Closed jcblaisecruz02 closed 3 years ago
Hi, I'm seeing the following stack trace in the logs:
Traceback (most recent call last):
File "/usr/local/lib/python3.6/dist-packages/ts/service.py", line 100, in predict
ret = self._entry_point(input_batch, self.context)
File "/home/model-server/tmp/models/7b8eea2d40534878b6e9830342ab8fb0/handler.py", line 145, in handle
input_data = _service.preprocess(data)
File "/home/model-server/tmp/models/7b8eea2d40534878b6e9830342ab8fb0/handler.py", line 88, in preprocess
input_data = self.sp.encode('[{}] [{}] '.format(ISO2M100[example['sourceLanguage']], ISO2M100[example['targetLanguage']]) + example['sourceText'], out_type=str)
TypeError: bytearray indices must be integers or slices, not str
Since it's a type error, I think this should be visible by running the tests at the bottom of the reference handler: https://github.com/gwenzek/flores/blob/master/dynalab/handler.py
I think you should start from the above handler and try to put your model specific code inside the Handler
class.
Ah, thanks for the catch. This handler looks very different from the template in the main Dynalab repo. I'll follow this one instead. Thanks for the quick response!
Hi!
We're submitting for the FLORES Small Track 2 and after configuring the handler to work for our model, both the local and integrated tests passed. Model upload finished without any problems as well, but when the model finished processing on dynabench, it immediately gave us a "takendown" status.
Here's some details on our submission:
model_files
in thesetup_config.json
file as suggested.Any ideas why the model failed to evaluate? Thanks in advance!