Add int8 quantized whisper large models (#1126)

scripts/whisper/export-onnx.py

@@ -582,9 +582,6 @@ def main():
             location=decoder_external_filename + ".weights",
         )
 
-    if "large" in args.model:
-        # it causes errors for large models, so skip it.
-        return
     # Generate int8 quantization models
     # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection