility check to match with serialization_lib by Jerry Ge · 10 months ago 0913dba [reference model] Code cleanup for SHAPE by Tai Ly · 10 months ago b77c2bb Remove unused struct to avoid warning message by Won Jeon · 10 months ago a21b2e8 Add DIM operator to reference model by Won Jeon · 10 months ago a641dd5 Add support for bias broadcasting by Tai Ly · 10 months ago be2e87c Update formatting based on clang-format 14 by Eric Kunze · 10 months ago 17fab3b Fix compiler warnings (NFC) by Eric Kunze · 11 months ago ed1a15b Fixed missing tensor inputs in fully_connected in model runner by Jiacheng Liang · 11 months ago e5cabbf Enable lazy tensor allocation by Jerry Ge · 11 months ago 9c9c8da Run clang-format and update copyright by Jerry Ge · 11 months ago 9a36755 Update tensor allocation logging by Eric Kunze · 11 months ago e7c7cab Add support for int32 tensor datatype in model runner by Jiacheng Liang · 11 months ago edac6ab Fix missing template instantiations by Eric Kunze · 12 months ago 78155c6 Add support for ERF operator to reference model by Won Jeon · 1 year ago eb74106 Upgrade to latest version of TOSA specification by Grant Watson · 12 months ago c8da1d2 Make verifiers to operate on const pointers by Georgios Pinitas · 12 months ago 135c955 Add ERROR_IF to incorrect broadcast shapes by Jerry Ge · 1 year, 1 month ago 41df428 Add TOSA MI verification methods by Georgios Pinitas · 1 year, 1 month ago 6168047 Correctly identify "axis" attributes. by Grant Watson · 1 year ago eb52cc1 Add support for boolean outputs in model runner by Jiacheng Liang · 1 year, 1 month ago 56a3a06 Update version to 0.80 draft by Eric Kunze · 1 year ago 9ccad3f Remove draft tag for 0.70 release by Eric Kunze · 1 year, 1 month ago v0.70 v0.70.0 307392a Add abs calculations under precise_mode by Tai Ly · 1 year, 1 month ago 264f7fa Add support for one dimension of size -1 in ReshapeOp by Jerry Ge · 1 year, 2 months ago 0bd4ec8 Refactor ref_model rank checking and add level check to argmax by Jerry Ge · 1 year, 1 month ago a4d748b [reference model] Add precise mode by Tai Ly · 1 year, 3 months ago 44eb88d AVG_POOL2D - Fix for stride renaming mistake by Jeremy Johnson · 1 year, 2 months ago a793f46 Add level checking to TOSA Ref model by Jerry Ge · 1 year, 2 months ago abf8718 revert to GetBlockByName "main" by Tai Ly · 1 year, 2 months ago 4e9a977 [reference model] support multiple regions by Tai Ly · 1 year, 3 months ago cf305db Refactor 1L to INT64_C() by Jerry Ge · 1 year, 3 months ago 2eb3d63 Update version to 0.70.0 draft by Eric Kunze · 1 year, 3 months ago 17def5f Update version number to 0.60 by Eric Kunze · 1 year, 3 months ago v0.60.0 57bc079 Update CAST fp to int rounding by Eric Kunze · 1 year, 5 months ago a4e48ca Update rank limits for SLICE, TILE and TRANSPOSE by Luke Hutton · 1 year, 4 months ago 5728713 Add FFT2d to the reference model by Luke Hutton · 1 year, 4 months ago d34b3fc Remove accumulator attributes from all but AVG_POOL2D by James Ward · 1 year, 5 months ago 736fd1a Create MI tests for Type Conversion: CAST by James Ward · 1 year, 5 months ago 261b7b6 Add RFFT2d to the reference model by Luke Hutton · 1 year, 5 months ago eef8667 Fix for sign extending LOGICAL LEFT/RIGHT SHIFT results by Jeremy Johnson · 1 year, 5 months ago 3407125 Create MI tests for Activation: CLAMP; Data Layout: PAD by James Ward · 1 year, 6 months ago df628d4 Update for RESCALE spec apply_add clarification by Jeremy Johnson · 1 year, 5 months ago 9e94af8 Reference model update for control flow operators support by Jerry Ge · 1 year, 8 months ago 98c281f Add explicit template instantiations for abstract classes by Jared Smolens · 1 year, 6 months ago 3d3d45d Add BF16 support to IModelRunner by James Ward · 1 year, 7 months ago 7f1ea8e Create MI tests for Image: RESIZE by James Ward · 1 year, 6 months ago 64285a1 Extend reference model API with eager operator execution entrypoints by Grant Watson · 1 year, 7 months ago c8774a1 Update version to 0.51.0 draft by Eric Kunze · 1 year, 6 months ago a682749 Fix reference model memory leaks for the following ops by Jerry Ge · 1 year, 7 months ago 417d874 Update version to 0.50 for release by Eric Kunze · 1 year, 6 months ago v0.50 v0.50.0 fa88828 Fix error messages/comments using old -C arguments. by Jeremy Johnson · 1 year, 7 months ago ee25669 FP16 improvements by James Ward · 1 year, 7 months ago a633bc6 Fix the floating point precision issue for Sigmoid FP32 by Jerry Ge · 1 year, 7 months ago 52460a8 Zero point addition in RESCALE should be int32 by Eric Kunze · 1 year, 7 months ago 24dbc42 Add BF16 support to reference model by James Ward · 1 year, 8 months ago 2e4d889 Add FP16 support to IModelRunner by Matthew Sloyan · 1 year, 8 months ago 4196491 Update to check tensor size is valid by Eric Kunze · 1 year, 8 months ago bc2a3db Rename FLOAT type to FP32 by Jeremy Johnson · 1 year, 9 months ago 93d4390 Minor fixes & add FP16 support to refmodel testing and conformance gen by Jeremy Johnson · 1 year, 9 months ago 71dfc70 Remove conflicting REQUIRE definition in unit tests by James Ward · 1 year, 8 months ago 8b39043 Reference model changes for fp16 support by James Ward · 1 year, 10 months ago ba5fad3 Add IModelRunner interface to TOSA Reference Model by Matthew Sloyan · 1 year, 9 months ago 366f297 Fix debug masks in refmodel. by Jeremy Johnson · 1 year, 9 months ago 45e42f4 Update reference model version to 0.41 draft by Eric Kunze · 1 year, 9 months ago a8206e3 Update revision to 0.40. by Eric Kunze · 1 year, 10 months ago v0.40 v0.40.0 fd62905 Update framework test generator to support TF/TFL conv3d. by TatWai Chong · 1 year, 11 months ago c1a9783 Align padding for transpose_conv2d to match spec by Eric Kunze · 2 years ago d511f9e Enabled 16-bit TABLE REQUIRE statement by Jerry Ge · 1 year, 10 months ago 286f834 Simplify the argument parsing and debug code by Eric Kunze · 2 years ago 3c59d5d Check that the shape calculation for PAD is correct by Eric Kunze · 1 year, 10 months ago f732609 Update TOSA resize to match specification by TatWai Chong · 2 years ago 750d27d Modify input limits for apply_scale_32 by Eric Kunze · 2 years ago 4b85a0b Update reference model version to 0.31 draft by Eric Kunze · 2 years ago f733783 Fix reference model use of weight zero point by Eric Kunze · 2 years ago v0.30 v0.30.0 95dfcb8 Remove draft status from reference model by Eric Kunze · 2 years ago b5fabec Remove quantization info from serialization attributes by Eric Kunze · 2 years ago 24594f5 Update transpose_conv2d to align with TOSA spec by TatWai Chong · 2 years ago 61f6622 Remove dilation from transpose_conv2d op by Eric Kunze · 2 years ago ec0327d Change size check in TILE to ERROR_IF by Eric Kunze · 2 years, 1 month ago 86c403b Align the serialization schema with TOSA 0.24.0 specification by TatWai Chong · 2 years ago c23fc3b Remove RESHAPE -1 dimensions support by Jeremy Johnson · 2 years, 1 month ago f7f78ae Add support for uint16_t to RESCALE by Jeremy Johnson · 2 years, 1 month ago 7edb34c Initialize accumulator with bias value for CONV3D by Eric Kunze · 2 years, 1 month ago 67a9155 Adjust divisor calculation by Eric Kunze · 2 years, 4 months ago 0e46364 Fix for NEGATE using 32-bit accumulator by Jeremy Johnson · 2 years, 1 month ago 4a6fb9b Update tensor ops ERROR_IF criteria by Jeremy Johnson · 2 years, 2 months ago 7de9b45 Add missing REQUIREs check to REDUCE_SUM in refmodel by Jeremy Johnson · 2 years, 2 months ago 81ee53d Add missing REQUIRE to NEGATE op by Jeremy Johnson · 2 years, 3 months ago 62a7b7f Usability cleanup by Jared Smolens · 2 years, 3 months ago 615feee Fix refmodel LOGICAL_XOR (was incorrectly LOGICAL_OR) by Jeremy Johnson · 2 years, 3 months ago 42c9bae Update refmodel apply_scale_32: adjust range checking by Jeremy Johnson · 2 years, 4 months ago 830add4 Rework divisor calculation for AVGPOOL2D by Eric Kunze · 2 years, 5 months ago 66bad80 Fix for LOGICAL_LEFT/RIGHT_SHIFT shift values by Jeremy Johnson · 2 years, 5 months ago 2131a4d Fix broadcast bug by Kevin Cheng · 2 years, 7 months ago 9fe1724 More ERROR_IF to check attribute for convolution ops by Kevin Cheng · 2 years, 7 months ago 1c3c847 Check valid broadcastable shape for binary and ternary ops by Kevin Cheng · 2 years, 7 months ago 25fbe52 Cleanup CMakeLists.txt by Kevin Cheng · 2 years, 7 months ago d4ea8f1 Update reference model version to 0.24.0 draft by Eric Kunze · 2 years, 7 months ago Next »