pytorch/PiPPy issues and pull requests

#842 - Migrate `local_test_autosplit.py` from pipeline driver to compile_stage

Pull Request - State: closed - Opened by eddogola over 1 year ago - 5 comments
Labels: cla signed

#841 - mkdir -p when creating ckpt dir

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#840 - Set output_grads correctly

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#839 - place optim state_dict loading inside `load_checkpoint`

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#838 - rename reference state dicts

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#837 - setup model trainer

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#836 - Disable bert CI

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#835 - Fix generation of loss spec from output spec

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 1 comment
Labels: cla signed

#834 - Fix PiPPy README typos for inference

Pull Request - State: closed - Opened by rohan-varma over 1 year ago
Labels: cla signed

#833 - Internal lints

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#832 - recreate "non-persistent buffers not loaded" problem

Pull Request - State: closed - Opened by eddogola over 1 year ago
Labels: cla signed

#831 - mnist example update to new compile stage API

Pull Request - State: closed - Opened by eddogola over 1 year ago - 4 comments
Labels: cla signed

#830 - selective 2d api/example added for fine-grained tp/pp demo

Pull Request - State: closed - Opened by moonbucks over 1 year ago
Labels: cla signed

#829 - update docstrings

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#828 - Enable pipeline + DDP (c10d version)

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 5 comments
Labels: cla signed

#827 - test optimizer using torch.load

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#826 - save optim state dict

Pull Request - State: closed - Opened by eddogola over 1 year ago - 4 comments
Labels: cla signed

#825 - rename ckpt test

Pull Request - State: closed - Opened by eddogola over 1 year ago
Labels: cla signed

#824 - [BE] Apply ufmt to all Python files except for pippy/fx/*.py and change the check.sh to use ufmt

Pull Request - State: closed - Opened by fegin over 1 year ago
Labels: cla signed

#823 - combine index-file-saving and params-saving into one API

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#822 - restore gitignore

Pull Request - State: closed - Opened by eddogola over 1 year ago
Labels: cla signed

#821 - test TRY_SAVE

Pull Request - State: closed - Opened by wz337 over 1 year ago
Labels: cla signed

#820 - write actual weights to files on disk

Pull Request - State: closed - Opened by eddogola over 1 year ago - 2 comments
Labels: cla signed

#819 - update metadata total size at once

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#818 - add test to check index file output

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#817 - Save checkpoint size to metadata index file

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#816 - Use stable pytorch version for code quality check

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#815 - Add test for optimizer

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#814 - Save module

Pull Request - State: closed - Opened by eddogola over 1 year ago - 3 comments
Labels: cla signed

#813 - Issue with optimizer instantiation

Issue - State: open - Opened by saiajaym over 1 year ago - 2 comments

#812 - save checkpoint index file

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#811 - Add c10d implementation for backward pass too

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#810 - add generate_json_file

Pull Request - State: closed - Opened by wz337 over 1 year ago
Labels: cla signed

#809 - tp+pp and gspmd examples not running

Issue - State: closed - Opened by access2rohit over 1 year ago - 1 comment

#808 - Why does parallel pipeline require a master

Issue - State: open - Opened by lengien over 1 year ago - 1 comment

#807 - Add docstring for load_checkpoint

Pull Request - State: closed - Opened by eddogola over 1 year ago - 1 comment
Labels: cla signed

#806 - How did this error happen when i run example about resnet?

Issue - State: open - Opened by lengien over 1 year ago

#805 - Pause some HF model tests for cleaner CI signal

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#804 - print module and code together in min_gpt_tracing

Pull Request - State: closed - Opened by moonbucks over 1 year ago
Labels: cla signed

#803 - Split each layer in multiple gpu

Issue - State: open - Opened by EnricoBeltramo over 1 year ago

#802 - Request for Examples of Pipeline Parallelism with Multiple Machines in PiPPy

Issue - State: open - Opened by littlefatfat over 1 year ago - 1 comment

#801 - How to run the gpt2 example on a single node with four GPU?

Issue - State: open - Opened by lsder over 1 year ago

#800 - Add c10d backend for PiPPy

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#799 - Incorrect loss value of huggingface bert example

Issue - State: open - Opened by pinxuezhao over 1 year ago

#798 - 0.1.1

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#797 - Add GitHub URL in long description

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#796 - Could pippy be coexisted with deepspeed?

Issue - State: open - Opened by leiwen83 over 1 year ago - 1 comment

#795 - Reduce chunks when batch size is too small to divide

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#794 - Pure c10d example

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 1 comment
Labels: cla signed

#793 - Use with torch.device(meta)

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#792 - Deserialize stage creation

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#791 - Deserialize stage init

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#790 - Load only necessary checkpoint files

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#789 - adding init with empty

Pull Request - State: closed - Opened by HamidShojanazeri over 1 year ago - 2 comments
Labels: cla signed

#788 - Generalize remap_qualname support to submodules

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#787 - Add test for remap_qualname

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 1 comment
Labels: cla signed

#786 - init_empty_weights only works with torchrun and is very slow

Issue - State: closed - Opened by HamidShojanazeri over 1 year ago - 6 comments

#785 - Add long_description

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#784 - Change package name to torchpippy

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#783 - 0.1.0

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#782 - Remove ddp2pipe as some users thought it is ddp+pipe

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#781 - Support upper case model name

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#780 - fix loading opt models

Pull Request - State: closed - Opened by jiqing-feng over 1 year ago - 3 comments
Labels: cla signed

#779 - Load lm_head from decoder.embed_tokens.weight

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#778 - Add VERSION_NO_GIT env to setup.py

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#777 - fix load module

Pull Request - State: closed - Opened by jiqing-feng over 1 year ago - 2 comments
Labels: cla signed

#776 - [WIP] Add benchmarks

Pull Request - State: closed - Opened by HamidShojanazeri over 1 year ago - 1 comment
Labels: cla signed

#775 - Remove spmd from repo

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 1 comment
Labels: cla signed

#774 - Inference example update

Pull Request - State: closed - Opened by HamidShojanazeri over 1 year ago - 2 comments
Labels: cla signed

#773 - Any plan to support PEFT LoRA models?

Issue - State: open - Opened by zsc over 1 year ago - 2 comments

#772 - Support HuggingFace generate method

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#771 - TP+PiPPy failing on HF examples.

Issue - State: open - Opened by HamidShojanazeri over 1 year ago - 4 comments

#770 - Load model before assign submodule to device to save cpu memory

Pull Request - State: closed - Opened by jiqing-feng over 1 year ago - 20 comments
Labels: cla signed

#769 - update the Inference example readme

Pull Request - State: closed - Opened by HamidShojanazeri over 1 year ago
Labels: cla signed

#768 - Binary publish

Pull Request - State: closed - Opened by HamidShojanazeri over 1 year ago
Labels: cla signed

#767 - Update README to include compile() and all_compile()

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#766 - Setting the default value of PIPPY_PIN_DEVICE to 0

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#765 - [checkpoint]Add fsspec writer/reader to Tau

Pull Request - State: closed - Opened by wz337 over 1 year ago - 1 comment
Labels: cla signed

#764 - [SPMD] Fix CI test and code quality errors

Pull Request - State: closed - Opened by wz337 over 1 year ago
Labels: cla signed

#763 - [checkpoint] Add fsspec to spmd dependency

Pull Request - State: closed - Opened by wz337 over 1 year ago
Labels: cla signed

#762 - Simplify tp+pp example by using pippy.all_compile

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 1 comment
Labels: cla signed

#761 - Use cond var instead of pp_group_barrier

Pull Request - State: closed - Opened by kwen2501 over 1 year ago
Labels: cla signed

#760 - Introduce pippy.all_compile

Pull Request - State: closed - Opened by kwen2501 over 1 year ago - 4 comments
Labels: cla signed

#759 - Inference example compile update

Pull Request - State: closed - Opened by HamidShojanazeri almost 2 years ago
Labels: cla signed

#758 - Clean up USE_CUDA env from tests

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#757 - Clean up DDP + PP test

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#756 - Refactor dynamo example

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#755 - Refactor dynamo example

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#754 - Enable pre-release torch versions for Dynamo CI

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#753 - Fix #752: update PyTorch version requirement

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#752 - Problem reproducing minimal example

Issue - State: closed - Opened by clessig almost 2 years ago - 2 comments

#751 - Adopt pippy.compile in HF examples

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#750 - update AOTConfig params for functionalization of inference

Pull Request - State: closed - Opened by lessw2020 almost 2 years ago - 1 comment
Labels: cla signed

#749 - Introduce pippy.compile

Pull Request - State: closed - Opened by kwen2501 almost 2 years ago
Labels: cla signed

#748 - [SPMD] Missing DT support NotImplementedError: Operator aten.amax.default does not have a DistributedTensor rule registered.

Issue - State: open - Opened by anj-s almost 2 years ago

GitHub / pytorch/PiPPy issues and pull requests