Martin Tutek

Agreed but I think the difference is that .from_dataset has the examples in-memory, while from_tabular_file doesn't actually load them (they are lazily loaded via a genexp). Correct me if I'm...

@mariosasko @ivansmokovic @FilipBolt please take a look at #274

We can either fix training of the model (you can make it deterministic at the cost of speed) or simply not care about performance metrics (unless they are relevant) as...

> > We can either fix training of the model (you can make it deterministic at the cost of speed) or simply not care about performance metrics (unless they are...

Will be closed via #318

This is fine, but due to the new direction that torchtext is taking, I'm not sure if it's necessary. @zhangguanheng66

This is enough to fix the issue. I would suggest we follow the example of huggingface/transformers (https://github.com/huggingface/transformers/blob/master/transformers/tokenization_utils.py#L52-L68) and having something akin to these special symbols predefined as properties in the...

Sure, it's ready for merging from my side, let me know if you see anything problematic.

You are right, it won't be backwards compatible as loading previously stored models will error out due to the missing attribute. I'll add a fix later this week, I'm swamped...

Same, FYI, Ubuntu 16.10, same problem occurred on make, both in version 2.2 & 2.3, latest pull fixes it. If you want some logs, let me know.