github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-11 13:59:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

[shardformer]: support gpt-j, falcon, Mistral and add interleaved pipeline for bert (#5088)

Browse Source 
* [shardformer] implement policy for all GPT-J models and test

* [shardformer] support interleaved pipeline parallel for bert finetune

* [shardformer] shardformer support falcon (#4883)

* [shardformer]: fix interleaved pipeline for bert model (#5048)

* [hotfix]: disable seq parallel for gptj and falcon, and polish code (#5093)

* Add Mistral support for Shardformer (#5103)

* [shardformer] add tests to mistral (#5105)

---------

Co-authored-by: Pengtai Xu <henryxu880@gmail.com>
Co-authored-by: ppt0011 <143150326+ppt0011@users.noreply.github.com>
Co-authored-by: flybird11111 <1829166702@qq.com>
Co-authored-by: eric8607242 <e0928021388@gmail.com>
This commit is contained in:
Wenhao Chen
2023-11-28 16:54:42 +08:00
committed by GitHub
parent 126cf180bc
commit 7172459e74
39 changed files with 4007 additions and 234 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
examples/language/bert/data.py
View File

@@ -88,20 +88,24 @@ class GLUEDataBuilder:
)
def val_dataloader(self):
# TODO: drop_last is set to True for now to avoid error when using PP
# as the last batch may not be divisible by the number of microbatches
if len(self.eval_splits) == 1:
return self.plugin.prepare_dataloader(self.dataset["validation"], batch_size=self.eval_batch_size)
return self.plugin.prepare_dataloader(
self.dataset["validation"], batch_size=self.eval_batch_size, drop_last=True
)
elif len(self.eval_splits) > 1:
return [
self.plugin.prepare_dataloader(self.dataset[x], batch_size=self.eval_batch_size)
self.plugin.prepare_dataloader(self.dataset[x], batch_size=self.eval_batch_size, drop_last=True)
for x in self.eval_splits
]
def test_dataloader(self):
if len(self.eval_splits) == 1:
return self.plugin.prepare_dataloader(self.dataset["test"], batch_size=self.eval_batch_size)
return self.plugin.prepare_dataloader(self.dataset["test"], batch_size=self.eval_batch_size, drop_last=True)
elif len(self.eval_splits) > 1:
return [
self.plugin.prepare_dataloader(self.dataset[x], batch_size=self.eval_batch_size)
self.plugin.prepare_dataloader(self.dataset[x], batch_size=self.eval_batch_size, drop_last=True)
for x in self.eval_splits
]