How to save finetuned model DETR Tensorlfow

[Protozzzer]

A use my own data in csv to finetune DETR tensorflow model. I cant find any implementation in source code how to save tensorflow model after each epoch. I want to implement saving weights, but there are some errors after this. Maybe you have some examples or recommendations how to save finetuned weights during or after training process.

[nathanterroir]

I am also having an issue saving the model after training using either the Keras API or saved_model formats. The specific error is:
Couldn't save model due to decoder_layer_call_fn(target, memory, target_mask, memory_mask, target_key_padding_mask, memory_key_padding_mask, pos_encoding, query_encoding, training) missing 1 required argument: memory

@Protozzzer , I was able to save model weights using detr.save_weights("save_path") where detr is the model I trained on my custom csv dataset.

[simpad2409]

Hi @Protozzzer ,
if you need help write me at [email protected] .
I have already solved this problem. Good day.

[Protozzzer]

Hi @nathanterroir Thx for your answer)
Hi @simpad2409
Can you put here your answer how to solve problem with saving full model ?

[simpad2409]

@Protozzzer In the finetuning file add :

for epoch_nb in range(180):
    if epoch_nb > 0:
        # After the first epoch, we finetune the transformers and the new layers
        config.train_transformers.assign(True)
        config.transformers_lr.assign(1e-4)
        config.nlayers_lr.assign(1e-3)

    training.eval(detr, valid_dt, config, class_names, evaluation_step=100)
    training.fit(detr, train_dt, optimzers, config, epoch_nb, class_names)

    detr.save_weights("detr-my-model.ckpt")

or, to save each checkpoint, add:
ckp = "detr-my-model-"+str(epoch_nb)+".ckpt"
detr.save_weights(ckp)

@Protozzzer Which dataset are you finetuning on?

I'm doing it on the Crowd Human dataset, but I think I'm doing something wrong ... The model I get is very bad!
Could anyone help me? Thank you so much. I would be immensely grateful.