wassname/alignment-handbook
1
0
Fork 0
mirror of https://github.com/wassname/alignment-handbook.git synced 2026-06-27 17:47:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

Apply quantization during DPO QLoRA (#115)

Browse Source 
* Add QLoRA fix

* Update script
This commit is contained in:
lewtun
2024-02-05 16:50:17 +01:00
committed by GitHub
parent d00e6f043e
commit 87cc800498
3 changed files with 11 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
recipes/zephyr-7b-beta/dpo/config_qlora.yaml
+4 -4
View File
@@ -1,12 +1,12 @@
# Model arguments
model_name_or_path: alignment-handbook/zephyr-7b-sft-qlora
torch_dtype: float16
torch_dtype: bfloat16
# LoRA arguments
use_peft: true
load_in_4bit: true
lora_r: 16
lora_alpha: 16
lora_r: 128
lora_alpha: 128
lora_dropout: 0.05
lora_target_modules:
- q_proj
@@ -32,7 +32,7 @@ beta: 0.01
do_eval: true
evaluation_strategy: steps
eval_steps: 100
gradient_accumulation_steps: 2
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false