fix: allow LoRA prompts without explicit scale

[andrewwhitecdw]

This PR addresses the following issue in scripts/trt.py: allow LoRA prompts without explicit scale.

Changes

• scripts/trt.py: allow LoRA prompts without explicit scale.

Details

--- a/scripts/trt.py+++ b/scripts/trt.py@@ -1,7 +1,11 @@- # Get pathes- print("Apllying LoRAs: " + str(loras))- available = modelmanager.available_loras()- for lora in loras:- lora_name, lora_scale = lora.split(":")[1:]- lora_scales.append(float(lora_scale))- if lora_name not in available:+ # Get pathes+ print("Apllying LoRAs: " + str(loras))+ available = modelmanager.available_loras()+ for lora in loras:+ lora_parts = lora.split(":")[1:]+ if len(lora_parts) < 1:+ raise ValueError(f"Invalid LoRA tag: <{lora}>")+ lora_name = lora_parts[0]+ lora_scale = float(lora_parts[1]) if len(lora_parts) > 1 else 1.0+ lora_scales.append(lora_scale)+ if lora_name not in available:

Tests

• tests/test_trt_lora.py

--- /dev/null+++ b/tests/test_trt_lora.py@@ -0,0 +1,56 @@+import unittest+from unittest.mock import MagicMock, patch++from scripts.trt import TensorRTScript+++class TestGetLoras(unittest.TestCase):+ def setUp(self):+ self.script = TensorRTScript()+ self.script.lora_hash = ""+ self.script.update_lora = False+ self.script.lora_refit_dict = {}++ def _prompt(self, text, model="model"):+ p = MagicMock()+ p.prompt = text+ p.sd_model_name = model+ return p++ @patch("scripts.trt.apply_loras")+ @patch("scripts.trt.modelmanager")+ def test_default_scale_for_lora_without_weight(self, modelmanager, apply_loras):+ modelmanager.available_loras.return_value = {"foo": "/path/foo.safetensors"}+ modelmanager.get_onnx_path.return_value = ("model", "/path/model.onnx")+ apply_loras.return_value = {}++ p = self._prompt("<lora:foo>")+ self.script.get_loras(p)++ self.assertEqual(p.prompt, "")+ apply_loras.assert_called_once_with(+ "/path/model.onnx",+ ["/path/foo.safetensors"],+ [1.0],+ )++ @patch("scripts.trt.apply_loras")+ @patch("scripts.trt.modelmanager")+ def test_explicit_scale_preserved(self, modelmanager, apply_loras):+ modelmanager.available_loras.return_value = {"bar": "/path/bar.safetensors"}+ modelmanager.get_onnx_path.return_value = ("model", "/path/model.onnx")+ apply_loras.return_value = {}++ p = self._prompt("<lora:bar:0.75>")+ self.script.get_loras(p)++ apply_loras.assert_called_once_with(+ "/path/model.onnx",+ ["/path/bar.safetensors"],+ [0.75],+ )+++if __name__ == "__main__":+ unittest.main()