Merge branch 'main' into xla_sana

Author: SahilCarterr

.github/workflows/nightly_tests.yml

@@ -359,6 +359,8 @@ jobs:
             test_location: "bnb"
           - backend: "gguf"
             test_location: "gguf"
+          - backend: "torchao"
+            test_location: "torchao"
     runs-on:
       group: aws-g6e-xlarge-plus
     container:

docs/source/en/quantization/torchao.md

@@ -25,6 +25,7 @@ Quantize a model by passing [`TorchAoConfig`] to [`~ModelMixin.from_pretrained`]
 The example below only quantizes the weights to int8.
 
 ```python
+import torch
 from diffusers import FluxPipeline, FluxTransformer2DModel, TorchAoConfig
 
 model_id = "black-forest-labs/FLUX.1-dev"
@@ -44,6 +45,10 @@ pipe = FluxPipeline.from_pretrained(
 )
 pipe.to("cuda")
 
+# Without quantization: ~31.447 GB
+# With quantization: ~20.40 GB
+print(f"Pipeline memory usage: {torch.cuda.max_memory_reserved() / 1024**3:.3f} GB")
+
 prompt = "A cat holding a sign that says hello world"
 image = pipe(
     prompt, num_inference_steps=50, guidance_scale=4.5, max_sequence_length=512
@@ -88,6 +93,63 @@ Some quantization methods are aliases (for example, `int8wo` is the commonly use
 
 Refer to the official torchao documentation for a better understanding of the available quantization methods and the exhaustive list of configuration options available.
 
+## Serializing and Deserializing quantized models
+
+To serialize a quantized model in a given dtype, first load the model with the desired quantization dtype and then save it using the [`~ModelMixin.save_pretrained`] method.
+
+```python
+import torch
+from diffusers import FluxTransformer2DModel, TorchAoConfig
+
+quantization_config = TorchAoConfig("int8wo")
+transformer = FluxTransformer2DModel.from_pretrained(
+    "black-forest-labs/Flux.1-Dev",
+    subfolder="transformer",
+    quantization_config=quantization_config,
+    torch_dtype=torch.bfloat16,
+)
+transformer.save_pretrained("/path/to/flux_int8wo", safe_serialization=False)
+```
+
+To load a serialized quantized model, use the [`~ModelMixin.from_pretrained`] method.
+
+```python
+import torch
+from diffusers import FluxPipeline, FluxTransformer2DModel
+
+transformer = FluxTransformer2DModel.from_pretrained("/path/to/flux_int8wo", torch_dtype=torch.bfloat16, use_safetensors=False)
+pipe = FluxPipeline.from_pretrained("black-forest-labs/Flux.1-Dev", transformer=transformer, torch_dtype=torch.bfloat16)
+pipe.to("cuda")
+
+prompt = "A cat holding a sign that says hello world"
+image = pipe(prompt, num_inference_steps=30, guidance_scale=7.0).images[0]
+image.save("output.png")
+```
+
+Some quantization methods, such as `uint4wo`, cannot be loaded directly and may result in an `UnpicklingError` when trying to load the models, but work as expected when saving them. In order to work around this, one can load the state dict manually into the model. Note, however, that this requires using `weights_only=False` in `torch.load`, so it should be run only if the weights were obtained from a trustable source.
+
+```python
+import torch
+from accelerate import init_empty_weights
+from diffusers import FluxPipeline, FluxTransformer2DModel, TorchAoConfig
+
+# Serialize the model
+transformer = FluxTransformer2DModel.from_pretrained(
+    "black-forest-labs/Flux.1-Dev",
+    subfolder="transformer",
+    quantization_config=TorchAoConfig("uint4wo"),
+    torch_dtype=torch.bfloat16,
+)
+transformer.save_pretrained("/path/to/flux_uint4wo", safe_serialization=False, max_shard_size="50GB")
+# ...
+
+# Load the model
+state_dict = torch.load("/path/to/flux_uint4wo/diffusion_pytorch_model.bin", weights_only=False, map_location="cpu")
+with init_empty_weights():
+    transformer = FluxTransformer2DModel.from_config("/path/to/flux_uint4wo/config.json")
+transformer.load_state_dict(state_dict, strict=True, assign=True)
+```
+
 ## Resources
 
 - [TorchAO Quantization API](https://github.com/pytorch/ao/blob/main/torchao/quantization/README.md)

src/diffusers/loaders/lora_pipeline.py

@@ -2286,6 +2286,50 @@ def unload_lora_weights(self):
             transformer.load_state_dict(transformer._transformer_norm_layers, strict=False)
             transformer._transformer_norm_layers = None
 
+        if getattr(transformer, "_overwritten_params", None) is not None:
+            overwritten_params = transformer._overwritten_params
+            module_names = set()
+
+            for param_name in overwritten_params:
+                if param_name.endswith(".weight"):
+                    module_names.add(param_name.replace(".weight", ""))
+
+            for name, module in transformer.named_modules():
+                if isinstance(module, torch.nn.Linear) and name in module_names:
+                    module_weight = module.weight.data
+                    module_bias = module.bias.data if module.bias is not None else None
+                    bias = module_bias is not None
+
+                    parent_module_name, _, current_module_name = name.rpartition(".")
+                    parent_module = transformer.get_submodule(parent_module_name)
+
+                    current_param_weight = overwritten_params[f"{name}.weight"]
+                    in_features, out_features = current_param_weight.shape[1], current_param_weight.shape[0]
+                    with torch.device("meta"):
+                        original_module = torch.nn.Linear(
+                            in_features,
+                            out_features,
+                            bias=bias,
+                            dtype=module_weight.dtype,
+                        )
+
+                    tmp_state_dict = {"weight": current_param_weight}
+                    if module_bias is not None:
+                        tmp_state_dict.update({"bias": overwritten_params[f"{name}.bias"]})
+                    original_module.load_state_dict(tmp_state_dict, assign=True, strict=True)
+                    setattr(parent_module, current_module_name, original_module)
+
+                    del tmp_state_dict
+
+                    if current_module_name in _MODULE_NAME_TO_ATTRIBUTE_MAP_FLUX:
+                        attribute_name = _MODULE_NAME_TO_ATTRIBUTE_MAP_FLUX[current_module_name]
+                        new_value = int(current_param_weight.shape[1])
+                        old_value = getattr(transformer.config, attribute_name)
+                        setattr(transformer.config, attribute_name, new_value)
+                        logger.info(
+                            f"Set the {attribute_name} attribute of the model to {new_value} from {old_value}."
+                        )
+
     @classmethod
     def _maybe_expand_transformer_param_shape_or_error_(
         cls,
@@ -2312,6 +2356,8 @@ def _maybe_expand_transformer_param_shape_or_error_(
 
         # Expand transformer parameter shapes if they don't match lora
         has_param_with_shape_update = False
+        overwritten_params = {}
+
         is_peft_loaded = getattr(transformer, "peft_config", None) is not None
         for name, module in transformer.named_modules():
             if isinstance(module, torch.nn.Linear):
@@ -2386,6 +2432,16 @@ def _maybe_expand_transformer_param_shape_or_error_(
                             f"Set the {attribute_name} attribute of the model to {new_value} from {old_value}."
                         )
 
+                    # For `unload_lora_weights()`.
+                    # TODO: this could lead to more memory overhead if the number of overwritten params
+                    # are large. Should be revisited later and tackled through a `discard_original_layers` arg.
+                    overwritten_params[f"{current_module_name}.weight"] = module_weight
+                    if module_bias is not None:
+                        overwritten_params[f"{current_module_name}.bias"] = module_bias
+
+        if len(overwritten_params) > 0:
+            transformer._overwritten_params = overwritten_params
+
         return has_param_with_shape_update
 
     @classmethod

src/diffusers/models/modeling_utils.py

@@ -718,10 +718,9 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
             hf_quantizer = None
 
         if hf_quantizer is not None:
-            is_bnb_quantization_method = hf_quantizer.quantization_config.quant_method.value == "bitsandbytes"
-            if is_bnb_quantization_method and device_map is not None:
+            if device_map is not None:
                 raise NotImplementedError(
-                    "Currently, `device_map` is automatically inferred for quantized bitsandbytes models. Support for providing `device_map` as an input will be added in the future."
+                    "Currently, providing `device_map` is not supported for quantized models. Providing `device_map` as an input will be added in the future."
                 )
 
             hf_quantizer.validate_environment(torch_dtype=torch_dtype, from_flax=from_flax, device_map=device_map)
@@ -820,7 +819,8 @@ def from_pretrained(cls, pretrained_model_name_or_path: Optional[Union[str, os.P
                     revision=revision,
                     subfolder=subfolder or "",
                 )
-                if hf_quantizer is not None and is_bnb_quantization_method:
+                # TODO: https://github.com/huggingface/diffusers/issues/10013
+                if hf_quantizer is not None:
                     model_file = _merge_sharded_checkpoints(sharded_ckpt_cached_folder, sharded_metadata)
                     logger.info("Merged sharded checkpoints as `hf_quantizer` is not None.")
                     is_sharded = False

src/diffusers/quantizers/torchao/torchao_quantizer.py

@@ -132,7 +132,7 @@ def validate_environment(self, *args, **kwargs):
     def update_torch_dtype(self, torch_dtype):
         quant_type = self.quantization_config.quant_type
 
-        if quant_type.startswith("int"):
+        if quant_type.startswith("int") or quant_type.startswith("uint"):
             if torch_dtype is not None and torch_dtype != torch.bfloat16:
                 logger.warning(
                     f"You are trying to set torch_dtype to {torch_dtype} for int4/int8/uintx quantization, but "

tests/lora/test_lora_layers_flux.py

@@ -558,6 +558,72 @@ def test_load_regular_lora(self):
         self.assertTrue(pipe.transformer.x_embedder.weight.data.shape[1] == in_features * 2)
         self.assertFalse(np.allclose(original_output, lora_output, atol=1e-3, rtol=1e-3))
 
+    def test_lora_unload_with_parameter_expanded_shapes(self):
+        components, _, _ = self.get_dummy_components(FlowMatchEulerDiscreteScheduler)
+
+        logger = logging.get_logger("diffusers.loaders.lora_pipeline")
+        logger.setLevel(logging.DEBUG)
+
+        # Change the transformer config to mimic a real use case.
+        num_channels_without_control = 4
+        transformer = FluxTransformer2DModel.from_config(
+            components["transformer"].config, in_channels=num_channels_without_control
+        ).to(torch_device)
+        self.assertTrue(
+            transformer.config.in_channels == num_channels_without_control,
+            f"Expected {num_channels_without_control} channels in the modified transformer but has {transformer.config.in_channels=}",
+        )
+
+        # This should be initialized with a Flux pipeline variant that doesn't accept `control_image`.
+        components["transformer"] = transformer
+        pipe = FluxPipeline(**components)
+        pipe = pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+
+        _, _, inputs = self.get_dummy_inputs(with_generator=False)
+        control_image = inputs.pop("control_image")
+        original_out = pipe(**inputs, generator=torch.manual_seed(0))[0]
+
+        control_pipe = self.pipeline_class(**components)
+        out_features, in_features = control_pipe.transformer.x_embedder.weight.shape
+        rank = 4
+
+        dummy_lora_A = torch.nn.Linear(2 * in_features, rank, bias=False)
+        dummy_lora_B = torch.nn.Linear(rank, out_features, bias=False)
+        lora_state_dict = {
+            "transformer.x_embedder.lora_A.weight": dummy_lora_A.weight,
+            "transformer.x_embedder.lora_B.weight": dummy_lora_B.weight,
+        }
+        with CaptureLogger(logger) as cap_logger:
+            control_pipe.load_lora_weights(lora_state_dict, "adapter-1")
+            self.assertTrue(check_if_lora_correctly_set(pipe.transformer), "Lora not correctly set in denoiser")
+
+        inputs["control_image"] = control_image
+        lora_out = control_pipe(**inputs, generator=torch.manual_seed(0))[0]
+
+        self.assertFalse(np.allclose(original_out, lora_out, rtol=1e-4, atol=1e-4))
+        self.assertTrue(pipe.transformer.x_embedder.weight.data.shape[1] == 2 * in_features)
+        self.assertTrue(pipe.transformer.config.in_channels == 2 * in_features)
+        self.assertTrue(cap_logger.out.startswith("Expanding the nn.Linear input/output features for module"))
+
+        control_pipe.unload_lora_weights()
+        self.assertTrue(
+            control_pipe.transformer.config.in_channels == num_channels_without_control,
+            f"Expected {num_channels_without_control} channels in the modified transformer but has {control_pipe.transformer.config.in_channels=}",
+        )
+        loaded_pipe = FluxPipeline.from_pipe(control_pipe)
+        self.assertTrue(
+            loaded_pipe.transformer.config.in_channels == num_channels_without_control,
+            f"Expected {num_channels_without_control} channels in the modified transformer but has {loaded_pipe.transformer.config.in_channels=}",
+        )
+        inputs.pop("control_image")
+        unloaded_lora_out = loaded_pipe(**inputs, generator=torch.manual_seed(0))[0]
+
+        self.assertFalse(np.allclose(unloaded_lora_out, lora_out, rtol=1e-4, atol=1e-4))
+        self.assertTrue(np.allclose(unloaded_lora_out, original_out, atol=1e-4, rtol=1e-4))
+        self.assertTrue(pipe.transformer.x_embedder.weight.data.shape[1] == in_features)
+        self.assertTrue(pipe.transformer.config.in_channels == in_features)
+
     @unittest.skip("Not supported in Flux.")
     def test_simple_inference_with_text_denoiser_block_scale_for_all_dict_options(self):
         pass