A high-performance inference engine for AI models
Fix OSS on M5 Max (#274)
Aleksei Savin30c0846e2da4
diff --git a/crates/uzu/src/backends/metal/kernel/moe/router_topk.metal b/crates/uzu/src/backends/metal/kernel/moe/router_topk.metalindex f88fecc4c..93badc6a7 100644--- a/crates/uzu/src/backends/metal/kernel/moe/router_topk.metal+++ b/crates/uzu/src/backends/metal/kernel/moe/router_topk.metal@@ -123,10 +123,8 @@ PUBLIC KERNEL(MoeRouterTopK)(if (lid == 0) {logits_shared[winner_idx] = NEG_INF;idx_shared[winner_idx] = 0xFFFFFFFFu;- device int* out_ids = topk_ids + token_idx * k;- device ScalarT* out_probs = topk_probs + token_idx * k;- out_ids[sel] = int(winner_idx);- out_probs[sel] = static_cast<ScalarT>(winner_val);+ topk_ids[token_idx * k + sel] = int(winner_idx);+ topk_probs[token_idx * k + sel] = static_cast<ScalarT>(winner_val);}}threadgroup_barrier(mem_flags::mem_threadgroup);