pip install crackedai-connect

        

          ML coding problems.

          Your machine. Your GPU.
        

        

100 problems in PyTorch and JAX. Implement attention from scratch,
          reproduce techniques from papers, build transformers. Code runs locally on your hardware.
        

        

          
            Browse Problems
          
          
            Setup Guide
          
        

      

    

  



  

    

      
        

          

            

              

                

                

                

              

              rotary-position-embeddings.py
            

            

              
hard
              
              research
              
                solve →
              
            

          

          

            
import torch

def apply_rope(x: torch.Tensor, positions: torch.Tensor) -> torch.Tensor:
"""Apply Rotary Position Embeddings to input tensor."""
pass

          

        

      
      

        random problem — refresh for another
      

    

  



  

    

      

        
          
Primitives

          
softmax, relu, batch norm, adam

          
25 problems

        
        
          
Framework

          
autograd, vmap, einsum, jit

          
25 problems

        
        
          
End-to-End

          
MLP, CNN, RNN, transformer

          
25 problems

        
        
          
Research

          
RoPE, GQA, LoRA, MoE

          
25 problems

        
      

    

  


  

    

      

        
$ crackedai-connect

        
  Detected: Apple M3 Pro | macOS

        
    pytorch 2.3.0 (MPS accelerated)

        
    jax 0.4.30

        
  Listening on http://127.0.0.1:52341

        

          Your code runs on your machine. We never see it.
          Setup takes 30 seconds.