Helpful PyTorch Tutorials
August 25, 2026 · IntelliSensei Team
Why generic embeddings underperform on domain corpora, mining positives and hard negatives, a contrastive fine-tune with sentence-transformers, recall@k evaluation, and pgvector integration.
Read moreAugust 25, 2026 · IntelliSensei Team
What fine-tuning and retrieval each actually change, a cost/latency/accuracy/governance matrix, and three anonymized scenarios where the answer was fine-tune, RAG, and both.
Read moreAugust 25, 2026 · IntelliSensei Team
When to move from DDP to FSDP2, wrapping a 7B model with fully_shard, activation checkpointing and mixed precision, distributed checkpointing, and debugging OOMs and stragglers.
Read moreAugust 25, 2026 · IntelliSensei Team
A precision cheat-sheet by GPU generation, autocast patterns, the state of FP8 training, a torchao INT8 walkthrough and a decision table for training versus serving.
Read more