Engineering AI agents and automation systems. Currently spending too much time thinking about orchestration, memory, and context windows.
🔥
Grilling Chicken Inasal
AI systems engineer building production agent architectures and distributed backend infrastructure.
-
Indie
- Bacolod City
- https://404.staytuned.com/
- @josephgabito
- in/joseph-gabito
Highlights
Pinned Loading
-
Single-file training script for a GP...
Single-file training script for a GPT-style decoder trained on TinyStories using GPT-2 BPE (tiktoken). Implements a modern pre-norm stack (RMSNorm + causal SDPA + GELU MLP) with weight tying, bf16 mixed precision (when available), warmup + cosine LR decay, gradient clipping, checkpointing (best + periodic), and a built-in sample generation at the end. 1"""2i16 — Train a GPT-style decoder on TinyStories (tiktoken GPT-2 BPE).34Usage:5python i16_train.py --data TinyStoriesV2-GPT4-train.txt --out checkpoints
Something went wrong, please refresh the page to try again.
If the problem persists, check the GitHub status page or contact support.
If the problem persists, check the GitHub status page or contact support.






