-
**Issue by [W4ngatang](https://github.com/W4ngatang)**
_Monday Jun 24, 2019 at 15:06 GMT_
_Originally opened as https://github.com/nyu-mll/jiant/pull/731_
----
Implements ReCoRD task.
We model t…
-
**Describe the bug**
When fine-tuning RoBERTa-Large on the MultiRC task, the following error occurs:
```
Traceback (most recent call last):
File "/home/js11133/jiant/jiant/proj/main/runscript.…
-
When starting a new experiment it's common to launch several runs at once that need the same data. This will create a race as the jobs all try to write and read the same preprocessed data at once, and…
-
**Issue by [pruksmhc](https://github.com/pruksmhc)**
_Sunday Mar 08, 2020 at 13:22 GMT_
_Originally opened as https://github.com/nyu-mll/jiant/pull/1030_
----
Adding Masked Language Modeling Task f…
-
**Description**
tokenize_and_cache.py fails for MultiRC with the following error:
```
Traceback (most recent call last):
File "/home/js11133/jiant/jiant/proj/main/tokenize_and_cache.py", line 1…
-
**Issue by [sleepinyourhat](https://github.com/sleepinyourhat)**
_Wednesday Jul 17, 2019 at 19:47 GMT_
_Originally opened as https://github.com/nyu-mll/jiant/pull/845_
----
There's a lot going on h…
-
Hi, I have tried to finetune the model with the run_finetune.sh script, but the accuracy is very low.
Here is the log:
`
INFO:root:18:19:05 Namespace(accumulate=None, batch_size=8, dataset='op…
-
This issue tracks progress on adding support for exporting SuperGLUE task test predictions in leaderboard submission format (see ["How do I use SuperGLUE"](https://super.gluebenchmark.com/faq) for det…
-
Hi,
We are trying to replicate some fine-tuning the results of the paper
but don't have the resources to finetune for 2^18 steps and evaluate all checkpoints (it is often not needed).
Can you sha…
-
# GPT-3 : Language Models are few-shot Learner
## 2. Approach
- GPT2의 Pretraining 방법론과 전반적으로 동일하며, GPT3는 모델의 사이즈와 데이터의 사이즈, 다양성, 학습의 길이를 키운 것임.
- in-context learning 같은 기법들도 역시 GPT2와 유사함 (입력 데이터를…