I'm training a huggingface xlnet-large-cased model with the following specs:
args = TrainingArguments( f"xlnet-large-finetuned", evaluation_strategy = "epoch", save_strategy = "epoch", learning_rate=2e-5, per_device_train_batch_size=1, per_device_eval_batch_size=1, num_train_epochs=3, gradient_accumulation_steps=16, weight_decay=0.01, load_best_model_at_end=True, metric_for_best_model="accuracy" )
and by calling this code: trainer = Trainer( model, args, train_dataset=tokenized_train_dataset, eval_dataset=tokenized_val_dataset, data_collator=data_collator, tokenizer=tokenizer, compute_metrics=compute_metrics ) , trainer.train().
I reduced the batch size to 1, emptied cuda cache and deleted all the variables in gc but I still get this error: RuntimeError: CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 15.78 GiB total capacity; 14.31 GiB already allocated; 2.75 MiB free; 14.78 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
Is there any way I could resolve this without having to acquire more GPU credits?
There is a method named "Mixed Precision", the idea is to convert parameters from float32 to float16 to speed up the training and reduce memory use, the detail of mixed precision.
In some repositories, you can see they implement "automatic mixed precision" by apex package. However, with the newest version of Pytorch, you can use it easily with torch.cuda.amp by wrapping the computation code in autocast() and control the gradient and loss scale by the scaler.
For a certain project purpose I am trying to store the 1 * 4096 embeddings (The output right before the final layer) of around 6000 images into a pkl file. For the same, I am running an iteration over the 6000 images on vgg16 modified model in google colab. But it returns 'CUDA out of memory. Tried to allocate 14.00 MiB (GPU 0; 15.90 GiB total capacity; 14.86 GiB already allocated; 1.88 MiB free; 342.26 MiB cached)' error.
Whereas I have used the same dataset split into test-train for training and validating my model and that runs fine. I am wondering why obtaining and storing the embedding alone is becoming a heavy task in colab.
Is there any other way I can obtain the embeddings and store in a pkl file other than the below code.
embedding = []
vgg16 = vgg16.to(device)
for x in range (0, len(inputImages)) :
input = transformations(inputImages[x]) //pre processing
input = torch.unsqueeze(input, 0)
input = input.to(device)
embedding.append(vgg16(input))
The code is interupted at the last line with the CUDA out of memory error.
The output that you have generated vgg16(input), thats still in cuda. This is so because this output is used for calculating the loss afterwards. So to avoid having your output being stored in CUDA and eat up your GPU memory, move it to CPU using .cpu().numpy(). If that throws an error, you might have to use .detach() as well to detach the variable.
I think it's a pretty common message for PyTorch users with low GPU memory:
RuntimeError: CUDA out of memory. Tried to allocate 😊 MiB (GPU 😊; 😊 GiB total capacity; 😊 GiB already allocated; 😊 MiB free; 😊 cached)
I tried to process an image by loading each layer to GPU and then loading it back:
for m in self.children():
m.cuda()
x = m(x)
m.cpu()
torch.cuda.empty_cache()
But it doesn't seem to be very effective. I'm wondering is there any tips and tricks to train large deep learning models while using little GPU memory.
Although
import torch
torch.cuda.empty_cache()
provides a good alternative for clearing the occupied cuda memory and we can also manually clear the not in use variables by using,
import gc
del variables
gc.collect()
But still after using these commands, the error might appear again because pytorch doesn't actually clears the memory instead clears the reference to the memory occupied by the variables.
So reducing the batch_size after restarting the kernel and finding the optimum batch_size is the best possible option (but sometimes not a very feasible one).
Another way to get a deeper insight into the alloaction of memory in gpu is to use:
torch.cuda.memory_summary(device=None, abbreviated=False)
wherein, both the arguments are optional. This gives a readable summary of memory allocation and allows you to figure the reason of CUDA running out of memory and restart the kernel to avoid the error from happening again (Just like I did in my case).
Passing the data iteratively might help but changing the size of layers of your network or breaking them down would also prove effective (as sometimes the model also occupies a significant memory for example, while doing transfer learning).
Just reduce the batch size, and it will work.
While I was training, it gave following error:
CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 10.76 GiB
total capacity; 4.29 GiB already allocated; 10.12 MiB free; 4.46 GiB
reserved in total by PyTorch)
And I was using batch size of 32. So I just changed it to 15 and it worked for me.
Send the batches to CUDA iteratively, and make small batch sizes. Don't send all your data to CUDA at once in the beginning. Rather, do it as follows:
for e in range(epochs):
for images, labels in train_loader:
if torch.cuda.is_available():
images, labels = images.cuda(), labels.cuda()
# blablabla
You can also use dtypes that use less memory. For instance, torch.float16 or torch.half.
Try not drag your grads too far.
I got the same error when I tried to sum up loss in all batches.
loss = self.criterion(pred, label)
total_loss += loss
Then I use loss.item instead of loss which requires grads, then solved the problem
loss = self.criterion(pred, label)
total_loss += loss.item()
The solution below is credited to yuval reina in the kaggle question
This error is related to the GPU memory and not the general memory => #cjinny comment might not work.
Do you use TensorFlow/Keras or Pytorch?
Try using a smaller batch size.
If you use Keras, Try to decrease some of the hidden layer sizes.
If you use Pytorch:
do you keep all the training data on the GPU all the time?
make sure you don't drag the grads too far
check the sizes of you hidden layer
Most things are covered, still will add a little.
If torch gives error as "Tried to allocate 2 MiB" etc. it is a mis-leading message. Actually, CUDA runs out of total memory required to train the model. You can reduce the batch size. Say, even if batch size of 1 is not working (happens when you train NLP models with massive sequences), try to pass lesser data, this will help you confirm that your GPU does not have enough memory to train the model.
Also, Garbage collection and cleaning cache part has to be done again, if you want to re-train the model.
Follow these steps:
Reduce train,val,test data
Reduce batch size {eg. 16 or 32}
Reduce number of model parameters {eg. less than million}
In my case, when I am training common voice dataset in kaggle kernels the same error raises. I delt with reducing training dataset to 20000,batch size to 16 and model parameter to 112K.
If you are done training and just want to test with an image, make sure to add a with torch.no_grad() and m.eval() at the beginning:
with torch.no_grad():
for m in self.children():
m.cuda()
m.eval()
x = m(x)
m.cpu()
torch.cuda.empty_cache()
This may seem obvious but it worked on my case. I was trying to use BERT to transform sentences into an embbeding representation. As BERT is a pre-trained model I didn't need to save all the gradients, and they were consuming all the GPU's memory.
There are ways to avoid, but it certainly depends on your GPU memory size:
Loading the data in GPU when unpacking the data iteratively,
features, labels in batch:
features, labels = features.to(device), labels.to(device)
Using FP_16 or single precision float dtypes.
Try reducing the batch size if you ran out of memory.
Use .detach() method to remove tensors from GPU which are not needed.
If all of the above are used properly, PyTorch library is already highly optimizer and efficient.
Implementation:
Feed the image into gpu batch by batch.
Using a small batch size during training or inference.
Resize the input images with a small image size.
Technically:
Most networks are over parameterized, which means they are too large for the learning tasks. So finding an appropriate network structure can help:
a. Compact your network with techniques like model compression, network pruning and quantization.
b. Directly using a more compact network structure like mobileNetv1/2/3.
c. Network architecture search(NAS).
I have the same error but fix it by resize my images from ~600 to 100 using the lines:
import torchvision.transforms as transforms
transform = transforms.Compose([
transforms.Resize((100, 100)),
transforms.ToTensor()
])
Although this seems bizarre what I found is there are many sessions running in the background for collab even if we factory reset runtime or we close the tab. I conquered this by clicking on "Runtime" from the menu and then selecting "Manage Sessions". I terminated all the unwanted sessions and I was good to go.
I would recommend using mixed precision training with PyTorch. It can make training way faster and consume less memory.
Take a look at https://spell.ml/blog/mixed-precision-training-with-pytorch-Xuk7YBEAACAASJam.
There is now a pretty awesome library which makes this very simple: https://github.com/rentruewang/koila
pip install koila
in your code, simply wrap the input with lazy:
from koila import lazy
input = lazy(input, batch=0)
As long as you don't cross a batch size of 32, you will be fine. Just remember to refresh or restart runtime or else even if you reduce the batch size, you will encounter the same error.
I set my batch size to 16, it reduces zero gradients from occurring during my training and the model matches the true function much better. Rather than using a batch size of 4 or 8 which causes the training loss to fluctuate than
I meet the same error, and my GPU is GTX1650 with 4g video memory and 16G ram. It worked for me when I reduce the batch_size to 3.
Hope this can help you
I faced the same problem and resolved it by degrading the PyTorch version from 1.10.1 to 1.8.1 with code 11.3.
In my case, I am using GPU RTX 3060, which works only with Cuda version 11.3 or above, and when I installed Cuda 11.3, it came with PyTorch 1.10.1. So I degraded the PyTorch version, and now it is working fine.
$ pip3 install torch==1.8.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
2- You can check by reducing train batch size also.
If you are working with images, just reduce the input image shape. For example, if you are using 512x512, try 256x256. It worked for me!
Best way would be lowering down the batch size. Usually it works. Otherwise try this:
import gc
del variable #delete unnecessary variables
gc.collect()
I'm just playing around with pytorch and I'm wondering why it consumes so much memory of my GPU?
I'm using Cuda 10.0 with pythorch 1.2.0 and torchvision 0.4.0.
import torch
gpu = torch.device("cuda")
x = torch.ones(int(4e8), device=gpu)
y = torch.ones(int(1e5), device=gpu)
Running the above code I get the error:
RuntimeError: CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 2.00 GiB total capacity; 1.49 GiB already allocated; 0 bytes free; 0 bytes cached)
So, does pytorch needs ~500MB of the gpu memory as overhead? Or what is the problem here?
More information and testing done by xymeng in github could be seen in the given link
Referencing xymeng's words :
PyTorch has its own cuda kernels. From my measurement the cuda runtime allocates ~1GB memory for them. If you compile pytorch with cudnn enabled the total memory usage is 1GB + 750M + others = 2GB+
Note that this is just my speculation as there is no official documentation about this. What puzzles me is that the cuda runtime allocates much more memory than the actual code size (they are approx. linearly correlated. If I remove half of pytorch's kernels the memory usage is also reduced by half). I suspect either the kernel binaries have been compressed or they have to be post-processed by the runtime.
Seems it suits your situation.
The same model ran fine for training with batch-size=5. I reduced the batch size from 80 to 5 during training because of the same error. I am using a GPU with 11GB of memory instead of Titan X (12GB memory), the one used by the author in actual experiment.
However, now in testing, which only has batch-size=1, it is not running.
The issue is in I-frame model testing phase, the other two models have successfully produced results for testing.
Following is my testing command:
time python test.py --arch resnet152 --data-name ucf101 --representation iframe --data-root data/ucf101/mpeg4_videos --test-list data/datalists/ucf101_split1_test.txt --weights ucf101_iframe_model_iframe_model_best.pth.tar --save-scores iframe_score_file
I have used nvidia-smi to make sure nothing else is running on the GPU.
Following is the actual error message:
RuntimeError: CUDA out of memory. Tried to allocate 384.00 MiB (GPU 0; 10.92 GiB total capacity; 10.12 GiB already allocated; 245.50 MiB free; 21.69 MiB cached)
What could be the issue and how it can be fixed?
EDIT: By removing the following two lines from test.py, it starts running without an memeory issue, but it is taking ages to process:
net = torch.nn.DataParallel(net.cuda(devices[0]), device_ids=devices)
net.eval()
Yes, the above lines are for GPU based parallel processing.
Still, is there a solution to my problem?
I suggest that you may check your test code first.
You can try:
with torch.no_grad():
It will reduce memory consumption for computations that would otherwise have requires_grad=True.
Original Answer(you can try it if you have a bigger GPU):
Maybe the model itself and parameters take up a lot of memory.
You can try "batch-size=1" on your Titan X GPU which you used before and watch whether GPU memory usage is more than 11 GB. If so, the GPU you use now(11 GB memory) may not suitable for this work.
I have run this model/testing on GPU with memory upto 8GB, by adding the following flag in the testing command given in the question:
--test-crops 1