tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

revert renaming of deepspeed stage3 args that use auto (#2964) [skip ci]

Browse Source 
* Revert "fix deprecate deepspeed stage3_gather_16bit_weights_on_model_save arg…"

This reverts commit e207762928.

* don't revert the values that don't use 'auto'

---------

Co-authored-by: Wing Lian <wing@axolotl.ai>
This commit is contained in:
NanoCode012
2025-07-22 20:59:47 +07:00
committed by GitHub
parent 3a208cfd84
commit 631268a0ca
4 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
deepspeed_configs/zero3.json
View File

@@ -5,8 +5,8 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true

4
deepspeed_configs/zero3_bf16.json
View File

@@ -5,8 +5,8 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true

4
deepspeed_configs/zero3_bf16_cpuoffload_all.json
View File

@@ -15,8 +15,8 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true

4
deepspeed_configs/zero3_bf16_cpuoffload_params.json
View File

@@ -11,8 +11,8 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true