enable qwen-1.8b (#72)

Signed-off-by: intellinjun <[email protected]>
intel · Jan 24, 2024 · ea4b713 · ea4b713
1 parent 6625539
commit ea4b713
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/neural_speed/models/qwen/qwen.h b/neural_speed/models/qwen/qwen.h
@@ -30,6 +30,8 @@ static const model_scratch qwen_mem_req(int n_layers) {
       return {2048ull * MB, 2048ull * MB, 4096ull * MB};
     case 32:
       return {1024ull * MB, 1024ull * MB, 1608ull * MB};
+    case 24:
+      return {512ull * MB, 512ull * MB, 1026ull * MB};
     default:
       MODEL_ASSERT(false);
   }