Tensor wrong shape (swag)

[klxiao]

I tried following the instructions in @38 but ran into the following bug when computing the tensors:
ValueError: Cannot reshape a tensor with 128 elements to shape [8,4] (32 elements) for 'loss/Reshape' (op: 'Reshape') with input shapes: [32,4], [2] and with input tensors computed as partial shapes: input[1] = [8,4].

To note, my input_ids/input_masks/segment_id's all have shape [4,max_seq_length], my model_fn rehapes to [32, max_seq_length], and I used the exact code for computing logits as run_classifier.py except that I reshaped logits to be [8,4], which resulted in the aforementioned error.

[klxiao]

Resolved!

[omerarshad]

@klxiao can you please share your code?