lycoris-lora 2.1.0.dev1__tar.gz → 2.1.0.dev3__tar.gz

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lycoris_lora
-Version: 2.1.0.dev1
+Version: 2.1.0.dev3
 Summary: Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion
 Home-page: https://github.com/KohakuBlueleaf/LyCORIS
 Author: Shih-Ying Yeh(KohakuBlueLeaf), Yu-Guan Hsieh, Zhidong Gao

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/README.md

@@ -100,7 +100,8 @@ You can use this package's kohya module to run kohya's training script to train
 
   ```bash
   accelerate launch train_network.py \
-    --config_file example_configs/training_configs/kohya/loha_config.toml
+    --config_file example_configs/training_configs/kohya/loha_config.toml \
+    --dataset_config example_configs/training_configs/kohya/dataset_config.toml
   ```
 
   For your convenience, some example `toml` files for kohya LyCORIS training are provided in [example/training_configs/kohya](example_configs/training_configs/kohya).
@@ -120,6 +121,7 @@ For the moment being the outputs of HCP-Diffusion are not directly compatible wi
 You can perform conversion with [tools/batch_hcp_convert.py](tools/batch_hcp_convert.py).
 
 In the case of pivotal tuning, [tools/batch_bundle_convert.py](tools/batch_bundle_convert.py) can be further used to convert to and from bundle formats.
+Check [docs/Conversion-scripts.md](docs/Conversion-scripts.md) for more information.
 
 #### As standalone wrappers
 
@@ -151,7 +153,7 @@ forward_with_lyco = your_model(x)
 
 You can check my [HakuPhi](https://github.com/KohakuBlueleaf/HakuPhi) project to see how I utilize LyCORIS to finetune the Phi-1.5 models.
 
-#### Graphical Interfaces and Colabs (via kohya trainer)
+#### Graphical interfaces and Colabs (via kohya trainer)
 
 You can also train LyCORIS with the following graphical interfaces
 
@@ -189,7 +191,7 @@ usage: extract_locon.py [-h] [--is_v2] [--is_sdxl] [--device DEVICE] [--mode MOD
 
 ### Merge LyCORIS back to model
 
-You can merge your LyCORIS model back to your checkpoint(base model)
+You can merge your LyCORIS model back to your checkpoint (base model).
 
 ```bash
 python3 merge.py <settings> <base_model> <lycoris_model> <output>
@@ -202,6 +204,36 @@ $ python3 merge.py --help
 usage: merge.py [-h] [--is_v2] [--is_sdxl] [--device DEVICE] [--dtype DTYPE] [--weight WEIGHT] base_model lycoris_model output_name
 ```
 
+### Conversion of LoRA, LyCORIS and full models between HCP and sd-webui format
+
+This script allows you to use the LyCORIS models trained with HCP-Diffusion in sd-webui.
+
+```bash
+python3 batch_hcp_convert.py \
+--network_path /path/to/ckpts \
+--dst_dir /path/to/stable-diffusion-webui/models/Lora \
+--output_prefix something \
+--auto_scale_alpha --to_webui
+```
+
+See [docs/Conversion-scripts.md](docs/Conversion-scripts.md) for more information. 
+
+
+### Conversion from and to bundle format
+
+This script is particularly useful in the case of pivotal tuning.
+
+```bash
+python3 batch_bundle_convert.py \
+--network_path /path/to/sd-webui-ssd/models/Lora  \
+--emb_path /path/to/ckpts \
+--dst_dir /path/to/sd-webui-ssd/models/Lora/bundle \
+--to_bundle --verbose 2 
+```
+
+See [docs/Conversion-scripts.md](docs/Conversion-scripts.md) for more information. 
+
+
 ## Change Log
 
 For full log, please see [Change.md](Change.md)

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/lycoris/modules/__init__.py

@@ -18,7 +18,7 @@ def make_module(lyco_type, params, lora_name, orig_module):
             orig_module,
             1,
             down.size(0),
-            torch.tensor(alpha).item(),
+            float(alpha),
             use_cp=mid is not None,
         )
         module.lora_up.weight.data.copy_(up)
@@ -28,7 +28,7 @@ def make_module(lyco_type, params, lora_name, orig_module):
     elif lyco_type == "hada":
         w1a, w1b, w2a, w2b, t1, t2, alpha = params
         module = LohaModule(
-            lora_name, orig_module, 1, w1b.size(0), alpha.item(), use_cp=t1 is not None
+            lora_name, orig_module, 1, w1b.size(0), float(alpha), use_cp=t1 is not None
         )
         module.hada_w1_a.copy_(w1a)
         module.hada_w1_b.copy_(w1b)
@@ -74,7 +74,7 @@ def make_module(lyco_type, params, lora_name, orig_module):
             orig_module,
             1,
             lora_dim,
-            torch.tensor(alpha).item(),
+            float(alpha),
             use_cp=t2 is not None,
             decompose_both=w1 is None and w2 is None,
             factor=factor,

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/lycoris/utils/__init__.py

@@ -373,7 +373,7 @@ def extract_diff(
         return loras
 
     all_loras = {}
-    for idx, te1, te2 in enumerate(zip(base_tes, db_tes)):
+    for idx, (te1, te2) in enumerate(zip(base_tes, db_tes)):
         if len(base_tes) > 1:
             prefix = f"{LORA_PREFIX_TEXT_ENCODER}{idx+1}"
         else:
@@ -685,7 +685,7 @@ def merge(tes, unet, lyco_state_dict, scale: float = 1.0, device="cpu"):
             )
 
     for idx, te in enumerate(tes):
-        if len(tes):
+        if len(tes)>1:
             prefix = LORA_PREFIX_TEXT_ENCODER + str(idx + 1)
         else:
             prefix = LORA_PREFIX_TEXT_ENCODER

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/lycoris/wrapper.py

@@ -146,7 +146,8 @@ def create_lycoris_from_weights(multiplier, file, module, weights_sd=None, **kwa
         if lora_name in loras:
             loras[lora_name] = modules
 
-    network = LycorisNetwork(module)
+    network = LycorisNetwork(module, init_only=True)
+    network.multiplier = multiplier
     network.loras = []
 
     for lora_name, orig_modules in loras.items():
@@ -156,6 +157,9 @@ def create_lycoris_from_weights(multiplier, file, module, weights_sd=None, **kwa
         module = make_module(lyco_type, params, lora_name, orig_modules)
         if module is not None:
             network.loras.append(module)
+            network.algo_table[module.__class__.__name__] = (
+                network.algo_table.get(module.__class__.__name__, 0) + 1
+            )
 
     for lora in network.loras:
         lora.multiplier = multiplier
@@ -200,10 +204,24 @@ class LycorisNetwork(torch.nn.Module):
         network_module: str = "locon",
         norm_modules=NormModule,
         train_norm=False,
+        init_only=False,
         **kwargs,
     ) -> None:
         super().__init__()
         root_kwargs = kwargs
+        if init_only:
+            self.multiplier = 1
+            self.lora_dim = 0
+            self.alpha = 1
+            self.conv_lora_dim = 0
+            self.conv_alpha = 1
+            self.dropout = 0
+            self.rank_dropout = 0
+            self.module_dropout = 0
+            self.use_tucker = False
+            self.loras = []
+            self.algo_table = {}
+            return
         self.multiplier = multiplier
         self.lora_dim = lora_dim
 

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/lycoris_lora.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lycoris-lora
-Version: 2.1.0.dev1
+Version: 2.1.0.dev3
 Summary: Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion
 Home-page: https://github.com/KohakuBlueleaf/LyCORIS
 Author: Shih-Ying Yeh(KohakuBlueLeaf), Yu-Guan Hsieh, Zhidong Gao

{lycoris_lora-2.1.0.dev1 → lycoris_lora-2.1.0.dev3}/setup.py

@@ -4,7 +4,7 @@ from setuptools import setup, find_packages
 setup(
     name="lycoris_lora",
     packages=find_packages(),
-    version="2.1.0.dev1",
+    version="2.1.0.dev3",
     url="https://github.com/KohakuBlueleaf/LyCORIS",
     description="Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion",
     author="Shih-Ying Yeh(KohakuBlueLeaf), Yu-Guan Hsieh, Zhidong Gao",