Fix qwen 2.5 VL multimodal example (#1541)

SUMMARY:
Fix a failing qwen 2.5 VL example, just had incorrect dataset split
configuration.


TEST PLAN:
Runs on H100 on latest llm-compressor

Signed-off-by: Brian Dellabetta <[email protected]>

Author: brian-dellabetta

examples/multimodal_vision/qwen_2_5_vl_example.py

@@ -20,7 +20,7 @@
 
 # Oneshot arguments
 DATASET_ID = "lmms-lab/flickr30k"
-DATASET_SPLIT = {"calibration": "test[:512]"}
+DATASET_SPLIT = "test[:512]"
 NUM_CALIBRATION_SAMPLES = 512
 MAX_SEQUENCE_LENGTH = 2048