{ "tokenizer_dir": "tokenizer", "tokenizer_max_length": 77, "vocab_size": 49408, "preferred_token_dtype": "int32", "fallback_token_dtype": "int64", "dtype_attempt_order": [ "int32", "int64" ], "variants": { "fp32": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true }, "int8": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true }, "android-qnn-npu": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true, "source_variant": "fp32", "profile_name": "android-qnn-npu" }, "android-cpu": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true, "source_variant": "fp32", "profile_name": "android-cpu" }, "android-gpu": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true, "source_variant": "fp32", "profile_name": "android-gpu" }, "ios-coreml": { "requested_token_dtype": "int32", "exported_input_name": "serving_default_args_0", "exported_input_dtype": "INT32", "exported_input_shape": [ 1, 77 ], "token_range": [ 267, 49407 ], "gpu_delegate_friendly": true, "source_variant": "fp32", "profile_name": "ios-coreml" } }, "notes": "Tokenizer output is external to the exported model. Token IDs are vocabulary indices and are not int8-quantized." }