Fix the vllm deepspeedfp not found issue

quant_config.json

@@ -0,0 +1,6 @@
+{
+    "bits": 8,
+    "rounding": "nearest",
+    "mantissa_bits": 3,
+    "group_size": 512
+}