search

google-coral / edgetpu

Coral issue tracker (and legacy Edge TPU API source)
https://coral.ai
Apache License 2.0
411 stars 124 forks source link

Vertex AI tflite converted to EdgeTPU poor performance #826

Open bhakav opened 4 months ago

bhakav commented 4 months ago

Description

We have created a model with 100 hrs or training on vertex ai. Exported the model successfully and converted successfully into edge_tpu model .

Edge TPU Compiler version 16.0.384591198 Input: 100hr_tflite.tflite Output: 100hr_tflite_edgetpu.tflite

Click to expand! ### Issue Type Performance ### Operating System Mendel Linux, Ubuntu ### Coral Device Dev Board, USB Accelerator ### Other Devices _No response_ ### Programming Language Python 3.6, Python 3.7, Python 3.8 ### Relevant Log Output ```shell Log information Edge TPU Compiler version 16.0.384591198 Input: 100hr_tflite.tflite Output: 100hr_tflite_edgetpu.tflite Operator Count Status CONCATENATION 2 Operation is otherwise supported, but not mapped due to some unspecified limitation RESHAPE 36 Tensor has unsupported rank (up to 3 innermost dimensions mapped) RESHAPE 2 Operation is otherwise supported, but not mapped due to some unspecified limitation RESHAPE 8 Mapped to Edge TPU PACK 24 Tensor has unsupported rank (up to 3 innermost dimensions mapped) CUSTOM 1 Operation is working on an unsupported data type LOGISTIC 16 Mapped to Edge TPU LOGISTIC 1 Operation is otherwise supported, but not mapped due to some unspecified limitation REDUCE_MAX 8 Mapped to Edge TPU MUL 16 Mapped to Edge TPU DEPTHWISE_CONV_2D 45 Mapped to Edge TPU MEAN 8 Mapped to Edge TPU ADD 31 Mapped to Edge TPU MAX_POOL_2D 15 Mapped to Edge TPU CONV_2D 89 Mapped to Edge TPU Operator Count Status CONCATENATION 2 Operation is otherwise supported, but not mapped due to some unspecified limitation RESHAPE 36 Tensor has unsupported rank (up to 3 innermost dimensions mapped) RESHAPE 2 Operation is otherwise supported, but not mapped due to some unspecified limitation RESHAPE 8 Mapped to Edge TPU PACK 24 Tensor has unsupported rank (up to 3 innermost dimensions mapped) CUSTOM 1 Operation is working on an unsupported data type LOGISTIC 16 Mapped to Edge TPU LOGISTIC 1 Operation is otherwise supported, but not mapped due to some unspecified limitation REDUCE_MAX 8 Mapped to Edge TPU MUL 16 Mapped to Edge TPU DEPTHWISE_CONV_2D 45 Mapped to Edge TPU MEAN 8 Mapped to Edge TPU ADD 31 Mapped to Edge TPU MAX_POOL_2D 15 Mapped to Edge TPU CONV_2D 89 Mapped to Edge TPU ********************************************************************************** edgetpu_compiler -a 100hr_tflite.tflite Edge TPU Compiler version 16.0.384591198 Started a compilation timeout timer of 180 seconds. Model compiled successfully in 2821 ms. Input model: 100hr_tflite.tflite Input size: 2.76MiB Output model: 100hr_tflite_edgetpu.tflite Output size: 4.89MiB On-chip memory used for caching model parameters: 3.11MiB On-chip memory remaining for caching model parameters: 1014.75KiB Off-chip memory used for streaming uncached model parameters: 0.00B Number of Edge TPU subgraphs: 5 Total number of operations: 302 Operation log: 100hr_tflite_edgetpu.log Model successfully compiled but not all operations are supported by the Edge TPU. A percentage of the model will instead run on the CPU, which is slower. If possible, consider updating your model to use only operations supported by the Edge TPU. For details, visit g.co/coral/model-reqs. Number of operations that will run on Edge TPU: 236 Number of operations that will run on CPU: 66 See the operation log file for individual operation details. Compilation child process completed within timeout period. Compilation succeeded! ```
bhakav commented 4 months ago

Is this board monitored ?