GX Kok

What does FixedDropout layer does?

5

Dear author, your work on implementing EfficientNets for Keras is much appreciated. I'm using your **segmentation_models** library which in turn uses this library. I trained a model in pure Keras...

Dear authors, I have read your paper Lee, C.-H., Liu, Z., Wu, L., & Luo, P. (2019). MaskGAN: Towards Diverse and Interactive Facial Image Manipulation. http://arxiv.org/abs/1907.11922 and am interested to...

Less than k results returned

6

Hi, I'm trying to use nmslib to return nearest neighbours less than a certain threshold distance. To do this, I am setting a high value for k. I found that...

How to add samples in batches?

12

I have some 10 M vectors and I tried to build an index by batch. Here is what I have and they are wrong. ``` # Instantiate index index =...

Permission for commercial use?

2

According to your README.md, we can only use this for research purposes. Is it possible for us to train on our own data and then used the resultant trained model...

How to do high resolution unconditional sampling?

3

The high resolution images from this repo seems to be sampled from conditional transformer models. How do I sample a high resolution image from an unconditional transformer? Do I have...

Mean BGR values for PyTorch pretrained model?

3

What is the mean BGR values for the PyTorch pretrained model? [104.00699, 116.66877, 122.67892, 137.86], [103.939,116.779,123.68], or [104.00699, 116.66877, 122.67892] But they seemed to be quite similar.

Regarding bounding box values during training and testing.

1

During training, we fit in batches of images of the same dimensions for training, 512 in code by default. During training the position of activated pixels to the rotated box...

Text-to-image LDM training loss stucked at around 1.00

4

Hi, I took a first stage KL-regularized autoencoder from one of the pretained models (models/first_stage_models/kl-f8/model.ckpt) and tried to train a LDM model. Training proceeds, but the loss ((train/loss_simple_step)) hardly budge....

In the config file `configs/stable-diffusion/v1-finetune_unfrozen.yaml`, ``` personalization_config: target: ldm.modules.embedding_manager.EmbeddingManager params: placeholder_strings: ["*"] initializer_words: ["sculpture"] per_image_tokens: false num_vectors_per_token: 1 progressive_words: False ``` Should the parameters here be changed, especially `placeholder_strings` and...