Questions tagged [pytorch]
PyTorch is an open-source deep learning framework and API that creates a Dynamic Computational Graph, which allows you to flexibly change the way your neural network behaves on the fly and is capable of performing automatic backward differentiation.
pytorch
24,206
questions
0
votes
1
answer
28
views
Simple DiffSharp example
How to convert this simple PyTorch snippet to DiffSharp? It's supposed to be similar but many functions are not to be found.
# Define tensors with requires_grad=True to track computation history
x = ...
- 171
-4
votes
0
answers
26
views
PYTHON GIVING ERROR ON MODEL TRAINING ON GPU [duplicate]
I am trying to run a YOLO V8 model training on some images on GPU. After installing CUDA TOOLKIT and cuDNN matching versions with pytorch finally the code ran but giving some kind of error which I ...
1
vote
1
answer
24
views
--shm_size change for Docker in Nextflow pipeline: containerOptions not seeming to work
I am trying to increase the shared memory size for the Docker image being used to run a Nextflow pipeline. Per some indications online, I have added the following to my config file:
docker {
...
- 419
0
votes
0
answers
18
views
Cuda 121 - ERROR : Could not find a version that satisfies the requirement torch (from versions: none)
Trying to implement cuda 12.1 for some project, impossible to download torch, torchvision and torchaudio. Went to PyTorch.org but the command line (both from stable version and nightly version) e.g. ...
0
votes
0
answers
14
views
Runnig a simple model on PyTorch and Flask on NPU/GPU
My code is below: The code works when I run it on CPU however, when I move to a device such as NPU I get an error in the step when I try to move the tensor to the device tensor = tensor.to(device). ...
- 1,047
0
votes
0
answers
6
views
How to use an Iterable dataloader for training in pytorch
This is the code i use for creating my dataloaders:
import torch
import os
import pickle
import numpy as np
from torch.utils.data import DataLoader, IterableDataset
from datasets import Dataset, ...
0
votes
0
answers
12
views
Training LLM uses unexpected amount of GPU memory
I'm training model with self-implemented training loops. A 1.5B Qwen2 occupies 40G of GPU memory. When I did the same training using llama factory, it only takes about 24G.
I tried to delete some ...
- 13
0
votes
0
answers
8
views
Instance segmentation of very close objects of the same class
I have a project where I have to detect and segment a variable number of cross section layers of a given material. Each layer is right above or below of the other one and the bounding boxes overlap on ...
-3
votes
0
answers
28
views
Why the Intel Iris performs better than RTX 3050 laptop GPU in training a Yolov8 model [closed]
I have Dell G-15 Gaming laptop with Core i5 12500h and Rtx 3050. I am new to deep learning and AI. I am training a model with Yolo v8 for Automated Number Plate Detection with a custom dataset of 28k ...
-2
votes
0
answers
38
views
More CPU cores lead to less performance in Deep Reinforcement Learning
We are working on a deep RL project and are encountering an issue where switching the training from our laptops to a workstation with more CPU cores causes the performance to drastically drop.
An ...
- 7
0
votes
0
answers
12
views
Set torch.nn.Parameter's data, including the gradient
I have an existing nn.Parameter, and a tensor with a gradient. I want to set the tensor to the parameter in a way that the parameter will have the same gradient (including perhaps .grad_fn and so on). ...
- 73
0
votes
0
answers
17
views
Finetuning BERT on classification task, tensor device mismatch error
I'm having trouble on fine-tuning a BERT model on a classification task, as I'm quite new to this. My data is composed of two columns, "item_title" (my input) and "meta_categ_id" (...
0
votes
1
answer
19
views
When calling torchvision.transforms.Normalize and converting to PIL.Image, how are values above 1 handled?
Applying torchvision.Normalize should lead to values below 0 and above 1, and thus below 0 and 255 when switching to integer values.
However, when watching the values, this seems NOT to be the case. ...
- 340
0
votes
0
answers
15
views
Inconsistent tensor device management detected by autograd
In my code, I encountered issues with tensor device management, specifically with tensors stored in custom containers (TensorPack objects). Initially, tensors were defaulting to CPU, even when the ...
- 149
1
vote
1
answer
60
views
Generate n random 2D points within a valid region
I want to generate a fixed number of random-uniformly distributed xy points within a valid mask.
The generated points should be continuous, i.e. can be "between pixels".
I'm looking for an ...
- 33