NF4 loading working... I think.

invokeai/backend/bnb.py

@@ -51,6 +51,42 @@ import torch
 #             self.SCB = SCB
 
 
+class InvokeLinear4Bit(bnb.nn.Linear4bit):
+    def _load_from_state_dict(
+        self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
+    ):
+        """This method is based on the logic in the bitsandbytes serialization unit tests for `Linear4bit`:
+        https://github.com/bitsandbytes-foundation/bitsandbytes/blob/6d714a5cce3db5bd7f577bc447becc7a92d5ccc7/tests/test_linear4bit.py#L52-L71
+
+        I'm not sure why this was not included in the original `Linear4bit` implementation.
+        """
+        # During serialization, the quant_state is stored as subkeys of "weight.". Here we extract those keys.
+        quant_state_keys = [k for k in state_dict.keys() if k.startswith(prefix + "weight.")]
+
+        if len(quant_state_keys) > 0:
+            # We are loading a quantized state dict.
+            quant_state_sd = {k: state_dict.pop(k) for k in quant_state_keys}
+            weight = state_dict.pop(prefix + "weight")
+            bias = state_dict.pop(prefix + "bias", None)
+
+            if len(state_dict) != 0:
+                raise RuntimeError(f"Unexpected keys in state_dict: {state_dict.keys()}")
+
+            self.weight = bnb.nn.Params4bit.from_prequantized(
+                data=weight, quantized_stats=quant_state_sd, device=weight.device
+            )
+            if bias is None:
+                self.bias = None
+            else:
+                self.bias = torch.nn.Parameter(bias)
+
+        else:
+            # We are loading a non-quantized state dict.
+            return super()._load_from_state_dict(
+                state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
+            )
+
+
 class Invoke2Linear8bitLt(torch.nn.Linear):
     """This class is the base module for the [LLM.int8()](https://arxiv.org/abs/2208.07339) algorithm."""
 
@@ -509,7 +545,7 @@ def _convert_linear_layers_to_nf4(
         fullname = f"{prefix}.{name}" if prefix else name
         if isinstance(child, torch.nn.Linear) and not any(fullname.startswith(s) for s in ignore_modules):
             has_bias = child.bias is not None
-            replacement = bnb.nn.Linear4bit(
+            replacement = InvokeLinear4Bit(
                 child.in_features,
                 child.out_features,
                 bias=has_bias,

invokeai/backend/load_flux_model_bnb_nf4.py

@@ -4,7 +4,7 @@ from pathlib import Path
 import accelerate
 import torch
 from diffusers.models.transformers.transformer_flux import FluxTransformer2DModel
-from safetensors.torch import load_file, save_file
+from safetensors.torch import load_file
 
 from invokeai.backend.bnb import quantize_model_nf4
 
@@ -43,6 +43,7 @@ def load_flux_transformer(path: Path) -> FluxTransformer2DModel:
         model.to_empty(device="cpu")
         sd = load_file(model_nf4_path / "model.safetensors")
         model.load_state_dict(sd, strict=True)
+        model = model.to("cuda")
 
     else:
         # The quantized model does not exist yet, quantize and save it.
@@ -79,7 +80,7 @@ def load_flux_transformer(path: Path) -> FluxTransformer2DModel:
         model = model.to("cuda")
 
         model_nf4_path.mkdir(parents=True, exist_ok=True)
-        save_file(model.state_dict(), model_nf4_path / "model.safetensors")
+        # save_file(model.state_dict(), model_nf4_path / "model.safetensors")
 
         # ---------------------