# I always fail when trying to save the model as a tflite file **URL:** \ **Category:** General Discussion\ **Tags:** python, tflite\ **Created:** [April 18, 2024, 3:45am UTC](https://discuss.ai.google.dev/t/i-always-fail-when-trying-to-save-the-model-as-a-tflite-file/30837 "2024-04-18T03:45:45Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![Luka\_Luka](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/luka_luka/32/16849_2.png) [@Luka\_Luka](https://discuss.ai.google.dev/u/Luka_Luka)\ **Post date:** [April 18, 2024, 3:45am UTC](https://discuss.ai.google.dev/t/i-always-fail-when-trying-to-save-the-model-as-a-tflite-file/30837/1 "2024-04-18T03:45:46Z")
* * * AttributeError Traceback (most recent call last) Cell In[20], line 3 1 # Convert the model to TensorFlow Lite format 2 converter = tf.lite.TFLiteConverter.from\_keras\_model(model) ----\> 3 tflite\_model = converter.convert() 5 # Save the TensorFlow Lite model to a file 6 with open(‘model.tflite’, ‘wb’) as f: File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\lite.py:1175, in \_export\_metrics..wrapper(self, \*args, \*\*kwargs) 1172 @functools.wraps(convert\_func) 1173 def wrapper(self, \*args, \*\*kwargs): 1174 # pylint: disable=protected-access → 1175 return self.\_convert\_and\_export\_metrics(convert\_func, \*args, \*\*kwargs) File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\lite.py:1129, in TFLiteConverterBase.\_convert\_and\_export\_metrics(self, convert\_func, \*args, \*\*kwargs) 1127 self.\_save\_conversion\_params\_metric() 1128 start\_time = time.process\_time() → 1129 result = convert\_func(self, \*args, \*\*kwargs) 1130 elapsed\_time\_ms = (time.process\_time() - start\_time) \* 1000 1131 if result: File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\lite.py:1641, in TFLiteKerasModelConverterV2.convert(self) 1637 if saved\_model\_convert\_result: 1638 return saved\_model\_convert\_result 1640 graph\_def, input\_tensors, output\_tensors, frozen\_func = ( → 1641 self.\_freeze\_keras\_model() 1642 ) 1644 graph\_def = self.\_optimize\_tf\_model( 1645 graph\_def, input\_tensors, output\_tensors, frozen\_func 1646 ) 1648 return super(TFLiteKerasModelConverterV2, self).convert( 1649 graph\_def, input\_tensors, output\_tensors 1650 ) File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\convert\_phase.py:215, in convert\_phase..actual\_decorator..wrapper(\*args, \*\*kwargs) 213 except Exception as error: 214 report\_error\_message(str(error)) → 215 raise error from None File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\convert\_phase.py:205, in convert\_phase..actual\_decorator..wrapper(\*args, \*\*kwargs) 202 @functools.wraps(func) 203 def wrapper(\*args, \*\*kwargs): 204 try: → 205 return func(\*args, \*\*kwargs) 206 except ConverterError as converter\_error: 207 if converter\_error.errors: File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\lite.py:1582, in TFLiteKerasModelConverterV2.\_freeze\_keras\_model(self) 1573 # If the model’s call is not a `tf.function`, then we need to first get its 1574 # input signature from `model_input_signature` method. We can’t directly 1575 # call `trace_model_call` because otherwise the batch dimension is set 1576 # to None. 1577 # Once we have better support for dynamic shapes, we can remove this. 1578 if not isinstance(self.\_keras\_model.call, \_def\_function.Function): 1579 # Pass `keep_original_batch_size=True` will ensure that we get an input 1580 # signature including the batch dimension specified by the user. 1581 # TODO(b/169898786): Use the Keras public API when TFLite moves out of TF → 1582 input\_signature = \_model\_input\_signature( 1583 self.\_keras\_model, keep\_original\_batch\_size=True 1584 ) 1586 # TODO(b/169898786): Use the Keras public API when TFLite moves out of TF 1587 func = \_trace\_model\_call(self.\_keras\_model, input\_signature) File ~\AppData\Local\Programs\Python\Python312\Lib\site-packages\tensorflow\lite\python\tflite\_keras\_util.py:84, in model\_input\_signature(model, keep\_original\_batch\_size) 82 input\_specs = input\_specs[0][0] 83 else: —\> 84 input\_specs = model.\_get\_save\_spec( # pylint: disable=protected-access 85 dynamic\_batch=not keep\_original\_batch\_size) 86 if input\_specs is None: 87 return None AttributeError: ‘Functional’ object has no attribute ‘\_get\_save\_spec’ ---
**Author:** ![Kiran\_Sai\_Ramineni](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/kiran_sai_ramineni/32/8859_2.png) [@Kiran\_Sai\_Ramineni](https://discuss.ai.google.dev/u/Kiran_Sai_Ramineni)\ **Post date:** [April 24, 2024, 8:23am UTC](https://discuss.ai.google.dev/t/i-always-fail-when-trying-to-save-the-model-as-a-tflite-file/30837/3 "2024-04-24T08:23:25Z")
Hi @Luka_Luka, I have tried to convert both the sequential and functional models to lite using ` tf.lite.TFLiteConverter.from_keras_model` with Tensorflow 2.15 and did not face any error. please refer to [gist](https://gist.github.com/kiransair/ca718d53b64f97f81eb504676568a222) this for a working code example. If possible could you please share the model you are trying to convert to debug the cause of the error. Thank You.