wassname/pytorch-lightning
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-lightning.git synced 2026-09-09 11:32:07 +08:00
Code Issues Packages Projects Releases Wiki Activity

Allow user to select individual TPU core to train on (#1729)

Browse Source 
* added tpu_id

added tpu_id to mixins

* train on individual tpu

* parallel loader if tpu_id is None

* removed progress_bar_refresh_rate

* chlog

* replaced num_tpu_cores with tpu_cores

* set tpu_id to None if int

* changed num_tpu_cores to tpu_cores in docs

* updated docs

* updated __init__.py
removed self.tpu_id for ParallelLoader

* Update pytorch_lightning/trainer/__init__.py

* check if tpu_cores is a list

Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>

* xla device conditional

* num_tpu_cores deprecation

* removed duplicate warning

* fixed pep8 error

* Revert "removed duplicate warning"

This reverts commit 8adb0a9b

* deprecated api update

* fixed recursion error

* fixed tests

* fixed flake errors

* removed current_tpu_index

* Update CHANGELOG.md

* Update trainer.py

Co-authored-by: Jirka <jirka.borovec@seznam.cz>
Co-authored-by: Jirka Borovec <Borda@users.noreply.github.com>
Co-authored-by: William Falcon <waf2107@columbia.edu>
This commit is contained in:
Lezwon Castelino
2020-05-17 16:30:54 -04:00
committed by GitHub
co-authored by Jirka Borovec Jirka William Falcon
parent 1a797bdad5
commit 7c7e50ca47
14 changed files with 104 additions and 126 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/multi_gpu.rst
+1 -1
View File
@@ -130,7 +130,7 @@ Lightning allows multiple ways of training
- DistributedDataParallel (`distributed_backend='ddp'`) (multiple-gpus across many machines).
- DistributedDataParallel2 (`distributed_backend='ddp2'`) (dp in a machine, ddp across machines).
- Horovod (`distributed_backend='horovod'`) (multi-machine, multi-gpu, configured at runtime)
- TPUs (`num_tpu_cores=8|x`) (tpu or TPU pod)
- TPUs (`tpu_cores=8|x`) (tpu or TPU pod)
.. note:: If you request multiple GPUs without setting a mode, ddp will be automatically used.