-
Hello, I encountered a problem while using diffvg. I used a ResNet18 network added with a 3-layer MLP network to predict the control point coordinates of a path, which was rendered by diffvg to obtain…
-
Hi,
This is more of a question than a feature request which I don't know where else to post.
So I'm trying to perform quantization aware training to a model that's not of tf.keras.Model type but o…
-
Wonderful work!
May I know the compatibility with ZeRO mechanism? E.g., Torch redundancy optimizer, deepspeed zero-1 to zero-3, and fairscale FSDP. Becaused I noticed that QLoRA relies on particula…
-
https://github.com/huaishuiweizhu/stylegan2-tiny/blob/e4537803b4ffb891c7ce0fb5a20a83829e3278b3/train.py#L141-L154
This implentation repeat train the same data with accumulation steps. I think you m…
-
Your code currently operates one image at a time, is it possible to adjust the batch_size operation? I look forward to hearing from you!
-
Hello,
I would like to contribute to PyTorch with a new feature. I am currently working on research using ESNs and have written some ESN models in PyTorch for the GPU and CPU. I think these could b…
-
## Issue description
Hello.
During training, after replacing a parameter, gradients are not being correctly updated.
Specifically:
There is a fixed tensor adj_mat without gradients.
A m…
ithok updated
9 months ago
-
Why is there a external dependency for computing the cross-correlation between feature pyramids? If we assume that f1 and f2 are the features for image1 and image2, I believe more or less a code like …
-
The main limitation of LLMs is the huge model size, plus, during training, the required VRAM/RAM necessary to store the model + the backpropagation parameters are much higher than during inference.
A…
-
I'm curious what the current list of items needed to catch native up to the current feature set is, across all the repositories. (My ulterior motive here is not having CUDA-compatible hardware right n…