Use PyTorchModelHubMixin for InstructionDataGuardNet (#416)

Signed-off-by: Sarah Yurick <[email protected]>

Author: sarahyurick

nemo_curator/classifiers/aegis.py

@@ -24,9 +24,8 @@
 import torch.nn.functional as F
 from crossfit import op
 from crossfit.backend.torch.hf.model import HFModel
-from huggingface_hub import hf_hub_download
+from huggingface_hub import PyTorchModelHubMixin
 from peft import PeftModel
-from safetensors.torch import load_file
 from torch.nn import Dropout, Linear
 from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer
 
@@ -75,7 +74,7 @@ class AegisConfig:
 ]
 
 
-class InstructionDataGuardNet(torch.nn.Module):
+class InstructionDataGuardNet(torch.nn.Module, PyTorchModelHubMixin):
     def __init__(self, input_dim, dropout=0.7):
         super().__init__()
         self.input_dim = input_dim
@@ -180,12 +179,14 @@ def load_model(self, device: str = "cuda"):
             add_instruction_data_guard=self.config.add_instruction_data_guard,
         )
         if self.config.add_instruction_data_guard:
-            weights_path = hf_hub_download(
-                repo_id=self.config.instruction_data_guard_path,
-                filename="model.safetensors",
+            model.instruction_data_guard_net = (
+                model.instruction_data_guard_net.from_pretrained(
+                    self.config.instruction_data_guard_path
+                )
+            )
+            model.instruction_data_guard_net = model.instruction_data_guard_net.to(
+                device
             )
-            state_dict = load_file(weights_path)
-            model.instruction_data_guard_net.load_state_dict(state_dict)
             model.instruction_data_guard_net.eval()
 
         model = model.to(device)