Create llama4.json

inference-cache-config/llama4.json

@@ -0,0 +1,18 @@
+{
+  "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
+    {
+      "task": "text-generation",
+      "batch_size": 1,
+      "sequence_length": 4096,
+      "num_cores": 16,
+      "auto_cast_type": "bf16"
+    },
+    {
+      "task": "text-generation",
+      "batch_size": 4,
+      "sequence_length": 4096,
+      "num_cores": 16,
+      "auto_cast_type": "bf16"
+    }
+  ]
+}