fixes for dpo and orpo template loading (#1424)
Browse files
.github/workflows/tests.yml
CHANGED
|
@@ -34,7 +34,7 @@ jobs:
|
|
| 34 |
fail-fast: false
|
| 35 |
matrix:
|
| 36 |
python_version: ["3.10", "3.11"]
|
| 37 |
-
timeout-minutes:
|
| 38 |
|
| 39 |
steps:
|
| 40 |
- name: Check out repository code
|
|
|
|
| 34 |
fail-fast: false
|
| 35 |
matrix:
|
| 36 |
python_version: ["3.10", "3.11"]
|
| 37 |
+
timeout-minutes: 20
|
| 38 |
|
| 39 |
steps:
|
| 40 |
- name: Check out repository code
|
src/axolotl/prompt_strategies/dpo/__init__.py
CHANGED
|
@@ -5,4 +5,4 @@ from functools import partial
|
|
| 5 |
|
| 6 |
from ..base import load as load_base
|
| 7 |
|
| 8 |
-
load = partial(load_base,
|
|
|
|
| 5 |
|
| 6 |
from ..base import load as load_base
|
| 7 |
|
| 8 |
+
load = partial(load_base, module_base="axolotl.prompt_strategies.dpo")
|
src/axolotl/prompt_strategies/orpo/chat_template.py
CHANGED
|
@@ -36,6 +36,7 @@ def load(
|
|
| 36 |
chat_template = chat_templates(chat_template)
|
| 37 |
except ValueError:
|
| 38 |
pass
|
|
|
|
| 39 |
|
| 40 |
return ORPOTokenizingStrategy(
|
| 41 |
ORPOPrompter(chat_template, tokenizer),
|
|
|
|
| 36 |
chat_template = chat_templates(chat_template)
|
| 37 |
except ValueError:
|
| 38 |
pass
|
| 39 |
+
tokenizer.chat_template = chat_template
|
| 40 |
|
| 41 |
return ORPOTokenizingStrategy(
|
| 42 |
ORPOPrompter(chat_template, tokenizer),
|