Piotr Skalski’s Post

xAI has released the weights and architecture for Grok-1. Grok-1 is a Mixture-of-Experts model with 314 billion parameters. The released model is a raw base model checkpoint from the Grok-1 pretraining phase, which means that it is not fine-tuned for any specific application. It was trained from scratch by xAI using a custom training stack. #x #ai #xai #grok #MoE

😮 Introducing the advanced math scripting engine in GenFEA: DesignPad! This powerful feature allows you to create custom design templates, link them to analysis results, and design any type of structural element according to your preferred design standards. No limitations! Boost your productivity and streamline your workflows with GenFEA! #StructuralEngineering #DesignAutomation #EngineeringInnovation #GenFEA #ai #structuraldesign

🌟 Dive into the fascinating world of image similarity estimation with my latest project: "Fuzzy Pattern Recognition System using Image Similarity"! 📸 In this project, I've harnessed the power of fuzzy logic to precisely gauge the likeness between grayscale images, overcoming the challenges posed by uncertainties and imprecise data. 💡 Key Highlights: 🔹 Leveraging NumPy, scikit-fuzzy, Pillow, and Matplotlib libraries for robust implementation. 🔹 Crafting a sophisticated fuzzy system with inputs like image disparity and edge coherence, culminating in a comprehensive similarity metric. 🔹 Unveiling the intricacies of image processing, from grayscale conversion to edge detection and feature alignment. 🔹 Employing the Mamdani method for fuzzy inference, ensuring precise computation of image similarity. Experience the project in action as it seamlessly computes and visualizes the similarity between two images, unlocking a myriad of applications in image processing, pattern recognition, and beyond. 🚀 #FuzzyLogic #PatternRecognition #ComputerVision

🚀 Google just released #PaliGemma: A New Family of Vision-Language Models 🔹 Model Types: - Pretrained (pt) models: Fine-tunable to various tasks. - Mix models: Fine-tuned for mixed tasks, ideal for general-purpose inference. - Fine-tuned (ft) models: Specialized for specific academic benchmarks. 🔹 Key Features: - Architecture: SigLIP-So400m (image encoder) + Gemma-2B (text decoder). - Combines image and text understanding, akin to CLIP and PaLI-3. - Pre-trained on image-text data; easily fine-tuned for tasks like captioning and segmentation. - Models available in three resolutions (224x224, 448x448, 896x896) and precisions (bfloat16, float16, float32). 🔹 Availability   - Released in the Hugging Face Hub model repositories.   - Includes model cards and licenses.   - Integrated with Hugging Face transformers. More: https://lnkd.in/dZ-AcWm6

🎵 Just Built an AI Composer Using Transformers! 🤖 Inspired by: https://lnkd.in/gFYENcgn Here's the tech breakdown: 🧠 Model: MusicgenForConditionalGeneration 🚀 Framework: PyTorch + 🤗 Transformers 🎛️ Architecture: T5 encoder + EnCodec 32kHz decoder 🔢 Params: 300M (small), 1.5B (large) ⚙️ Generation: Classifier-free guidance, temperature sampling 🎨 UI: Gradio for interactive demo Key features: 👉Text-to-music in 5-30 seconds 👉Custom duration and randomness control 👉Multi-instrument, multi-genre capability code link: https://lnkd.in/gP5VBE9J From "upbeat EDM" to "melancholic piano", it delivers! 💬 Happy coding! 👉 If you found this helpful, give it a like! 🔁 Share this with your network to spread the word! ♻️ Repost to share these exciting updates with your connections! #AIMusic #PyTorch #HuggingFace #DeepLearning

Introduction to TensorFlow’s Functional API Learn what the Functional API is, and how to build complex keras models using it Photo by Hunter Harritt on Unsplash TensorFlow’s Sequential API helps the user to stack layers one on top of another, easily creating linear models, where the input of each layer is always the output of the previous one. But what happens when your model needs to handle multiple inputs or outputs, shared layers, or non-linear connections? In those scenarios, TensorFlow’s Functional API allows to build more advanced, flexible, and customizable models, giving the user the power to design the complex architectures mentioned above with ease. This article will explain: What the Functional API is How to build a simple model with the Functional API What is Tensorflow’s Functional API? TensorFlow’s Functional API is a way to create models where layers are connected like a network graph, not just stacked linearly like in the Sequential API. Imagine building with LEGO bricks: while the Sequential API would just allow us to stack blocks one on top of another, the Functional API allows us to construct bridges, towers, and paths connecting different areas. Three of the main reasons why TensorFlow’s functional API allows you to create more complex models than the sequential one are: Allows to build models with multiple inputs and outputs (like models that take both an image and text as input or models that output both classification and regression values). Allows to build models with shared layers (where two paths in the model reuse the same layer). Allows to build models with branching paths (like ResNet, where layers “skip” connections). Versatility difference between the Sequential and Functional API. Image by author How the Functional API Works The three components necessary for creating a model with the Functional API: the Input, the Layers and the Model. Image by author As the above image shows, there are three fundamental components that will need to be defined each time you work with the TensorFlow’s functional API: the input, the layers, and the model. In the following code blocks, the definition of each of these components will be shown and explained. Input from tensorflow.keras.layers import Input # Define input inputs = Input(shape=(28, 28, 1)) As shown, the tf.keras.layers.Input class receives the shape of the data in the constructor, in this case (28, 28, 1), which is the shape of a 28×28 grayscale image. Layers from tensorflow.keras.layers import Dense, Flatten # Define layers flatten_layer = Flatten() hidden_layer_1 = Dense(128, activation='relu') hidden_layer_2 = Dense(64, activation='relu') output_layer = Dense(10, activation='softmax') In the Functional API, when you call a layer instance with some input, it will return the result of forwarding the received input through the layer. Because of this, layer chains are created by calling layer instances with the outputs...

This is huge. #xAI has released an open source base model #grok1 with both the weights and architecture (real #opensource ) The model is trained on 314 billion parameters, and not fine-tuned on any task, which makes it a lucrative checkpoint to fine tune on any task. It’s exciting to see more players pushing LLMs more towards the open edge. Exciting times. #llms #ai #opensourceai

I'm thrilled to share that I've successfully completed Task4 at SkillCraft Technology focused on developing a hand gesture recognition model! This model can accurately identify and classify different hand gestures from images or video data, paving the way for more intuitive human-computer interaction and gesture-based control systems. In completing the hand gesture recognition model project, I gained valuable insights into several key areas: Data Augmentation and Preprocessing: Enhanced my understanding of how to effectively augment and preprocess image data to improve model performance. CNN Architecture: Deepened my knowledge of designing and training CNNs for image classification tasks. Model Evaluation: Learned to interpret and visualize model performance using confusion matrices and classification reports. Real-Time Systems: Gained hands-on experience in deploying machine learning models for real-time applications using OpenCV. #Machinelearning #SkillCraft Here is the link to the task4: https://lnkd.in/g8xqjx4E

Completion of DeepLearningAI course "Open Source Models with Hugging Face"

"Just finished creating a seminar post for Machine Learning and Artificial Intelligence! 🤖✨ I designed placeholders for three images, added a dynamic pattern background, and worked with layers to give the design depth and structure. It's amazing how layers can bring everything together—sometimes I feel like I'm stacking up ideas like pancakes! 🥞 #GraphicDesignWorkshop #PhotoshopSkills #MachineLearning #CreativeJourney #LayerItUp"