Skip to content

Feat trust remote enable #62

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
codelion merged 3 commits into from
Oct 7, 2025
Merged

Feat trust remote enable #62

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ec7a4be
Update classifier.py
codelion Oct 7, 2025
3c3a542
Update classifier.py
codelion Oct 7, 2025
1aef514
Bump version to 0.1.2
codelion Oct 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion setup.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@

setup(
name="adaptive-classifier",
version="0.1.1",
version="0.1.2",
author="codelion",
author_email="[email protected]",
description="A flexible, adaptive classification system for dynamic text classification",
Expand Down
2 changes: 1 addition & 1 deletion src/adaptive_classifier/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
from .multilabel import MultiLabelAdaptiveClassifier, MultiLabelAdaptiveHead
from huggingface_hub import ModelHubMixin

__version__ = "0.1.1"
__version__ = "0.1.2"

__all__ = [
"AdaptiveClassifier",
Expand Down
33 changes: 22 additions & 11 deletions src/adaptive_classifier/classifier.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,8 @@ def __init__(
device: Optional[str] = None,
config: Optional[Dict[str, Any]] = None,
seed: int = 42, # Add seed parameter
use_onnx: Optional[Union[bool, str]] = "auto" # "auto", True, False
use_onnx: Optional[Union[bool, str]] = "auto", # "auto", True, False
trust_remote_code: bool = False
):
"""Initialize the adaptive classifier.

Expand All @@ -44,6 +45,7 @@ def __init__(
seed: Random seed for initialization
use_onnx: Whether to use ONNX Runtime ("auto", True, False).
"auto" uses ONNX on CPU, PyTorch on GPU.
trust_remote_code: Whether to trust remote code when loading models (default: False)
"""
# Set seed for initialization
torch.manual_seed(seed)
Expand All @@ -60,7 +62,8 @@ def __init__(
logger.info(f"Initializing ONNX model for {model_name}")
self.model = ORTModelForFeatureExtraction.from_pretrained(
model_name,
export=True # Auto-export to ONNX if not already in ONNX format
export=True, # Auto-export to ONNX if not already in ONNX format
trust_remote_code=trust_remote_code
)
logger.info("Successfully loaded ONNX model")
except ImportError:
Expand All @@ -69,17 +72,17 @@ def __init__(
"Install with: pip install optimum[onnxruntime]"
)
self.use_onnx = False
self.model = AutoModel.from_pretrained(model_name).to(self.device)
self.model = AutoModel.from_pretrained(model_name, trust_remote_code=trust_remote_code).to(self.device)
except Exception as e:
logger.warning(
f"Failed to load ONNX model: {e}. Falling back to PyTorch."
)
self.use_onnx = False
self.model = AutoModel.from_pretrained(model_name).to(self.device)
self.model = AutoModel.from_pretrained(model_name, trust_remote_code=trust_remote_code).to(self.device)
else:
self.model = AutoModel.from_pretrained(model_name).to(self.device)
self.model = AutoModel.from_pretrained(model_name, trust_remote_code=trust_remote_code).to(self.device)

self.tokenizer = AutoTokenizer.from_pretrained(model_name)
self.tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=trust_remote_code)

# Initialize memory system
self.embedding_dim = self.model.config.hidden_size
Expand Down Expand Up @@ -637,6 +640,7 @@ def _from_pretrained(
token: Optional[Union[str, bool]] = None,
use_onnx: Optional[Union[bool, str]] = "auto",
prefer_quantized: bool = True,
trust_remote_code: bool = False,
**kwargs
) -> "AdaptiveClassifier":
"""Load a model from the HuggingFace Hub or local directory.
Expand All @@ -653,6 +657,7 @@ def _from_pretrained(
use_onnx: Whether to use ONNX Runtime ("auto", True, False)
prefer_quantized: Use quantized ONNX model if available (default: True)
Set to False to use unquantized model for maximum accuracy
trust_remote_code: Whether to trust remote code when loading models (default: False)
**kwargs: Additional arguments passed to from_pretrained

Returns:
Expand All @@ -667,6 +672,9 @@ def _from_pretrained(
>>>
>>> # Force PyTorch (no ONNX)
>>> classifier = AdaptiveClassifier.load("adaptive-classifier/llm-router", use_onnx=False)
>>>
>>> # Load model requiring custom code
>>> classifier = AdaptiveClassifier.load("model-with-custom-code", trust_remote_code=True)
"""

# Check if model_id is a local directory
Expand Down Expand Up @@ -814,9 +822,10 @@ def _from_pretrained(

classifier.model = ORTModelForFeatureExtraction.from_pretrained(
onnx_path,
file_name=onnx_file
file_name=onnx_file,
trust_remote_code=trust_remote_code
)
classifier.tokenizer = AutoTokenizer.from_pretrained(config_dict['model_name'])
classifier.tokenizer = AutoTokenizer.from_pretrained(config_dict['model_name'], trust_remote_code=trust_remote_code)

# Initialize memory and other components
classifier.embedding_dim = classifier.model.config.hidden_size
Expand Down Expand Up @@ -852,7 +861,8 @@ def _from_pretrained(
config_dict['model_name'],
device=device,
config=config_dict.get('config', None),
use_onnx=final_use_onnx if isinstance(final_use_onnx, bool) else False
use_onnx=final_use_onnx if isinstance(final_use_onnx, bool) else False,
trust_remote_code=trust_remote_code
)

# Restore label mappings
Expand Down Expand Up @@ -1187,19 +1197,20 @@ def save(self, save_dir: str, include_onnx: bool = True, quantize_onnx: bool = T
)

@classmethod
def load(cls, save_dir: str, device: Optional[str] = None, use_onnx: Optional[Union[bool, str]] = "auto", prefer_quantized: bool = True) -> 'AdaptiveClassifier':
def load(cls, save_dir: str, device: Optional[str] = None, use_onnx: Optional[Union[bool, str]] = "auto", prefer_quantized: bool = True, trust_remote_code: bool = False) -> 'AdaptiveClassifier':
"""Legacy load method for backwards compatibility.

Args:
save_dir: Directory to load from
device: Device to load model on
use_onnx: Whether to use ONNX Runtime ("auto", True, False)
prefer_quantized: Use quantized ONNX model if available (default: True)
trust_remote_code: Whether to trust remote code when loading models (default: False)
"""
kwargs = {}
if device is not None:
kwargs['device'] = device
return cls._from_pretrained(save_dir, use_onnx=use_onnx, prefer_quantized=prefer_quantized, **kwargs)
return cls._from_pretrained(save_dir, use_onnx=use_onnx, prefer_quantized=prefer_quantized, trust_remote_code=trust_remote_code, **kwargs)

def to(self, device: str) -> 'AdaptiveClassifier':
"""Move the model to specified device.
Expand Down