ML Engineer, LLM Engineer, take note: TorchCode
A platform with practice tasks for basic implementations in PyTorch and questions on Transformer, which are often encountered in interviews.
β Gathers in 39 structured tasks typical for #ML #interviews - implementations of operators, modules, and architectures in #PyTorch.
β Provides auto-checking, gradient checking, time measurement, and instant feedback, so that the practice more closely resembles #LeetCode for interviews.
β Built on the basis of Jupyter Notebook, while supporting one-click reset, hints, reference solutions, and progress tracking.
β Covers such frequent topics as ReLU, Softmax, LayerNorm, Attention, RoPE, Flash Attention, #LoRA, $MoE, and others.
β Supports online mode via Hugging Face Spaces, opening individual tasks in #Google #Colab, and local launch via #Docker.
π https://github.com/duoan/TorchCode
