timctho / convolutional-pose-machines-tensorflow

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

compile tflite cpm hand model to edge TPU

akssieg opened this issue · comments

I successfully converted the provide checkpoints (cpm_hand) to tflite model and presently trying to compile the tflite model for edge TPU. During compilation the edge TPU compiler provides the following output:

`Edge TPU Compiler version 14.1.317412892

Model compiled successfully in 68 ms.

Input model: tflite_model/hand_quant.tflite
Input size: 23.38MiB
Output model: hand_quant_edgetpu.tflite
Output size: 23.38MiB
On-chip memory used for caching model parameters: 0.00B
On-chip memory remaining for caching model parameters: 0.00B
Off-chip memory used for streaming uncached model parameters: 0.00B
Number of Edge TPU subgraphs: 0
Total number of operations: 42
Operation log: hand_quant_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: 0
Number of operations that will run on CPU: 42

Operator                       Count      Status

CONV_2D                        31         Operation is working on an unsupported data type
QUANTIZE                       6          Operation is otherwise supported, but not mapped due to some unspecified limitation
CONCATENATION                  2          Operation is working on an unsupported data type
MAX_POOL_2D                    3          Operation is working on an unsupported data type`

This shows that none of the layers are executing on the TPU and will use the CPU instead.

Anybody tried and succeeded with the full edge TPU compilation?