Fix RuntimeError: CUDA out of memory (PyTorch)
Fix PyTorch CUDA out-of-memory errors with profiling, batch and sequence controls, AMP, checkpointing, allocator tuning, distributed fixes, and GPU diagnostics.
pytorch
3 articles
Resolve "CUDA error: device-side assert triggered"
The traceback lies because CUDA runs async. Get the real line with CUDA_LAUNCH_BLOCKING=1 or the CPU, then fix the out-of-range index behind it.
Fix torch.cuda.is_available() Returns False
The fastest test is torch.version.cuda — if it's None you have a CPU-only wheel. PyTorch bundles its own CUDA runtime, so you only need the NVIDIA driver.
Other topics
Debugging (50)Python (33)TensorFlow & GPUs (21)AI Engineering (17)JavaScript (15)llm (11)System Design (11)Angular (10)gpu (9)Interview Prep (9)Algorithms (8)cuda (8)Engineering (7)Data Engineering (6)React (6)huggingface (4)Svelte (4)typescript (4)nlp (3)ai-agents (2)ai-architecture (2)Java (2)mcp (2)networking (2)nodejs (2)pandas (2)performance (2)rag (2)security (2)SQL (2)api-development (1)authentication (1)bitsandbytes (1)chromadb (1)computer-vision (1)context-engineering (1)databases (1)devops (1)environment (1)gcp (1)hybrid-search (1)jdbc (1)json (1)jupyter (1)keras (1)langchain (1)llm-evaluation (1)llm-observability (1)machine-learning (1)monitoring (1)nextjs (1)npm (1)openai (1)opencv (1)opentelemetry (1)owasp (1)prompt-engineering (1)prompt-injection (1)pydantic (1)requests (1)reranking (1)scalability (1)sentence-transformers (1)typeerror (1)vector-database (1)voice-ai (1)