Support T5&Clip Text Encoder LoRA from OneTrainer

requested by #1727
and some cleanups/licenses
PS: LoRA request must give download URL to at least one LoRA

backend/patcher/lora.py

@@ -1,24 +1,13 @@
 import torch
-import time
 
 import packages_3rdparty.webui_lora_collection.lora as lora_utils_webui
 import packages_3rdparty.comfyui_lora_collection.lora as lora_utils_comfyui
 
-from tqdm import tqdm
 from backend import memory_management, utils
-from backend.args import dynamic_args
-
-
-class ForgeLoraCollection:
-    # TODO
-    pass
 
 
 extra_weight_calculators = {}
-
-lora_utils_forge = ForgeLoraCollection()
-
-lora_collection_priority = [lora_utils_forge, lora_utils_webui, lora_utils_comfyui]
+lora_collection_priority = [lora_utils_webui, lora_utils_comfyui]
 
 
 def get_function(function_name: str):
@@ -32,12 +21,31 @@ def load_lora(lora, to_load):
     return patch_dict, remaining_dict
 
 
+def inner_str(k, prefix="", suffix=""):
+    return k[len(prefix):-len(suffix)]
+
+
 def model_lora_keys_clip(model, key_map={}):
-    return get_function('model_lora_keys_clip')(model, key_map)
+    model_keys, key_maps = get_function('model_lora_keys_clip')(model, key_map)
+
+    for model_key in model_keys:
+        if model_key.endswith(".weight"):
+            if model_key.startswith("t5xxl.transformer."):
+                # Flux OneTrainer T5
+                formatted = inner_str(model_key, "t5xxl.transformer.", ".weight")
+                formatted = formatted.replace(".", "_")
+                formatted = f"lora_te2_{formatted}"
+                key_map[formatted] = model_key
+
+    return key_maps
 
 
 def model_lora_keys_unet(model, key_map={}):
-    return get_function('model_lora_keys_unet')(model, key_map)
+    model_keys, key_maps = get_function('model_lora_keys_unet')(model, key_map)
+
+    # TODO: OFT
+
+    return key_maps
 
 
 @torch.inference_mode()