Open FengMu1995 opened 2 years ago
Describe the bug I convert my onnx to tf saved model by onnx-tensorflow.
When I tested by onnx_tf, it runs successfully:
_# onnx_model = onnx.load("./rvm_144*256.onnx") # load onnx model
# tf_rep.export_graph("model.tf") # export the model # rec = [np.zeros([1, 1, 1, 1], dtype=np.float32)] * 4 # Must match dtype of the model. # rec[0] = np.zeros([1, 16, 72, 128], dtype=np.float32) # rec[1] = np.zeros([1, 20, 36, 64], dtype=np.float32) # rec[2] = np.zeros([1, 40, 18, 32], dtype=np.float32) # rec[3] = np.zeros([1, 64, 9, 16], dtype=np.float32) # img = cv2.imread('/home/wudi/project/Pytorch_Retinaface_1/meeting.png') # img = np.transpose(img, (2, 0, 1)) / 255. # src = np.expand_dims(img, 0) # test = src.astype(np.float32) # out = tf_rep.run([test, rec[0], rec[1], rec[2], rec[3]]) # pha = out[1]*255 # pha = np.squeeze(pha) # im = Image.fromarray(np.uint8(pha)) # im.save("out" + ".jpg") # print(out)_
However, when I tested by tf, the error :valueError callback pyfunc_0 is not found.
img = cv2.imread('./meeting.png') img = np.transpose(img, (2, 0, 1)) / 255. img = torch.from_numpy(img).unsqueeze(0).float() input = tf.constant(img.numpy()) rec = [tf.constant(0.)] * 4 saved_model = tf.saved_model.load("model.tf") rvm = saved_model.signatures["serving_default"] outputs = rvm(r1i=rec[0], r2i=rec[1], r3i=rec[2], r4i=rec[3], src=input)
I upload my model(model-onnx.zip/tf.zip) here.
model-onnx.zip
Describe the bug I convert my onnx to tf saved model by onnx-tensorflow.
When I tested by onnx_tf, it runs successfully:
_# onnx_model = onnx.load("./rvm_144*256.onnx") # load onnx model
tf_rep = prepare(onnx_model) # prepare tf representation
However, when I tested by tf, the error :valueError callback pyfunc_0 is not found.
img = cv2.imread('./meeting.png') img = np.transpose(img, (2, 0, 1)) / 255. img = torch.from_numpy(img).unsqueeze(0).float() input = tf.constant(img.numpy()) rec = [tf.constant(0.)] * 4 saved_model = tf.saved_model.load("model.tf") rvm = saved_model.signatures["serving_default"] outputs = rvm(r1i=rec[0], r2i=rec[1], r3i=rec[2], r4i=rec[3], src=input)
I upload my model(model-onnx.zip/tf.zip) here.
model-onnx.zip