From 2202c9a51bd1f9f4ed4dd495688207950226169f Mon Sep 17 00:00:00 2001
From: Andrej Karpathy <andrej.karpathy@gmail.com>
Date: Thu, 2 May 2024 18:09:27 +0000
Subject: [PATCH] add kernel 4 to docs. have to improve these docs more and
 document them better

---
 dev/cuda/classifier_fused.cu | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/dev/cuda/classifier_fused.cu b/dev/cuda/classifier_fused.cu
index f03d2ce59..df6894113 100644
--- a/dev/cuda/classifier_fused.cu
+++ b/dev/cuda/classifier_fused.cu
@@ -10,6 +10,7 @@ nvcc -O3 --use_fast_math classifier_fused.cu -o classifier_fused
 ./classifier_fused 1
 ./classifier_fused 2
 ./classifier_fused 3
+./classifier_fused 4
 */
 
 #include <stdio.h>
@@ -448,7 +449,7 @@ __global__ void fused_classifier_kernel4(float* dlogits, float* losses, float* p
     // calculate the probability needed for the loss and update (single-threaded)
     if(threadIdx.x == 0) {
         float prob = expf(logits[idx * P + ix] - sp.Offset) * sp.Scale;
-        losses[idx] = -logf(prob); 
+        losses[idx] = -logf(prob);
     }
 
     // very sensible default for dlosses is 1/(B*T), which is the uniform loss