From 7166dc47c9bd764e2cd1d7c3d9e046cdc439b58c Mon Sep 17 00:00:00 2001
From: CaptainJack2491 <jayrupnakawala@gmail.com>
Date: Mon, 17 Aug 2026 16:35:28 +0100
Subject: perf(cuda): add AMP fp16 autocast with GradScaler and torch.compile
 reduce-overhead mode

---
 src/config.py | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'src/config.py')

diff --git a/src/config.py b/src/config.py
index 25dcdda..4db66b0 100644
--- a/src/config.py
+++ b/src/config.py
@@ -39,6 +39,8 @@ class Config:
     log_n_examples: int = 10
     out_dir: str = "runs"
     device: str = "auto"             # "auto" | "cuda" | "cpu" | "mps"
+    use_amp: bool = True             # FP16 mixed precision on CUDA (Turing Tensor Cores)
+    compile_model: bool = False      # torch.compile (reduce-overhead / CUDA Graphs)
 
     @property
     def vocab(self) -> int:
-- 
cgit v1.2.3