chore: upgrade transformers 4.52.3 (#6575)

Co-authored-by: Mick <mickjagger19@icloud.com>

python/sglang/test/runners.py

@@ -26,6 +26,7 @@ from transformers import (
     AutoModelForCausalLM,
     AutoModelForVision2Seq,
     AutoProcessor,
+    GenerationConfig,
 )
 
 from sglang.srt.entrypoints.engine import Engine
@@ -382,13 +383,17 @@ class HFRunner:
                 model = base_model
 
             outputs = model.generate(
-                input_ids,
-                do_sample=False,
-                temperature=None,
-                top_p=None,
-                max_new_tokens=max_new_tokens,
-                return_dict_in_generate=True,
-                output_scores=(not output_str_only),
+                input_ids=input_ids,
+                generation_config=GenerationConfig(
+                    do_sample=False,
+                    temperature=None,
+                    top_p=None,
+                    max_new_tokens=max_new_tokens,
+                    return_dict_in_generate=True,
+                    output_scores=(not output_str_only),
+                    # make sure to disable compile
+                    disable_compile=True,
+                ),
             )
 
             text = tokenizer.decode(