- Видео 22
- Просмотров 55 372
Uygar Kurt
Добавлен 26 июн 2023
AI researcher and engineer. I created this channel to share my technical knowledge and share some projects I do. It's still very new and small channel. Single subscription goes a looong way :)
Implement Llama 3 From Scratch - PyTorch
We're going to implement Llama 3 from scratch using PyTorch and do a sample text generation with it! This is the annotated Llama. GitHub below ↓
Want to support the channel? Hit that like button and subscribe!
Since all Llama models are built on top of each other, you can use this for your other Llama implementations as well.
GitHub Link of the Code
github.com/uygarkurt/Llama-3-PyTorch
Llama 3 is introduced in the paper: "The Llama 3 Herd of Models"
arxiv.org/abs/2407.21783
Meta Website to Download Llama 3 Weights
www.llama.com/llama-downloads/
Meta's Official Implementation
github.com/meta-llama/llama3
What should I implement next? Let me know in the comments!
00:00:00 Introduction
00:00:56 Download ...
Want to support the channel? Hit that like button and subscribe!
Since all Llama models are built on top of each other, you can use this for your other Llama implementations as well.
GitHub Link of the Code
github.com/uygarkurt/Llama-3-PyTorch
Llama 3 is introduced in the paper: "The Llama 3 Herd of Models"
arxiv.org/abs/2407.21783
Meta Website to Download Llama 3 Weights
www.llama.com/llama-downloads/
Meta's Official Implementation
github.com/meta-llama/llama3
What should I implement next? Let me know in the comments!
00:00:00 Introduction
00:00:56 Download ...
Просмотров: 1 094
Видео
Model Merging: Merge LLMs to Create Frankestein Models - Python, HuggingFace, Mergekit
Просмотров 6004 месяца назад
How do I merge multiple LLMs and create a better one out of them without any training? Or can I? This video answers these question with a quick tutorial. Enjoy. GitHub below ↓ Want to support the channel? Hit that like button and subscribe! GitHub Link of the Code github.com/uygarkurt/Model-Merge Mergekit github.com/arcee-ai/mergekit Julien Simon Explanation ruclips.net/video/cvOpX75Kz4M/видео....
Conversational Memory for LLMs Using LangChain and Huggingface - Python
Просмотров 8996 месяцев назад
In order to have chain of conversations with LLMs you have to take additional steps. In this video I'll show how to do that with open source libraries. Langchain and Huggingface. GitHub below ↓ Want to support the channel? Hit that like button and subscribe! GitHub Link of the Code github.com/uygarkurt/SFT-TinyLlama LangChain Huggingface Pipeline python.langchain.com/docs/integrations/llms/hugg...
Instruct-Tune Llama to Create ChatGPT Like Chatbots | Custom Dataset, Huggingface, SFT
Просмотров 1 тыс.7 месяцев назад
We're going instrcut-tune or SFT TinyLlama so that it'll follow our instructions and respond accordingly. It's the first step to create ChatGPT like applications or chatbots. GitHub below ↓ I'm going to use TinyLlama but you can use Llama 2 either. With any other LLM's you'll probably need to follow a similar procedure too. Want to support the channel? Hit that like button and subscribe! GitHub...
Train Diffusion Model For Image Generation | PyTorch, Diffusers, Custom Dataset
Просмотров 1,5 тыс.8 месяцев назад
I'm going to show you how to train a DDPM/Diffusion Model with your own data for image generation. We'll be using Diffusers from Huggingface and PyTorch with Python. Want to support the channel? Hit that like button and subscribe! Github below ↓ GitHub Link of the Code github.com/uygarkurt/DDPM-Image-Generation Notebook github.com/uygarkurt/DDPM-Image-Generation/blob/main/DDPM_Image_Generartion...
Implement and Train DCGAN From Scratch for Image Generation - PyTorch
Просмотров 60510 месяцев назад
We're going to implement DCGAN (Deep Convolutional Generative Adversarial Networks) and train our implementation on the CelebA dataset to generate images! Want to support the channel? Hit that like button and subscribe! Github below ↓ GitHub Link of the Code github.com/uygarkurt/DCGAN-PyTorch Notebook github.com/uygarkurt/DCGAN-PyTorch/blob/main/DCGAN.ipynb DCGAN (Deep Convolutional Generative ...
Lip-Sync With AI | Video-Retalking
Просмотров 2,9 тыс.10 месяцев назад
We're going to synchornize lips of a video and an audio using Video-Retalking. In this case I experimented as a dubbing task. GitHub below ↓ Want to support the channel? Hit that like button and subscribe! GitHub Used In Video github.com/uygarkurt/video-retalking Original GitHub github.com/OpenTalker/video-retalking 00:00 Intro 01:10 Sample Results 01:49 Usage
Transformer Based Punctuation Restoration for Turkish - Paper Presentation
Просмотров 13310 месяцев назад
My paper presentation for the "A Modern Turkish Poet: Fine-Tuned GPT-2" published at UBMK23. You can find the paper and related material below ↓ Paper ieeexplore.ieee.org/document/10286720 GitHub github.com/uygarkurt/Turkish-Punctuation-Restoration
Atatürk'ün Yapay Zeka Tarafından Üretilen Sesiyle Gençliğe Hitabe - Cumhuriyetin 100. Yılı Projesi
Просмотров 87910 месяцев назад
Hepimizin bildiği Gençliğe Hitabe'yi bir de Ulu Önder Gazi Mustafa Kemal Atatürk'ün yapay zeka ile üretilen sesiyle dinleyelim.
A Modern Turkish Poet: Fine-Tuned GPT-2 - Paper Presentation
Просмотров 14410 месяцев назад
My paper presentation for the "A Modern Turkish Poet: Fine-Tuned GPT-2" published at UBMK23. You can find the paper and the pdf version of the book below ↓ Paper ieeexplore.ieee.org/document/10286720 GitHub github.com/uygarkurt/A-Modern-Turkish-Poet-Fine-Tuned-GPT-2
ViT (Vision Transformer) - An Image Is Worth 16x16 Words (Paper Explained)
Просмотров 67611 месяцев назад
Description We're going to read and explain ViT (Vision Transformer) from the paper "An Image Is Worth 16x16 Words: Transformers for Image Recognition at Scale"! Video where I implement and train the ViT from scratch below ↓ Want to support the channel? Hit that like button and subscribe! Implement and Train ViT From Scratch ruclips.net/video/Vonyoz6Yt9c/видео.html ViT (Vision Transformer) is i...
100 Subscribers Special. Thank You!
Просмотров 62911 месяцев назад
It's been a long way. But here we are. Thank you!
Implement and Train ViT From Scratch for Image Recognition - PyTorch
Просмотров 14 тыс.11 месяцев назад
We're going to implement ViT (Vision Transformer) and train our implementation on the MNIST dataset to classify images! Video where I explain the ViT paper and GitHub below ↓ Want to support the channel? Hit that like button and subscribe! ViT (Vision Transformer) - An Image Is Worth 16x16 Words (Paper Explained) ruclips.net/video/8phM16htKbU/видео.html GitHub Link of the Code github.com/uygark...
Kaggle Competition Walkthrough [NLP] - CommonLit Evaluate Student Summaries
Просмотров 593Год назад
We're going to solve another Kaggle competition! We'll implement our solution step by step. We'll be using BERT with PyTorch and 🤗 Transformers. In the end we'll make our own submission. Hit that like button and subscribe! Solution notebook below ↓ www.kaggle.com/code/uygarkk/youtube-debertav3-autocorrect Which competition should I solve next? Let me know in the comments! 00:00 Introduction 00:...
Kaggle Competition Walkthrough [NLP] - U.S. Patent Phrase to Phrase Matching
Просмотров 356Год назад
Kaggle Competition Walkthrough [NLP] - U.S. Patent Phrase to Phrase Matching
Kaggle Competition Walkthrough [NLP] - CommonLit Readability Prize
Просмотров 465Год назад
Kaggle Competition Walkthrough [NLP] - CommonLit Readability Prize
LoRA BERT vs Non-LoRA BERT: Comparison and Implementation
Просмотров 712Год назад
LoRA BERT vs Non-LoRA BERT: Comparison and Implementation
Fine-Tune BERT For Question-Answering(Q&A) - PyTorch & HuggingFace
Просмотров 2,8 тыс.Год назад
Fine-Tune BERT For Question-Answering(Q&A) - PyTorch & HuggingFace
Implement and Train U-NET From Scratch for Image Segmentation - PyTorch
Просмотров 16 тыс.Год назад
Implement and Train U-NET From Scratch for Image Segmentation - PyTorch
Scaled Dot Product Attention Explained + Implemented
Просмотров 442Год назад
Scaled Dot Product Attention Explained Implemented
Implement BERT From Scratch - PyTorch
Просмотров 9 тыс.Год назад
Implement BERT From Scratch - PyTorch
Thanks for the tutorial! I was able to run Re-Talking once with the provided example video and audio files, but since then, my PC keeps crashing (restarting) on every trial, even when running on the example files. It always crashes on the FaceDet phase (100%), or [Step 6] Lip Synthesis, as can be seen here: ruclips.net/video/eMuM4go2iXs/видео.html. What could be the reason? How do I troubleshoot? I have a very strong PC: CPU: Intel, 13900K GPUL ASUS TUF Gaming GeForce RTX 3090 OC Edition Mother board: Gigabyte Z690 AORUS Ultra RAM: Corsair VENGEANCE 64GB (2x32GB) DDR5 DRAM 5600MHz C40 4 NVME Drives Thanks, Edo
Hey. It seems to me like a hardware problem. It may be a problem that you're not aware of. Please try to run it on cloud too.
Indeed, RAG applications behave more like "searching" for pieces of content, re-ranking them, and then inserting them into a prompt to generate answers. Additionally, when generating later answers, RAG does not go back to modify the previously generated ones. Strictly speaking, RAG cannot "read" an entire document and summarize it. LLMs are also limited by the maximum context length, usually around 8K tokens. Although some closed-source models can handle over 128K tokens, it's not feasible to run such models on local machines. LangGraph operates in a way that's closer to real reading. It reads piece by piece and builds entity-relation (ER) nodes, then searches through these ER nodes to generate answers. Currently, the closest technology to human reading behavior seems to be LangGraph. Unfortunately, transformer models can only predict based on existing data and lack the ability to infer new knowledge. Recent advancements mainly involve agents predicting and executing plans to simulate reasoning, but this is still essentially prediction, and far from achieving true AGI. Even models like OpenAI’s GPT, which are claimed to be at a PhD level, fall short. For example, when I asked an LLM to "help me create a Vim editor application," it couldn't provide a complete solution.
Let’s make llama4 before llama4 🤝
With enough gpus 🤝
As always, great job!👏🏻
@@aykutcayir64 thank you!
Data, algorithms, and computational power are the three key elements. Why hasn't anyone added more complex connection models to Transformers? We should consider increasing the algorithmic complexity of large language models (LLMs), which can be likened to the complexity of connections in the human brain. This way, we wouldn't need to endlessly increase the number of parameters, especially since the number of artificial neurons already exceeds that of human neurons. Moreover, we haven't seen designs similar to the short-term memory neuron models from the runtime period. We should aim to design a model that can, like humans, quickly read relevant articles when faced with a problem. During the reading process, it could summarize related content into short-term memory and continuously update it. Then, based on this short-term memory, the model could verify the correctness of answers, for instance, by writing code to check the answers. Wouldn't this approach allow us to make the model smaller?
It's a very good research question. Attention mechanism can be viewed like the "short-term" memory you mentioned too. I remember some articles to make NN's like human brain sinapses. However the problem is that they didn't perform that well.
@@uygarkurtai The variety of neurons in the human brain far exceeds the range of functions used in artificial neural networks. How can we expect a single model, like the transformer, to handle everything? Shouldn't we focus on designing more diverse neural functions to better reflect the complexity of the brain?
@@flashlin1 in that case we again end up with a computationally expensive model. There's such a trade-off that is difficult to overcome. You may want to check multi-models that's closest to what you mention. Combination of several models. If you're curious about mimicking the brain also check out spiking neural networks.
@@uygarkurtai Why haven't we seen much progress with Spiking Neural Networks? My ideal concept of short-term memory should function during the inference phase, not be fixed during the training phase. Specifically, as the model processes an input question or reads through a large volume of articles, it should be able to summarize and store useful and relevant information in short-term memory, and only then generate an answer based on that. Moreover, during the process of generating an answer, the model should be able to dynamically update the short-term memory. For example, if later predictions impact the earlier generated content, the model should revise the previous answers based on the new information before producing the final result. Is there any model that works like this?
@@flashlin1 we haven't seem them because usually there're points where they fall short compared to regular MLPs. To me what you mentioned seems a bit like RAG applications.
I was waiting for new video. Thanks for awesome work ❤😊
Thank you!
WOW! You are something else dude! No one provides content like you! Exceptional!
Thank you!
Awesome!
Thank you!
The Video that I have been waiting for !!! Thank you 🙏🏻
Thank you!
Woowwww awesome thanks for this ❤❤
Thank you!
I've followed your video to the letter, but I am getting this message and no video at all, please, help: /usr/local/lib/python3.10/dist-packages/torch/utils/cpp_extension.py:1965: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST']. warnings.warn( /usr/local/lib/python3.10/dist-packages/torch/utils/cpp_extension.py:1965: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST']. warnings.warn( Traceback (most recent call last): File "/content/drive/MyDrive/lip-sync/video-retalking-main/inference.py", line 16, in <module> from third_part.GPEN.gpen_face_enhancer import FaceEnhancement File "/content/drive/MyDrive/lip-sync/video-retalking-main/third_part/GPEN/gpen_face_enhancer.py", line 11, in <module> from utils.inference_utils import Laplacian_Pyramid_Blending_with_mask File "/content/drive/MyDrive/lip-sync/video-retalking-main/utils/inference_utils.py", line 5, in <module> from models import load_network, load_DNet File "/content/drive/MyDrive/lip-sync/video-retalking-main/models/__init__.py", line 2, in <module> from models.DNet import DNet File "/content/drive/MyDrive/lip-sync/video-retalking-main/models/DNet.py", line 10, in <module> from models.base_blocks import LayerNorm2d, ADAINHourglass, FineEncoder, FineDecoder File "/content/drive/MyDrive/lip-sync/video-retalking-main/models/base_blocks.py", line 9, in <module> from basicsr.archs.arch_util import default_init_weights File "/usr/local/lib/python3.10/dist-packages/basicsr/__init__.py", line 4, in <module> from .data import * File "/usr/local/lib/python3.10/dist-packages/basicsr/data/__init__.py", line 22, in <module> _dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames] File "/usr/local/lib/python3.10/dist-packages/basicsr/data/__init__.py", line 22, in <listcomp> _dataset_modules = [importlib.import_module(f'basicsr.data.{file_name}') for file_name in dataset_filenames] File "/usr/lib/python3.10/importlib/__init__.py", line 126, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "/usr/local/lib/python3.10/dist-packages/basicsr/data/realesrgan_dataset.py", line 11, in <module> from basicsr.data.degradations import circular_lowpass_kernel, random_mixed_kernels File "/usr/local/lib/python3.10/dist-packages/basicsr/data/degradations.py", line 8, in <module> from torchvision.transforms.functional_tensor import rgb_to_grayscale ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'
thanks so much
Thank you!
Do we need to add .UNSQUEEZE(1) to mask = img_mask[1].float().to(device) and why?Thanks!
@@ElminsterWindwalker unsqueeze adjusts the dimensions of the tensors. If it works we don't need to.
i've trained your model through 30 epoch, but when i use model to predict It's too bad !!! then i continue to train 10 epoch more, but it's still to bad, i don't know why, i use your dataset, i use your scripts;<
Me too!
@@ElminsterWindwalker i have fixed, his code is perfect, It was my fault
@@TinLee99 that usually happens when you use a different dataset with different image channels. In that case you have to do slight modifications in your code. Are you using a different dataset?
@@uygarkurtai i use pet dataset, i trained each through 30 epoch and i can see result, but when i modify upsample from transposedv2 to upsample of pytorch, i see my result that is better, i don't no why
Me waiting for the syntax error like: 😈🤭
Can I contact you on Gmail, please?
I like this video , so clear , and I was able to follow and do the same thing. I understook Unet thanks to you . Thank you so much . Could you do a video on DDPM ? specially conditional or unconditional DDPM using Unet ? Thanks a million.
@@Samuel-san-x9x thank you! I already have a DDPM video. Check it out here: ruclips.net/video/LGe0xhRseeg/видео.htmlsi=QSAnwVGYrL5Vdafz
@@uygarkurtai thank you.
Shouldn't x be first in x = torch.cat([x, cls_token], dim=1) ?
Hey, I'm not sure if it makes a difference. You can do like that too.
@@uygarkurtai I take it back, the cls_token has to be first.
Hallo, danke Ihnen für dieses Video . ich habe aber nicht verstanden,woher haben Sie die manual_test und manual_test_mask bekommen . bei der daten ordner haben wir nur die"test.zip",test_hq.zip",,"train.zip","train_hq.zip","train_mask.zip" ich habe schon immer error für die path des manual_mask und manual_test_mask beim die Inferance Teil .
Hey, thank you! I got them from Kaggle competition. I supposed to have given the link in the video. I supposed to be showing the competition page too. You can just download from there.
The best teacher ever!
Thank you!
which notebook did you say you are using?
Hey, it's in the video description. You can also find from the github page.
The code reads okay but RuntimeError: Given groups=1, weight of size [64, 3, 3, 3], expected input[1, 64, 512, 512] to have 3 channels, but got 64 channels instead Can't seem to figure what's going wrong!
you have a shape mismatch error. You have to change variables according to your input data.
I am working on a project and the input videos may have frames where there is no face. How do I change the face detection code to just apply audio but no lip sync at those frames instead of throwing me a error
I'm not sure if we can directly do it with this package. However you can follow a 2 step approach. Maybe detect the frames first which has faces, then apply this to those frames. Skip the other frames.
thank you!
Thank you!
Can u tell me which version of python , torch , sckit learn , and other used
Hey I didn't use a specific version. You can just use the latest one.
That's a good one! Thank you for the video!
Thank you!
Uygar hocam, selamlar bir çok diğer tutorialda olduğu gibi, görseli küçük parçalara ayırmak için nn.Conv2d kullanmışız (@14:15). ancak benim bildiğim, eğer Conv2d varsa, random initialize edilen weightler de olacaktır. dolayısyla, evet resmi küçük parçalara ayırıyoruz ama aynı zamanda convolutionların weightlerini de resimlere uygulamış oluyoruz. istersen nn.Conv2d/patcher'dan sonra oluşan küçük resimleri, orjinal resim parçalarıyla karşılaştır. farklı olduklarını göreceksin. belki de ben hata yapıyorumdur. iyi çalışmalar ve başarılar
Merhabalar, kendileri de implementationlarinda patchlere ayirmak icin conv layerlar kullaniyorlar. Nasil patchele ayirdiklarini ilk basta soylemiyorlar. Conv kullanmalarinin 2 sebebi var. 1: Conv layer kullanmak performansi arttiriyor. arxiv.org/abs/2106.14881 buna bakilabilir. 2: conv layer kullandigimizda gpu'da paralel kullanabiliyoruz. Bu da her seyin daha hizli olmasini sagliyor.
Another great video!
thank you!
At last! Missed your videos😊
Thank you!
great video
Thank you!
Hi, great video! It already helped me a lot! Thanks! I'll try to train a U-Net with my own data. Can you tell me if there is anything else I need to be aware of? Does it matter what kind of data files (jpg, png, gif or tiff) I use or do they have to be jpg files + gif files with labels for the training?
Great to hear that! You can choose any format as long as you can load it.
Why you used random seed in training, Its generating similar samples everytime we run the generate image models. When I remove the generator=torch.manual_seed(random_seed), code from the sample_image_generation() method, The performance of the model decreases.
I used random seed for the training so that the resulting model would be the same. However this resulting model can generate different images. I wouldn't be expecting the performance to drop. Generator is an optional argument. Are you sure you haven't changed anything?
@@uygarkurtai No, I just removed the generator, then it starts to generating different images at all batches but the results were dropped to some extent.
@@uygarkurtai What would happen if I remove the Random seed during training and train the model. Because I see Ramdom Seed no where in the huggingface document.
@@MeDenTec You can remove the random seed. You'll still get butterfiels. It's not a vital part.
@@uygarkurtai Thanks allot, your videos and guidance are really helpful. Keep it high
could you implement a DiT ? difussion transformer?
Hey that's a great idea! I added it to my list.
Nice work. Please, my work entails using a similar model to segment multiple parasitized cells from the uninfected cells in a microscope slide image. any hint will be appreciated, please.
Hey. There're much more up-to-date models. If you want to use segmentation on a project I suggest you check them out.
@@uygarkurtai Thanks for your timely response, it means a lot to me. Would you please recommend videos or material for me?
@@afolabiowoloye804 It's a pleasure. I found this repo. You can check it out. github.com/mrgloom/awesome-semantic-segmentation
@@uygarkurtai Thanks, a lot
@@uygarkurtai such a great repo. tnx
Please implement CLIP if you can.....
Noted.
Uygar hocam selam, ben modelinizi çalıştırdım fakat çıktıda predicted mask tamamen siyah olmaktadır. Aynı dataseti kullanıyorum. Bu datasette doğru çalıştırabilirsem başka bir datasette proje yapmama gerekiyor. sebebi ne olabilir?
merhaba, data preperation yaparken bir yeri atliyorsun yuksek ihtimal. Image channellar'i karistirdiginda bu durum oluyor cogunlukla.
Hey Uygar, Thanks a lot for the tutorial, you're like my coding sensei! I was wondering about something while coding the ViT. Why do you define hidden_dim if you're not using it later on? Or maybe you are using it and I just haven't noticed? Appreciate your help!
Thank you! Seems like I don't use it yes. I don't remember exactly why I put it in the first place. Probably make a deeper MLP or something. In this case you can skip it.
you are genius brooo 🤩🤩🤩
Thank you!
Hocam "OSError: uygarkurt/bert-restore-punctuation-turkish does not appear to have a file named tf_model.h5 but there is a file for PyTorch weights. Use `from_pt=True` to load this model from those weights." hatası alıyorum modelinizi kullanmaya çalışırken
merhaba, modeli yuklerken `from_pt=True` kullanmanız ve pytorch uyumlu fonksiyonları kullanmanız gerekli.
@@uygarkurtai from_pt = True kullandım ancak uyumlu pytorch versiyonunu kullandığımdan emin değilim. Hangi versiyonu tavsiye ediyorsunuz?
@@bilalsedef9545 PyTorch versiyonunun çok onemli oldugunu düşünmüyorum. Modeli yanlış yüklüyorsunuz yüksek ihtimal. stackoverflow.com/questions/65099455/loading-pre-trained-bert-model-error-error-no-file-named-pytorch-model-bin bunun çözeceğini düşünüyorum
i got error when training : RuntimeError: Expected memory formats of X and dY are same.
Hey, can you check which device the tensors at?
i used colab pro with V100
@@YohaneesHutagalung no not which gpu. Your tensors are at which device? coda, CPU?
i am using cuda,,
Love how you explain everything
Thank you!
Thats cool man, your coding skills and how smooth you are coding that is even scary, maybe AI is not for me xdddd. Anyways my question is here: You are using only one layer , what If i want to use multiple layers? 22:44 after encoder_layer should I add another encoder_layer_2 with different parameters?
Hey, thank you for your kind words :) You can do that. The thing is you have to experiment these kind of stuff. In AI, let's say there's this architecture that works. Why is it like that? Because it works. Adding another encoder will work? Probably. Will it improve performance? I don't know. You got to try.
acaba hangi uygulamayı kullandınız?
import torch import maxvit # from .maxvit import MaxViT, max_vit_tiny_224, max_vit_small_224, max_vit_base_224, max_vit_large_224 # Tiny model network: maxvit.MaxViT = maxvit.max_vit_tiny_224(num_classes=1000) input = torch.rand(1, 3, 224, 224) output = network(input) my purpose is to do give an input as an image (1,3,224,224) and generate output as its description for that. how should i do that, what should i add more to this code?
Hey, I have no idea about maxvit. If your input channels with models input channels and sizes match, there should be no problem. I suggest you check those out.
Thanks a lot man! This helps a lot
Thank you!
Thank you so much, a video that difficult to find on the internet again
Thank you!
I have an error its No such file or directory as i put the paths of image and mask. Can you help me solving this error?
hey. Probably you typed a wrong path to images and masks. Doulbe check please
ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor' I followed your video tutorial Operation Colab started to run, prompted this, how to solve?
Hey. Sometimes colab update their default libraries. Have you tried !pip3 install torchvision ?
@@uygarkurtai fixed by using !pip install torchvision==0.12.0, it also work in a local installation on conda when you get the same error
😊 does this new implementation include flash attention?
Hey, this is just the vanilla implementation. However I believe you can enable flash attention while loading the model. I suggest you check huggingface documentation.
You are great! Adamsın :)
Thank you! Tesekkurler :)