tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

fix deprecate deepspeed stage3_gather_16bit_weights_on_model_save arg (#2956) [skip ci]

Browse Source 
* fix deprecate deepspeed stage3_gather_16bit_weights_on_model_save arg

* replace the rest of the migrated deepspeed params
This commit is contained in:
Wing Lian
2025-07-21 11:41:31 -04:00
committed by GitHub
parent fefb0797ee
commit e207762928
4 changed files with 20 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
deepspeed_configs/zero3.json
View File

@@ -5,11 +5,11 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 0,
"stage3_max_reuse_distance": 0,
"stage3_gather_16bit_weights_on_model_save": true
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true
},
"bf16": {
"enabled": "auto"

10
deepspeed_configs/zero3_bf16.json
View File

@@ -5,11 +5,11 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 0,
"stage3_max_reuse_distance": 0,
"stage3_gather_16bit_weights_on_model_save": true
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true
},
"bf16": {
"enabled": true

10
deepspeed_configs/zero3_bf16_cpuoffload_all.json
View File

@@ -15,11 +15,11 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 0,
"stage3_max_reuse_distance": 0,
"stage3_gather_16bit_weights_on_model_save": true
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true
},
"bf16": {
"enabled": true

10
deepspeed_configs/zero3_bf16_cpuoffload_params.json
View File

@@ -11,11 +11,11 @@
"contiguous_gradients": true,
"sub_group_size": 0,
"reduce_bucket_size": "auto",
"stage3_prefetch_bucket_size": "auto",
"stage3_param_persistence_threshold": "auto",
"stage3_max_live_parameters": 0,
"stage3_max_reuse_distance": 0,
"stage3_gather_16bit_weights_on_model_save": true
"prefetch_bucket_size": "auto",
"param_persistence_threshold": "auto",
"max_live_parameters": 0,
"max_reuse_distance": 0,
"gather_16bit_weights_on_model_save": true
},
"bf16": {
"enabled": true