Update inference/bf16_cast_channel_int8.py

inference/bf16_cast_channel_int8.py

@@ -35,8 +35,21 @@ def main(bf16_path, int8_path, model_name="deepseek-ai/DeepSeek-R1"):
 
         # modify config.json and save it
         config = json.load(open(config_file))
-        # delete quantization_config
-        config.pop("quantization_config", None)
+        if "quantization_config" in config:
+            quant_config = config["quantization_config"]
+            quant_config.pop("fmt", None)
+            quant_config.pop("weight_block_size", None)
+            quant_config["quant_method"] = "w8a8_int8"
+            quant_config["group_size"] = -1
+            quant_config["activation_scheme"] = "dynamic"
+            quant_config["bits"] = 8
+        else:
+            config["quantization_config"] = {
+                "activation_scheme": "dynamic",
+                "quant_method": "w8a8_int8",
+                "group_size": -1,
+                "bits": 8
+            }
         with open(config_file, "w", encoding="utf-8") as f:
             json.dump(config, f, indent=2, ensure_ascii=False, sort_keys=True)
         print(f"config.json modified and saved to {config_file}")