search

nyu-mll / jiant-v1-legacy

The jiant toolkit for general-purpose text understanding models
MIT License
21 stars 9 forks source link

Using different datasets fro probing tasks #1010

Open jeswan opened 4 years ago

jeswan commented 4 years ago

Issue by mahsash Friday Feb 21, 2020 at 13:09 GMT Originally opened as https://github.com/nyu-mll/jiant/issues/1010

Hi.

As I do not have access to Ontonotes dataset, I tried to reproduce results for NER probing task using a different NER dataset. I changed its format to the one you use for probing tasks and replaced the rel_path in edges-ner-ontonotes task with the address to point to where I saved my dataset (after retokenizing and populating it with labels.txt file). But when I run it I get following error: "RuntimeError: No dimension to distribute: torch.Size([768])"

This is an example from my train.json file: {"text": "The Commonwealth Bank Tennis Classic is a tennis tournament on the WTA Tour held in Bali , Indonesia .", "targets": [{"span1": [1, 5], "label": "I-MISC"}, {"span1": [11, 12], "label": "I-ORG"}, {"span1": [15, 16], "label": "I-LOC"}, {"span1": [17, 18], "label": "I-LOC"}]}

Can you help me to resolve this error?

jeswan commented 4 years ago

Comment by sleepinyourhat Friday Feb 21, 2020 at 15:34 GMT

@iftenney - Mind giving this a first look?

jeswan commented 4 years ago

Comment by mahsash Monday Feb 24, 2020 at 10:39 GMT

Hi again.

I also ran it with semeval data for relation classification task and got the same error. Can you please help me? I am in a hurry.