Release 0.17.0 (#2691)

- Bump versions
- Fix a few TODO comments
- A bit of clean up in test_target_paramters.py

Author: BenjaminBossan

setup.py

@@ -15,7 +15,7 @@
 from setuptools import find_packages, setup
 
 
-VERSION = "0.16.1.dev0"
+VERSION = "0.17.0"
 
 extras = {}
 extras["quality"] = [

src/peft/__init__.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.16.1.dev0"
+__version__ = "0.17.0"
 
 from .auto import (
     MODEL_TYPE_TO_PEFT_MODEL_MAPPING,

tests/test_custom_models.py

@@ -1259,7 +1259,7 @@ def forward(self, X):
 
 
 class _LinearUsingParameter(nn.Module):
-    # TODO
+    # Linear layer equivalent
     def __init__(self, in_features, out_features, bias=None):
         super().__init__()
         self.in_features = in_features
@@ -1273,7 +1273,7 @@ def forward(self, x):
 
 
 class MlpUsingParameters(nn.Module):
-    # TODO
+    # MLP that uses layers whose parameters need to be targeted with target_parameters
     def __init__(self, bias=True):
         super().__init__()
 

tests/test_gpu_examples.py

@@ -3825,7 +3825,7 @@ def setUp(self):
         # torchao breaks with fp16 and if a previous test uses fp16, transformers will set this env var, which affects
         # subsequent tests, therefore the env var needs to be cleared explicitly
         #
-        # TODO: remove this once https://github.com/huggingface/transformers/pull/37259 is merged
+        # TODO: remove this once https://github.com/huggingface/transformers/pull/39483 is merged
         os.environ.pop("ACCELERATE_MIXED_PRECISION", None)
 
     def tearDown(self):

tests/test_target_parameters.py

@@ -16,7 +16,7 @@
 import torch
 from transformers import AutoModelForCausalLM
 
-from peft import LoraConfig, get_peft_model
+from peft import LoraConfig, TaskType, get_peft_model
 
 from .testing_common import PeftCommonTester, hub_online_once
 from .testing_utils import set_init_weights_false
@@ -26,18 +26,14 @@
     "trl-internal-testing/tiny-Llama4ForCausalLM",
 ]
 
-# TODO Missing from this list are LoKr, LoHa, LN Tuning, add them
 ALL_CONFIGS = [
     # target down_proj
     (
         LoraConfig,
         {
-            "task_type": "CAUSAL_LM",
-            "r": 8,
-            "lora_alpha": 32,
-            "target_modules": None,
+            "task_type": TaskType.CAUSAL_LM,
+            "target_modules": [],
             "lora_dropout": 0.0,
-            "bias": "none",
             "target_parameters": [
                 "feed_forward.experts.down_proj",
             ],
@@ -47,12 +43,9 @@
     (
         LoraConfig,
         {
-            "task_type": "CAUSAL_LM",
-            "r": 8,
-            "lora_alpha": 32,
-            "target_modules": None,
+            "task_type": TaskType.CAUSAL_LM,
+            "target_modules": [],
             "lora_dropout": 0.0,
-            "bias": "none",
             "target_parameters": [
                 "0.feed_forward.experts.gate_up_proj",
                 "1.feed_forward.experts.down_proj",
@@ -79,12 +72,9 @@
     (
         LoraConfig,
         {
-            "task_type": "CAUSAL_LM",
-            "r": 8,
-            "lora_alpha": 32,
+            "task_type": TaskType.CAUSAL_LM,
             "target_modules": ["q_proj", "v_proj"],
             "lora_dropout": 0.0,
-            "bias": "none",
             "target_parameters": [
                 "feed_forward.experts.down_proj",
             ],