hongy

CODEEPNEAT restoring state results in divide by zero.

3

CODEEPNEAT restoring state results in divide by zero when training on GPU. ``` Backed up generation 17 of the CoDeepNEAT evolutionary run to file: /content/tfne_state_backups/tfne_state_backup_2021-Apr-10_13-57-22/tfne_state_backup_gen_17.json --------------------------------------------------------------------------- ZeroDivisionError Traceback (most recent...

Evaluation on GPU runs in OOM Issues when trying to allocate 10k images all at once

1

https://github.com/PaulPauls/Tensorflow-Neuroevolution/blob/55c76f08ee4e4206d842565902b3d11c517c3756/tfne/environments/cifar10_environment.py#L87 The above line instructs the GPU to make a prediction on all 10k test images at once. Please switch to the model.predict(data, batchsize) method. `predictions = model.predict(self.test_images, batch_size =...

https://github.com/PaulPauls/Tensorflow-Neuroevolution/blob/55c76f08ee4e4206d842565902b3d11c517c3756/tfne/environments/cifar10_environment.py#L64 This isn't about the VRAM, which seems to be cleared properly. The problem lies with the system RAM when running the _eval_genome_fitness_weight_training method when using a GPU. The RAM...

Link for cars196 returns 404 error

4

/!\ PLEASE INCLUDE THE FULL STACKTRACE AND CODE SNIPPET DownloadError: Failed to get url https://ai.stanford.edu/~jkrause/cars/car_devkit.tgz. HTTP code: 404. **Short description** Its just a 404 link. Need a new link. **Environment...

My understanding from the readme is that there is some flexibility in the TPU Mesh, but all operations must replicated on all TPU cores. Will there ever be support for...