Shape mismatch: inputs 32 != attn_output 1024

[code-Kasim]

I am done trying to resolve it but couldn't do the thing, if anyone can help it would be a great support, i am trying to run it for generating IR images through my custom dataset(2562563). But ended up here..after being stuck for almost a month...will add my modified code files and here is the error message: ##############
TransGAN Train
##############

Depth list: [5, 4, 3, 2, 1, 1]
Length of depth list: 6
Heads list: [4, 4, 4, 4, 4, 4]
Length of heads list: 6
MLP dim list: [4096, 2048, 1024, 512, 256, 128]
Length of MLP dim list: 6
Initializing model from scratch

batch_size : 32
noise_dim : 256
g_dim : 1024
g_depth : [5, 4, 3, 2, 1, 1]
g_heads : [4, 4, 4, 4, 4, 4]
g_mlp : [4096, 2048, 1024, 512, 256, 128]
g_initializer : orthogonal
g_learning_rate : 0.0001
g_beta_1 : 0.0
g_beta_2 : 0.99
d_dim : [512, 256, 128]
d_depth : [2, 3]
d_heads : [4, 4, 4]
d_mlp : [2048, 1024, 1024, 1024]
d_initializer : orthogonal
d_patch_size : 16
d_learning_rate : 0.0001
d_beta_1 : 0.0
d_beta_2 : 0.99
d_steps : 1
loss : wgan
gp_weight : 10.0
policy : color,translation,cutout

<class 'keras.src.metrics.reduction_metrics.Mean'>
patches shape: (32, 1024, 1024)
pos_embeddings shape: (1, 1024, 1024)
patches shape: (32, 1024, 1024)
pos_embeddings shape: (1, 1024, 1024)
patches shape: (32, 1024, 1024)
pos_embeddings shape: (1, 1024, 1024)
Traceback (most recent call last):
File "/home/user/kasim/TransGAN-main(1)/train.py", line 252, in
main()
File "/home/user/kasim/TransGAN-main(1)/train.py", line 232, in main
run_training(args)
File "/home/user/kasim/TransGAN-main(1)/train.py", line 195, in run_training
train_step(image_batch)
File "/home/user/anaconda3/envs/transgan/lib/python3.9/site-packages/tensorflow/python/util/traceback_utils.py", line 153, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/tmp/autograph_generated_filexagp5mu3.py", line 63, in tf__train_step
ag.for_stmt(ag__.converted_call(ag__.ld(range), (ag__.ld(hparams)['d_steps'],), None, fscope), None, loop_body, get_state_1, set_state_1, (), {'iterate_names': '_'})
File "/tmp/autograph_generated_filexagp5mu3.py", line 28, in loop_body
generator_output = ag.converted_call(ag__.ld(generator), (ag__.ld(noise),), dict(training=True), fscope)
File "/home/user/anaconda3/envs/transgan/lib/python3.9/site-packages/keras/src/utils/traceback_utils.py", line 122, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/home/user/kasim/TransGAN-main(1)/model.py", line 230, in call
x = self.block_8(x, training=training)
File "/home/user/kasim/TransGAN-main(1)/model.py", line 143, in build
raise ValueError(f"Shape mismatch: inputs {input_shape[-1]} != attn_output {attn_output_shape[-1]}")
ValueError: in user code:

File "/home/user/kasim/TransGAN-main(1)/train.py", line 146, in train_step  *
    generator_output = generator(noise, training=True)
File "/home/user/anaconda3/envs/transgan/lib/python3.9/site-packages/keras/src/utils/traceback_utils.py", line 122, in error_handler  **
    raise e.with_traceback(filtered_tb) from None
File "/home/user/kasim/TransGAN-main(1)/model.py", line 230, in call
    x = self.block_8(x, training=training)
File "/home/user/kasim/TransGAN-main(1)/model.py", line 143, in build
    raise ValueError(f"Shape mismatch: inputs {input_shape[-1]} != attn_output {attn_output_shape[-1]}")

ValueError: Exception encountered when calling Generator.call().

Shape mismatch: inputs 32 != attn_output 1024

Arguments received by Generator.call():
  • z=tf.Tensor(shape=(32, 256), dtype=float32)
  • training=True

will share code files if needed..thank in advance and waiting eagerly!!!