Update README.md
Browse files
README.md
CHANGED
@@ -723,7 +723,7 @@ widget:
|
|
723 |
|
724 |
> still running some evals, etc. expect the model card to change a bit
|
725 |
|
726 |
-
\* No additional code. This model uses `position_embedding_type="relative_key"` to help
|
727 |
|
728 |
|
729 |
## this checkpoint
|
@@ -738,10 +738,10 @@ It achieves the following results on the evaluation set:
|
|
738 |
|
739 |
## GLUE benchmark
|
740 |
|
741 |
-
> WIP
|
742 |
|
743 |
|
744 |
-
Thus far, all completed in fp32 (
|
745 |
|
746 |
| GLUE Task | Accuracy | Combined Score | Pearson | Spearmanr | Matthews Correlation | Loss |
|
747 |
|-----------|----------|----------------|---------|-----------|----------------------|---------|
|
|
|
723 |
|
724 |
> still running some evals, etc. expect the model card to change a bit
|
725 |
|
726 |
+
\* No additional code. This model uses `position_embedding_type="relative_key"` to help with long ctx.
|
727 |
|
728 |
|
729 |
## this checkpoint
|
|
|
738 |
|
739 |
## GLUE benchmark
|
740 |
|
741 |
+
> WIP till this text is removed
|
742 |
|
743 |
|
744 |
+
Thus far, all completed in fp32 (_using nvidia tf32 dtype behind the scenes when supported_)
|
745 |
|
746 |
| GLUE Task | Accuracy | Combined Score | Pearson | Spearmanr | Matthews Correlation | Loss |
|
747 |
|-----------|----------|----------------|---------|-----------|----------------------|---------|
|