8 Replies Latest reply on Aug 15, 2018 11:39 PM by Intel Corporation

    TensorFlow 501 Week 6

    andrewblance

      Hi,

       

      I am working my way through the TensorFlow 501 course and have got slightly stuck on Week 6. I'm slightly unsure what the question "If I tell you that the inner-most values come as `W,b` pairs, can you get the *shapes* of the coefficients?" is asking of me. What does it mean by "inner-most"?  Does it mean 'fc6'?

      Apart from that I am struggling with how you would add these to the convolutional layer slightly further on in the notebook.

       

      I have completed the Machine Learning 501 course and the first 5 weeks of TensorFlow 501 and throughout both of them it references a set of answers. I would like to check what I have done against these but I cannot find them anywhere! Could anyone help me locate them? It would help a lot.

       

      Thanks,

      Andrew

        • 1. Re: TensorFlow 501 Week 6
          Intel Corporation
          This message was posted on behalf of Intel Corporation

          Hi 
          Thanks for reaching us. We will look into it and get back to you.

          Thanks

          • 2. Re: TensorFlow 501 Week 6
            andrewblance

            thank you!

            • 3. Re: TensorFlow 501 Week 6
              Intel Corporation
              This message was posted on behalf of Intel Corporation

              Hi
              """
              I am working my way through the TensorFlow 501 course and have got slightly stuck on Week 6. I'm slightly unsure what the question "If I tell you that the inner-most values come as `W,b` pairs, can you get the *shapes* of the coefficients?" is asking of me. What does it mean by "inner-most"?  Does it mean 'fc6'?
              """"


              For AlexNet the layers are 5 conv layers are 3 FC layers. Inner layers are the layers before FC layers - the 5 conv layers conv1, conv2, conv3, conv4, conv5. The innermost should be the layer before the FC6. But you can do the printing of weights and bias of all the conv layers

              """
              Apart from that I am struggling with how you would add these to the convolutional layer slightly further on in the notebook.
              """

              For adding pretrained weights and bias for convolutional layer.
                a)You need to fetch pretrained weights.
                b)Load them into variables using tf.variable function
                c)Use tf.nn.conb2d function to add them for convolutional layer

              """
              I have completed the Machine Learning 501 course and the first 5 weeks of TensorFlow 501 and throughout both of them it references a set of answers. I would like to check what I have done against these but I cannot find them anywhere! Could anyone help me locate them? It would help a lot
              """

              These are all about building code and check the results for learning purpose.


              Thanks

              • 4. Re: TensorFlow 501 Week 6
                Intel Corporation
                This message was posted on behalf of Intel Corporation

                Hi
                Small typo in the above reply. 
                """
                I am working my way through the TensorFlow 501 course and have got slightly stuck on Week 6. I'm slightly unsure what the question "If I tell you that the inner-most values come as `W,b` pairs, can you get the *shapes* of the coefficients?" is asking of me. What does it mean by "inner-most"? Does it mean 'fc6'?
                """"


                For AlexNet the layers are 5 conv layers are 3 FC layers. Inner layers are the layers before FC layers - the 5 conv layers conv1, conv2, conv3, conv4, conv5. The innermost should be the layer before the FC6. But you can do the printing of weights and bias of all the conv layers

                """
                Apart from that I am struggling with how you would add these to the convolutional layer slightly further on in the notebook.
                """

                For adding pretrained weights and bias for convolutional layer.
                a)You need to fetch pretrained weights.
                b)Load them into variables using tf.variable function
                c)Use tf.nn.conv2d function to add them for convolutional layer

                """
                I have completed the Machine Learning 501 course and the first 5 weeks of TensorFlow 501 and throughout both of them it references a set of answers. I would like to check what I have done against these but I cannot find them anywhere! Could anyone help me locate them? It would help a lot
                """

                These are all about building code and check the results for learning purpose.


                Thanks

                • 5. Re: TensorFlow 501 Week 6
                  Intel Corporation
                  This message was posted on behalf of Intel Corporation

                  Hi
                  Can you please let us whether the issue is resolved or not?

                  Thanks

                  • 6. Re: TensorFlow 501 Week 6
                    Intel Corporation
                    This message was posted on behalf of Intel Corporation

                    Hi 
                    Small typo,
                    Can you please let us know whether the issue is resolved or not?

                    Thanks

                    • 7. Re: TensorFlow 501 Week 6
                      andrewblance

                      yeah, i have it working now,

                      Thank you

                      • 8. Re: TensorFlow 501 Week 6
                        Intel Corporation
                        This message was posted on behalf of Intel Corporation

                        Hi
                        Thanks for your response. I am closing this thread. If any issue, Please raise a new thread.

                        Thanks