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

test

Browse Source
This commit is contained in:
sunny
2024-11-07 11:57:37 -05:00
parent d2e51406a1
commit 7ee7b4c493
1 changed files with 12 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
test2.yml
View File

@@ -1,5 +1,11 @@
base_model: JackFram/llama-68m base_model: JackFram/llama-68m
datasets:
- path: arcee-ai/distilabel-intel-orca-dpo-pairs-binarized
type: chatml.ultra
split: train
output_dir: ./outputs/lora-out
sequence_len: 1024 sequence_len: 1024
load_in_8bit: true load_in_8bit: true
adapter: lora adapter: lora
@@ -10,15 +16,16 @@ lora_target_linear: true
rl: dpo rl: dpo
dpo_use_weighting: true dpo_use_weighting: true
datasets: wandb_project: check_dpotrainer
- path: arcee-ai/distilabel-intel-orca-dpo-pairs-binarized wandb_entity: axolotl-ai
type: chatml.ultra wandb_watch:
split: train wandb_name: baseline/dpo_base/dpo_use_weighting
wandb_log_model:
num_epochs: 1 num_epochs: 1
micro_batch_size: 4 micro_batch_size: 4
gradient_accumulation_steps: 1 gradient_accumulation_steps: 1
output_dir: ./outputs/lora-out
learning_rate: 0.00001 learning_rate: 0.00001
optimizer: paged_adamw_8bit optimizer: paged_adamw_8bit
lr_scheduler: cosine lr_scheduler: cosine
@@ -31,8 +38,3 @@ gradient_checkpointing_kwargs:
#special_tokens: #special_tokens:
# pad_token: <|end_of_text|> # pad_token: <|end_of_text|>
wandb_project: check_dpotrainer
wandb_entity: axolotl-ai
wandb_watch:
wandb_name: baseline/dpo_base/dpo_use_weighting
wandb_log_model: