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

fix eval_steps to be a sane default (#797)

Browse Source 
* fix eval_steps to be a sane default

* update docs for fractional eval_steps
This commit is contained in:
Wing Lian
2023-10-27 22:36:30 -04:00
committed by GitHub
parent 0800885e2f
commit 8b79ff0e94
23 changed files with 36 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
examples/code-llama/13b/lora.yml
View File

@@ -34,7 +34,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -54,7 +54,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed:

4
examples/code-llama/13b/qlora.yml
View File

@@ -36,7 +36,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -56,7 +56,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed:

4
examples/code-llama/34b/lora.yml
View File

@@ -34,7 +34,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -54,7 +54,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed:

4
examples/code-llama/34b/qlora.yml
View File

@@ -36,7 +36,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -56,7 +56,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed:

4
examples/code-llama/7b/lora.yml
View File

@@ -34,7 +34,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -54,7 +54,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed:

4
examples/code-llama/7b/qlora.yml
View File

@@ -36,7 +36,7 @@ wandb_log_model:
gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
num_epochs: 4
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002
@@ -56,7 +56,7 @@ xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 20
eval_steps: 0.05
save_steps:
debug:
deepspeed: