'Tensor' object has no attribute 'numpy' when i run pose_estimation.py

[irojasc]

hello guys, i want to run pose_estimation.py file . But, when code enter to method predict() of class PoseEstimation:
def predict(self, img:np.ndarray, output_path:str, debug=True):
(more code here)

    joint_heatmap = np.squeeze(joint_heatmap.numpy())
    joint_locations = np.squeeze(joint_locations.numpy())
    joint_offsets = np.squeeze(joint_offsets.numpy())
    heatmap = np.squeeze(heatmap.numpy())
    offsets = np.squeeze(offsets.numpy())
    whs = np.squeeze(whs.numpy())```

this message apears : 'Tensor' object has no attribute 'numpy' . I suppose this trouble because numpy() work with eager mode enable. i´ve tried to enable eager mode but it didnt work yet with the followings commands:
tf.compat.v1.enable_eager_execution() and self.model.run_eagerly = True **but tf.executing_eagerly() prints False
Finally, I´m using these dependencies:

• Tensor flow 2.4.1
• keras 2.4.3
• opencv-python 4.5.1
• numpy 1.19.5
• pillow 8.1.0
• matplotlib 3.3.4
• pandas 1.2.2
• sckit-learn 0.24.1
• sckit-image 0.18.1
• imgaug 0.4.0
  I would appreciate very much help me to solve this issue. thanks

[Adeel-Intizar]

I will look into it, till then you can use CDCL Model for pose estimation

[shersoni610]

Hello,

I face the same issue with the CenterNet model. Any update on the issue?