invoke-ai
[bug]: SDXL Lora no longer works: The base model quantization format is not compatible with DoRA patches
Is there an existing issue for this problem?
- [x] I have searched the existing issues
Install method
Manual
Operating system
Linux
GPU vendor
AMD (ROCm)
GPU model
RX 6700XT
GPU VRAM
12GB
Version number
5.9.0
Browser
Firefox 143.0.4
System Information
{ "version": "6.9.0", "dependencies": { "absl-py" : "2.3.1", "accelerate" : "1.10.1", "annotated-types" : "0.7.0", "anyio" : "4.11.0", "attrs" : "25.4.0", "bidict" : "0.23.1", "bitsandbytes" : "0.48.1", "blake3" : "1.0.8", "certifi" : "2022.12.7", "cffi" : "2.0.0", "charset-normalizer" : "2.1.1", "click" : "8.3.0", "coloredlogs" : "15.0.1", "compel" : "2.1.1", "contourpy" : "1.3.3", "CUDA" : "N/A", "cycler" : "0.12.1", "Deprecated" : "1.2.18", "diffusers" : "0.33.0", "dnspython" : "2.8.0", "dynamicprompts" : "0.31.0", "einops" : "0.8.1", "fastapi" : "0.118.3", "fastapi-events" : "0.12.2", "filelock" : "3.19.1", "flatbuffers" : "25.9.23", "fonttools" : "4.60.1", "fsspec" : "2025.9.0", "gguf" : "0.17.1", "h11" : "0.16.0", "hf-xet" : "1.1.10", "httptools" : "0.7.1", "huggingface-hub" : "0.35.3", "humanfriendly" : "10.0", "idna" : "3.4", "importlib_metadata" : "8.7.0", "iniconfig" : "2.1.0", "invisible-watermark" : "0.2.0", "InvokeAI" : "6.9.0", "jax" : "0.7.1", "jaxlib" : "0.7.1", "Jinja2" : "3.1.6", "kiwisolver" : "1.4.9", "lion-pytorch" : "0.2.3", "MarkupSafe" : "2.1.5", "matplotlib" : "3.10.7", "mediapipe" : "0.10.14", "ml_dtypes" : "0.5.3", "mpmath" : "1.3.0", "networkx" : "3.5", "numpy" : "1.26.3", "onnx" : "1.16.1", "onnxruntime" : "1.19.2", "opencv-contrib-python": "4.11.0.86", "opencv-python" : "4.11.0.86", "opt_einsum" : "3.4.0", "packaging" : "24.1", "pandas" : "2.2.3", "picklescan" : "0.0.31", "pillow" : "11.3.0", "pluggy" : "1.5.0", "prompt_toolkit" : "3.0.52", "protobuf" : "4.25.8", "psutil" : "7.1.0", "pycparser" : "2.23", "pydantic" : "2.12.2", "pydantic-settings" : "2.11.0", "pydantic_core" : "2.41.4", "pyparsing" : "3.2.5", "PyPatchMatch" : "1.0.2", "pytest" : "8.3.5", "python-dateutil" : "2.9.0.post0", "python-dotenv" : "1.1.1", "python-engineio" : "4.12.3", "python-multipart" : "0.0.20", "python-socketio" : "5.14.2", "pytorch-triton-rocm" : "3.3.1", "pytz" : "2025.2", "PyWavelets" : "1.9.0", "PyYAML" : "6.0.3", "regex" : "2025.9.18", "requests" : "2.28.1", "safetensors" : "0.6.2", "scipy" : "1.16.2", "semver" : "3.0.4", "sentencepiece" : "0.2.0", "setuptools" : "70.2.0", "simple-websocket" : "1.1.0", "six" : "1.17.0", "sniffio" : "1.3.1", "sounddevice" : "0.5.2", "spandrel" : "0.4.1", "starlette" : "0.48.0", "sympy" : "1.14.0", "tokenizers" : "0.22.1", "torch" : "2.7.1+rocm6.2.4", "torchsde" : "0.2.6", "torchvision" : "0.22.1+rocm6.2.4", "tqdm" : "4.66.5", "trampoline" : "0.1.2", "transformers" : "4.57.1", "typing-inspection" : "0.4.2", "typing_extensions" : "4.15.0", "tzdata" : "2025.2", "urllib3" : "1.26.13", "uvicorn" : "0.37.0", "uvloop" : "0.22.1", "watchfiles" : "1.1.1", "wcwidth" : "0.2.14", "websockets" : "15.0.1", "wheel" : "0.43.0", "wrapt" : "1.17.3", "wsproto" : "1.2.0", "zipp" : "3.23.0" }, "config": { "schema_version": "4.0.2", "legacy_models_yaml_path": null, "host": "127.0.0.1", "port": 9090, "allow_origins": [], "allow_credentials": true, "allow_methods": [""], "allow_headers": [""], "ssl_certfile": null, "ssl_keyfile": null, "log_tokenization": false, "patchmatch": true, "models_dir": "models", "convert_cache_dir": "models/.convert_cache", "download_cache_dir": "models/.download_cache", "legacy_conf_dir": "configs", "db_dir": "databases", "outputs_dir": "outputs", "custom_nodes_dir": "nodes", "style_presets_dir": "style_presets", "workflow_thumbnails_dir": "workflow_thumbnails", "log_handlers": ["console"], "log_format": "color", "log_level": "info", "log_sql": false, "log_level_network": "warning", "use_memory_db": false, "dev_reload": false, "profile_graphs": false, "profile_prefix": null, "profiles_dir": "profiles", "max_cache_ram_gb": null, "max_cache_vram_gb": null, "log_memory_usage": false, "device_working_mem_gb": 8, "enable_partial_loading": true, "keep_ram_copy_of_weights": true, "ram": null, "vram": null, "lazy_offload": true, "pytorch_cuda_alloc_conf": null, "device": "auto", "precision": "auto", "sequential_guidance": false, "attention_type": "sliced", "attention_slice_size": "balanced", "force_tiled_decode": false, "pil_compress_level": 1, "max_queue_size": 10000, "clear_queue_on_startup": false, "allow_nodes": null, "deny_nodes": null, "node_cache_size": 512, "hashing_algorithm": "blake3_single", "remote_api_tokens": null, "scan_models_on_startup": false, "unsafe_disable_picklescan": false, "allow_unknown_models": true }, "set_config_fields": [ "outputs_dir", "attention_type", "legacy_models_yaml_path", "device_working_mem_gb", "enable_partial_loading", "attention_slice_size" ] }
What happened
An SDXL LoRA that previously worked (Either on 6.6.0 or 6.7.0) is no longer able to load.
The model file can be found here https://civitai.com/models/971952/stabilizer-ilnai
Note this is a LoRA for NoobAIXL Vpred, however base models of this flavour work fine.
Using it results in this stacktrace:
[2025-10-08 08:13:30,264]::[InvokeAI]::ERROR --> Error while invoking session 106a5ee4-548c-4cab-86cd-9c1139935d0d, invocation 9260f05b-234b-4fc5-9348-95a3ea847361 (denoise_latents): The base model quantization format (likely bitsandbytes) is not compatible with DoRA patches.
[2025-10-08 08:13:30,264]::[InvokeAI]::ERROR --> Traceback (most recent call last):
File ".venv2/lib/python3.12/site-packages/invokeai/app/services/session_processor/session_processor_default.py", line 130, in run_node
output = invocation.invoke_internal(context=context, services=self._services)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/app/invocations/baseinvocation.py", line 241, in invoke_internal
output = self.invoke(context)
^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/app/invocations/denoise_latents.py", line 825, in invoke
return self._old_invoke(context)
^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File ".local/share/uv/python/cpython-3.12.10-linux-x86_64-gnu/lib/python3.12/contextlib.py", line 81, in inner
return func(*args, **kwds)
^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/app/invocations/denoise_latents.py", line 1088, in _old_invoke
result_latents = pipeline.latents_from_embeddings(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/stable_diffusion/diffusers_pipeline.py", line 397, in latents_from_embeddings
step_output = self.step(
^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/stable_diffusion/diffusers_pipeline.py", line 548, in step
uc_noise_pred, c_noise_pred = self.invokeai_diffuser.do_unet_step(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py", line 199, in do_unet_step
) = self._apply_standard_conditioning(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/stable_diffusion/diffusion/shared_invokeai_diffusion.py", line 343, in _apply_standard_conditioning
both_results = self.model_forward_callback(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/stable_diffusion/diffusers_pipeline.py", line 611, in _unet_forward
return self.unet(
^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/diffusers/models/unets/unet_2d_condition.py", line 1279, in forward
sample = upsample_block(
^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/diffusers/models/unets/unet_2d_blocks.py", line 2457, in forward
hidden_states = resnet(hidden_states, temb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/diffusers/models/resnet.py", line 369, in forward
input_tensor = self.conv_shortcut(input_tensor.contiguous())
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/model_manager/load/model_cache/torch_module_autocast/custom_modules/custom_conv2d.py", line 39, in forward
return self._autocast_forward_with_patches(input)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/model_manager/load/model_cache/torch_module_autocast/custom_modules/custom_conv2d.py", line 22, in _autocast_forward_with_patches
aggregated_param_residuals = self._aggregate_patch_parameters(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/model_manager/load/model_cache/torch_module_autocast/custom_modules/custom_module_mixin.py", line 71, in _aggregate_patch_parameters
layer_params = patch.get_parameters(orig_params, weight=patch_weight)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File ".venv2/lib/python3.12/site-packages/invokeai/backend/patches/layers/dora_layer.py", line 99, in get_parameters
raise RuntimeError(
RuntimeError: The base model quantization format (likely bitsandbytes) is not compatible with DoRA patches.
What you expected to happen
For the model to load correctly.
How to reproduce the problem
Download the lora (currently version 0.271) and use it with an SDXL model.
Additional context
As I am on AMD, I do not have bitsandbytes support, so I am unsure why the log references it.
Discord username
cubethethird
I have found the cause of this error. It only occurs when setting enable_partial_loading=true, due to the way this chunk of the dora_layer.py code works:
def get_parameters(self, orig_parameters: dict[str, torch.Tensor], weight: float) -> dict[str, torch.Tensor]:
if any(p.device.type == "meta" for p in orig_parameters.values()):
# If any of the original parameters are on the 'meta' device, we assume this is because the base model is in
# a quantization format that doesn't allow easy dequantization.
raise RuntimeError(
"The base model quantization format (likely bitsandbytes) is not compatible with DoRA patches."
)