09_02_09_00

New in this Release

Description	Notes
Support for asymmetrically quantized LeakyRelu	This requires an updated version of C7x/MMA firmware (09_02_09_00) and needs to have the advaced_options:c7x_firmware set

---

Fixed in this Release

ID	Description	Affected Platforms
TIDL-4364	Softmax layer with height axis and no output transpose results in functional mismatch on EVM	All except AM62
TIDL-4355	16-bit Elementwise Mul (Unsigned x Unsigned) results in overflow in the output in target flow	All except AM62
TIDL-4266	Elementwise broadcast fails when both inputs have broadcasted dimension value = 1	All except AM62
TIDL-4034	Slice layer may functionally mismatch when preceeded by other slice/reshape layer combinations	All except AM62
TIDL-4031	Elementwise layers with repetitive inputs results into "segmentation fault" during host emulation	All except AM62
TIDL-4395	[High Precision Sigmoid] Host emulation output is wrong for '-128' input value	All except AM62
TIDL-3836	Resize-nearest neighbour fails when enableHighResOptimization flag is set	All except AM62
TIDL-4001	16-bit Softmax Kernel generates wrong output if plane width and height are not equal	All except AM62
TIDL-4381	Allowlisting failure for TIDL induced transpose when optimizing consecutive transposes in an ONNX network	All except AM62
TIDL-4380	TIDL Import gives "coeff cannot be found(or not match) in coef file, Random bias will be generated! Only for evaluation usage! Results are all random!" warning when ONNX Add/Mul operator has a constant input of size 1 in higher dimensions ([1], [1,1], .)	All except AM62
TIDL-4379	TIDL Import fails with "All the Tensor Dimensions has to be greater then Zero" when ONNX network has split with negative axis	All except AM62
TIDL-4261	Gather layer hangs on EVM when run in 16-bit	All except AM62
TIDL-4042	Compilation results in following warning and a potential segmentation fault if additional layer is added at the end of network by TIDL during compilation phase: "Warning : Couldn't find corresponding ioBuf tensor for onnx tensor with matching name"	All except AM62
TIDL-4029	ONNX QDQ model fails with "Import Error: Model with all ranges not supplied" message when it has LeakyRelu operator	All except AM62
TIDL-4028	Gather indices not parsed correctly during calibration when its datatype is INT32	All except AM62
TIDL-4012	Consecutive transpose operators do not fuse into a single transpose operator	All except AM62
TIDL-4011	Network having Reshape/Squeeze layers with (A) Multiple consumer layers and (B) No change in shape fails during compilation with error message as "missing inputs in the network and cannot be topologically sorted"	All except AM62
TIDL-4389	Asymmetric quantization results in lower accuracy compared to symmetric quantization for networks with LeakyReLU	All except AM62
TIDL-4377	Calibration for models which are not properly regularized, results in degraded accuracy	All except AM62
TIDL-4274	16-bit Elementwise Mul (Unsigned x Unsigned) results in overflow in the output	All except AM62
TIDL-4352	Transpose operator with any configuration other than (0,1,2,3) -> (0,2,3,1) or vice versa may result into functional mismatch on target	All except AM62
TIDL-4030	Model compilation may hang in networks containing 2x2 stride 2 depthwise separable convolution	All except AM62
TIDL-4405	Networks with height-wise/width-wise concatenate layers may result in segmentation fault during compilation	All except AM62
TIDL-4402	Setting option "advanced_options:high_resolution_optimization" = 1 may result in functionally incorrect output on target for some layers leading up to resize layer in the network	All except AM62
TIDL-4401	Layer level debug traces for inference are incorrect for host emulation mode on setting option "advanced_options:high_resolution_optimization" = 1	All except AM62
TIDL-4399	Depth-wise convolution layer followed by resize layer (not necessarily immediate consumer) may result in functional issue on target	All except AM62
TIDL-4376	2x2 stride 2 spatial max pooling layer with odd input dimensions functionally incorrect for useCeil = 0 attribute on target	All except AM62
TIDL-4044	Setting option "advanced_options:high_resolution_optimization" = 1 can result in segmentation fault in case of Reshape/Slice layer being present in network	All except AM62
TIDL-4043	Setting option "advanced_options:high_resolution_optimization" = 1 may result in functionally incorrect output for networks with multiple branches	All except AM62
TIDL-4041	High resolution optimization gets disabled for entire network despite setting option "advanced_options:high_resolution_optimization" = 1 in case Pad Layer is present in network	All except AM62
TIDL-3947	Segmentation fault on setting option "advanced_options:high_resolution_optimization" = 1 in cases where layer id and execution id of layer(the 2 numbers depicted for layer in visualization svg file) are not same	All except AM62
TIDL-4416	"Increased network binary size on 9.2 SDK may cause model inference to fail on EVM with following error: ""Failed to allocate memory record 13 @ space 17"""	All except AM62
TIDL-4410	TIDL Eltwise layer (Add/Mul in ONNX) gives incorrect results during host emulation inference run when using OPENACC GPU tools	All except AM62
TIDL-4394	"16-bit transpose ONNX operator with permute (0, 1, 3 , 2) may fail with following error during TIDLRT_create: ""Init: Error: DMA transfer needs more than 5 channels"""	All except AM62
TIDL-4393	"16-bit transpose ONNX operator with permute changing the last dimension (0,1,2,3 with 3 changing it's place in target permute) may give ""Workload not created for layer = N, TIDL_DataConvertLayer"" during TIDL Import"	All except AM62
TIDL-4392	TFLITE TransposeConv operator is incorrectly getting denied and being delegated to ARM in OSRT	All except AM62
TIDL-4358	Data convert layer with NHWC to NCHW format with WxCxelementSize > 320Kb results in functional mismatch on target	All except AM62
TIDL-4273	Documentation issues in edgeai-tidl-tools	All except AM62
TIDL-4027	"numOutputChannelsPerGroup > (256*K)/numParamsBit results into wrong functionality on Target, where k = 1 (AM62A, J722S) and k=2 for rest all devices"	All except AM62
TIDL-4003	Matmul in 16 bit mode may functionally fail on target if one of the input's plane Size is larger than 256KB	All except AM62
TIDL-3925	"QDQ model throws ""Could not find const or initializer of layer"" if the weights are kept in floating point in the model"	All except AM62
TIDL-3881	Segmentation fault during inference in a model with resize layers due to incorrect buffer reuse	All except AM62
TIDL-3879	"Matmul layer with inner most plane dimension ( M x K ) and ( K x N) may result in below error during compilation even when 2* 64 x K < 448 KB ""Dataflow for tensor N with high volume is not supported"" when K is large"	All except AM62
TIDL-3876	Incorrect output dimensions provided by TIDL-RT to ONNX-RT in a subgraph with multiple outputs	All except AM62
TIDL-3875	Networks resulting in situation of subgraphs on C7x with more than 1 inputs from previous sub graph on cortex-A CPU have issue	All except AM62
TIDL-3863	"Model compilation may fail for large networks with 7x7 depth wise separable layers with error message ""�Memory limit exceeded for Workload Creation�"	All except AM62
TIDL-3804	Data convert layer with NCHW->NHWC and C x W x outputElementSize > 256k results into mismatch on EVM	All except AM62

---

Known Issues

ID	Description	Affected Platforms	Occurrence	Workaround in this release
TIDL-2946	Fully grouped convolution with padding greater than 32 (8-bit inference) or padding greater than 16 (16-bit inference) results in incorrect outputs	All except AM62	Rare	None
TIDL-2947	Convolution with pad greater than the input width results in incorrect outputs	All except AM62	Rare	None
TIDL-2990	PReLU layer does not correctly parse the slope parameter and produces incorrect outputs	All except AM62	Rare	None
TIDL-3351	Model hangs during compilation when it has intermediate tensor as output	All except AM62	Rare	None
TIDL-3572	Incorrect data type for input buffer when the "add_data_convert_ops": 0 (ARM dataconversion) which results in functional mismatch	All except AM62	Rare	None
TIDL-3622	Quantization prototxt does not correctly fill information for tflite const layers	All except AM62	Rare	None
TIDL-3704	Intermediate subgraphs whose outputs are not 4D result in incorrect outputs	All except AM62	Rare	None
TIDL-3780	Prototext based scale input may result in slight degradation in quantized output	All except AM62	Rare	None
TIDL-3830	OSRT: Incorrect data type selection during float calibration pass when addDataConvert is disabled for ONNX networks	All except AM62	Rare	None
TIDL-3845	Running model compilation and inference back to back in the same python script results in a segfault	All except AM62	Rare	None
TIDL-3876	Incorrect output dimensions provided by TIDL EP to ONNX-RT in a subgraph with multiple outputs	All except AM62	Rare	None
TIDL-3880	TIDL GPU tools do not support NVIDIA GPUs with compute capability 8.9	All except AM62	Rare	None
TIDL-3883	Graph formation (ARM) failing for unsupported maxpool w/ asym stride due to extra dimensions	All except AM62	Rare	None
TIDL-3886	Maxpool 2x2 with stride 1x1 is considered supported but is incorrectly denied from being offloaded to C7x	All except AM62	Rare	None
TIDL-3895	2x2s2 Max Pooling with ceil_mode=0 and odd input dimensions results in incorrect outputs	All except AM62	Rare	None
TIDL-3897	Operators with attributes fed as runtime variable results in undefined behavior during compilation stage	All except AM62	Rare	None
TIDL-3898	Pow, Div (With 2 variable inputs) , Sqrt & Erf may get accidentally offloaded to C7x and result in incorrect outputs	All except AM62	Rare	None
TIDL-3899	Unsupported cast operators might get incorrectly offloaded and result in accuracy degradation	All except AM62	Rare	None
TIDL-3900	ONNX GEMM operator followed by a Transpose layer can result in compilation failure with message "topological sort error"	All except AM62	Rare	None
TIDL-3903	Global average pooling produces incorrect quantized output when a large plane (H*W > 1024) is being reduced	All except AM62	Rare	None
TIDL-3905	TFLite Prequantized models with "add_dataconvert_ops": 3 fails with error "Unable to split bias"	All except AM62	Rare	None
TIDL-3910	MatMul operator with signed inputs and unsigned output results in an MMALIB init error	All except AM62	Rare	None
TIDL-3917	Transpose layer with more than 4 dimensions and with the last dimension not changing gets incorrectly denied from offloading to C7x/MMA with message: "Unsupported (TIDL check) TIDL layer type --- Transpose"	All except AM62	Rare	None
TIDL-3925	QDQ model throws "Could not find const or initializer of layer" if the weights are kept in floating point in the model	All except AM62	Rare	None
TIDL-3928	Sub operator with variable input get's incorrectly offloaded to C7x and results in an init failure during inference	All except AM62	Rare	None
TIDL-3930	Limited set of operators support more than 4 dimensions (Reshape, Transpose, Batchnorm, Split/Slice, Softmax, MatMul & Eltwise) - other operators might accidentally get offloaded to C7x and produce incorrect results	All except AM62	Rare	None
TIDL-3935	Subgraphs terminating with Slice, Squeeze or Reshape result in incorrect graph formation in ONNXRuntime	All except AM62	Rare	None
TIDL-3936	Allocation failure during graph creation in ONNXRUNTIME when ConstantOfShape with unknown dimensions are present	All except AM62	Rare	None
TIDL-3937	Transpose (>4D) with batch dimension not being explicit in ONNXRT does not correctly set output shapes	All except AM62	Rare	None
TIDL-3938	Offloaded ReduceSum layer does not get correct input shape	All except AM62	Rare	None
TIDL-3939	Dynamic Quantize/Dequantize results in ONNXRUNTIME exception with offloaded subgraphs	All except AM62	Rare	None
TIDL-3941	Subgraphs with tensors whose output dimensions greater than or equal to 5 result in incorrect outputs	All except AM62	Rare	None
TIDL-4016	Fully unsupported onnx model with TIDL Compilation Execution provider results in a segmentation fault	All except AM62	Rare	None
TIDL-4017	Single operator ONNX subgraph with both inputs and outputs coming from ONNX-RT has incorrect shape information	All except AM62	Rare	None
TIDL-4018	Resize with attributes fed as variable input results in undefined output dimensions	All except AM62	Rare	None
TIDL-4019	Incorrect handling of operators expanded by ONNX (Due to implementation not existing in ONNX)	All except AM62	Rare	None
TIDL-4024	QDQ models with self-attention blocks error out during model compilation with "RUNTIME_EXCEPTION : Non-zero status code returned while running TIDL_0 node. Name:'TIDLExecutionProvider_TIDL_0_0' Status Message: CHECK failed: (index) < (current_size_)")	All except AM62	Rare	None
TIDL-4031	Network having layers with multiple inputs where one or more inputs repeat results into "segmentation fault" during host emulation	All except AM62	Rare	None
TIDL-4038	Reduce Min & Max operators do not correctly parse when the optional input for "axes" is not specified	All except AM62	Rare	None
TIDL-4039	Flatten layer with non-zero axis creates incorrect dimensions in TIDL graph	All except AM62	Rare	None
TIDL-4052	ONNX QDQ model compilation fails with error "Unable to merge Dequantize upwards" and produces a model with functional issues	All except AM62	Rare	None
TIDL-4054	ONNX QDQ models with more than one subgraph result in a segmentation fault and error "Missing dequantize layer inputs"	All except AM62	Rare	None
TIDL-4055	Models with batched input and tensors with less than 4 dimensions result in models with incorrect shape information and produce functionally incorrect outputModels with batched input and tensors with less than 4 dimensions result in models with incorrect shape information and produce functionally incorrect output	All except AM62	Rare	None

---