Upload folder using huggingface_hub

README.md

@@ -148,8 +148,8 @@ for i in range(NUM_ITER + WARMUP):
         torch.cuda.synchronize()
         timing_kernel.append(start_kernel.elapsed_time(end_kernel))
         timing_vanilla.append(start_vanilla.elapsed_time(end_vanilla))
-        loss_kernel_list[i-warmup] = loss_kernel.detach()
-        loss_vanilla_list[i-warmup] = loss_vanilla.detach()
+        loss_kernel_list[i-WARMUP] = loss_kernel.detach()
+        loss_vanilla_list[i-WARMUP] = loss_vanilla.detach()
     zero_grad()
 
 if torch.allclose(loss_kernel, loss_vanilla):