fix profile

karpathy · Jun 25, 2024 · 2413b0a · 2413b0a
1 parent 39f976f
commit 2413b0a
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/profile_gpt2.cu b/profile_gpt2.cu
@@ -61,7  61,7 @@ int main(int argc, char *argv[]) {
     // do a training step
     gpt2_forward(&model, x, y, B, T);
     gpt2_zero_grad(&model);
-    gpt2_backward_and_reduce(&model, x, true);
     gpt2_backward_and_reduce(&model, x, 1, true);
     gpt2_update(&model, 1e-4f, 0.9f, 0.999f, 1e-8f, 0.0f, 1.f, 1, &multi_gpu_config);
     cudaCheck(cudaDeviceSynchronize()); // finish all CUDA work to get correct precise timings