Revert bert-base.yaml benchmark to use uncompressed streaming datasets (

#1306)
mosaicml · Jul 21, 2022 · 8418a67 · 8418a67
1 parent 38ba394
commit 8418a67
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/composer/yamls/models/bert-base.yaml b/composer/yamls/models/bert-base.yaml
@@ -8,7 +8,7 @@ model:
 # Train the model on the English C4 corpus
 train_dataset:
   streaming_c4:
-    remote: s3://allenai-c4/mds/1-gz/
+    remote: s3://allenai-c4/mds/1/
     local: /tmp/mds-cache/mds-c4/
     split: train
     shuffle: true
@@ -32,7 +32,7 @@ evaluators:
     label: bert_pre_training
     eval_dataset:
       streaming_c4:
-        remote: s3://allenai-c4/mds/1-gz/
+        remote: s3://allenai-c4/mds/1/
         local: /tmp/mds-cache/mds-c4/
         split: val
         shuffle: false