github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-06 11:32:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

[Feature] auto-cast optimizers to distributed version (#5746)

Browse Source 
* auto-cast optimizers to distributed

* fix galore casting

* logger

---------

Co-authored-by: Edenzzzz <wtan45@wisc.edu>
This commit is contained in:
Edenzzzz
2024-05-24 17:24:16 +08:00
committed by GitHub
parent 2fc85abf43
commit 5f8c0a0ac3
13 changed files with 61 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/test_shardformer/test_model/_utils.py
View File

@@ -17,7 +17,7 @@ from colossalai.booster.plugin import HybridParallelPlugin, LowLevelZeroPlugin
from colossalai.booster.plugin.hybrid_parallel_plugin import HybridParallelModule
from colossalai.checkpoint_io.utils import gather_distributed_param
from colossalai.lazy import LazyInitContext
from colossalai.nn.optimizer import DistGaloreAwamW
from colossalai.nn.optimizer import GaLoreAdamW8bit
from colossalai.nn.optimizer.galore import get_galore_param_groups
from colossalai.pipeline.stage_manager import PipelineStageManager
from colossalai.shardformer import ShardConfig, ShardFormer
@@ -130,7 +130,7 @@ def build_model_from_hybrid_plugin(
if use_lazy_init:
ctx.materialize(org_model)
org_model = org_model.cuda()
if sharded_optim_class == DistGaloreAwamW:
if optim_class == GaLoreAdamW8bit:
# Disable clipping and block-wise quantization
org_optimizer = optim_class(
get_galore_param_groups(org_model, weight_decay=0, rank=4),