wassname/Open-Assistant
1
0
Fork 0
mirror of https://github.com/wassname/Open-Assistant.git synced 2026-07-26 13:07:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

Added deberta configs

Browse Source
This commit is contained in:
Szymon Ożóg
2023-01-07 16:36:55 +01:00
parent d0942a3256
commit f304921bd5
4 changed files with 56 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
model/reward/instructor/configs/deberta-v2-xlarge.yml
+15
View File
@@ -0,0 +1,15 @@
model_name: microsoft/deberta-v2-xlarge
learning_rate: 1e-5
freeze_layer: 15
scheduler: cosine
gradient_checkpointing: false
gradient_accumulation_steps: 16
per_device_train_batch_size: 1
warmup_steps: 600
eval_steps: 200
save_steps: 500
max_length: 512
num_train_epochs: 2
datasets:
- webgpt
- hfsummary
model/reward/instructor/configs/deberta-v3-base.yml
+14
View File
@@ -0,0 +1,14 @@
model_name: microsoft/deberta-v3-base
learning_rate: 1e-5
scheduler: cosine
gradient_checkpointing: false
gradient_accumulation_steps: 32
per_device_train_batch_size: 2
warmup_steps: 600
eval_steps: 200
save_steps: 500
max_length: 512
num_train_epochs: 2
datasets:
- webgpt
- hfsummary
model/reward/instructor/configs/deberta-v3-large-squad2.yml
+13
View File
@@ -0,0 +1,13 @@
model_name: deepset/deberta-v3-large-squad2
learning_rate: 1e-5
gradient_checkpointing: false
gradient_accumulation_steps: 32
per_device_train_batch_size: 1
warmup_steps: 600
eval_steps: 200
save_steps: 500
max_length: 512
num_train_epochs: 2
datasets:
- webgpt
- hfsummary
model/reward/instructor/configs/deberta-v3-large.yml
+14
View File
@@ -0,0 +1,14 @@
model_name: microsoft/deberta-v3-large
learning_rate: 1e-5
scheduler: cosine
gradient_checkpointing: false
gradient_accumulation_steps: 32
per_device_train_batch_size: 1
warmup_steps: 600
eval_steps: 200
save_steps: 500
max_length: 512
num_train_epochs: 2
datasets:
- webgpt
- hfsummary