Make old scaled fp8 format use the new mixed quant ops system. (#11000)

comfy/ops.py

@@ -23,6 +23,7 @@ from comfy.cli_args import args, PerformanceFeature
 import comfy.float
 import comfy.rmsnorm
 import contextlib
+import json
 
 def run_every_op():
     if torch.compiler.is_compiling():
@@ -422,22 +423,12 @@ def fp8_linear(self, input):
 
     if input.ndim == 3 or input.ndim == 2:
         w, bias, offload_stream = cast_bias_weight(self, input, dtype=dtype, bias_dtype=input_dtype, offloadable=True)
+        scale_weight = torch.ones((), device=input.device, dtype=torch.float32)
 
-        scale_weight = self.scale_weight
-        scale_input = self.scale_input
-        if scale_weight is None:
-            scale_weight = torch.ones((), device=input.device, dtype=torch.float32)
-        else:
-            scale_weight = scale_weight.to(input.device)
-
-        if scale_input is None:
-            scale_input = torch.ones((), device=input.device, dtype=torch.float32)
-            input = torch.clamp(input, min=-448, max=448, out=input)
-            layout_params_weight = {'scale': scale_input, 'orig_dtype': input_dtype}
-            quantized_input = QuantizedTensor(input.to(dtype).contiguous(), "TensorCoreFP8Layout", layout_params_weight)
-        else:
-            scale_input = scale_input.to(input.device)
-            quantized_input = QuantizedTensor.from_float(input, "TensorCoreFP8Layout", scale=scale_input, dtype=dtype)
+        scale_input = torch.ones((), device=input.device, dtype=torch.float32)
+        input = torch.clamp(input, min=-448, max=448, out=input)
+        layout_params_weight = {'scale': scale_input, 'orig_dtype': input_dtype}
+        quantized_input = QuantizedTensor(input.to(dtype).contiguous(), "TensorCoreFP8Layout", layout_params_weight)
 
         # Wrap weight in QuantizedTensor - this enables unified dispatch
         # Call F.linear - __torch_dispatch__ routes to fp8_linear handler in quant_ops.py!
@@ -458,7 +449,7 @@ class fp8_ops(manual_cast):
             return None
 
         def forward_comfy_cast_weights(self, input):
-            if not self.training:
+            if len(self.weight_function) == 0 and len(self.bias_function) == 0:
                 try:
                     out = fp8_linear(self, input)
                     if out is not None:
@@ -471,59 +462,6 @@ class fp8_ops(manual_cast):
             uncast_bias_weight(self, weight, bias, offload_stream)
             return x
 
-def scaled_fp8_ops(fp8_matrix_mult=False, scale_input=False, override_dtype=None):
-    logging.info("Using scaled fp8: fp8 matrix mult: {}, scale input: {}".format(fp8_matrix_mult, scale_input))
-    class scaled_fp8_op(manual_cast):
-        class Linear(manual_cast.Linear):
-            def __init__(self, *args, **kwargs):
-                if override_dtype is not None:
-                    kwargs['dtype'] = override_dtype
-                super().__init__(*args, **kwargs)
-
-            def reset_parameters(self):
-                if not hasattr(self, 'scale_weight'):
-                    self.scale_weight = torch.nn.parameter.Parameter(data=torch.ones((), device=self.weight.device, dtype=torch.float32), requires_grad=False)
-
-                if not scale_input:
-                    self.scale_input = None
-
-                if not hasattr(self, 'scale_input'):
-                    self.scale_input = torch.nn.parameter.Parameter(data=torch.ones((), device=self.weight.device, dtype=torch.float32), requires_grad=False)
-                return None
-
-            def forward_comfy_cast_weights(self, input):
-                if fp8_matrix_mult:
-                    out = fp8_linear(self, input)
-                    if out is not None:
-                        return out
-
-                weight, bias, offload_stream = cast_bias_weight(self, input, offloadable=True)
-
-                if weight.numel() < input.numel(): #TODO: optimize
-                    x = torch.nn.functional.linear(input, weight * self.scale_weight.to(device=weight.device, dtype=weight.dtype), bias)
-                else:
-                    x = torch.nn.functional.linear(input * self.scale_weight.to(device=weight.device, dtype=weight.dtype), weight, bias)
-                uncast_bias_weight(self, weight, bias, offload_stream)
-                return x
-
-            def convert_weight(self, weight, inplace=False, **kwargs):
-                if inplace:
-                    weight *= self.scale_weight.to(device=weight.device, dtype=weight.dtype)
-                    return weight
-                else:
-                    return weight.to(dtype=torch.float32) * self.scale_weight.to(device=weight.device, dtype=torch.float32)
-
-            def set_weight(self, weight, inplace_update=False, seed=None, return_weight=False, **kwargs):
-                weight = comfy.float.stochastic_rounding(weight / self.scale_weight.to(device=weight.device, dtype=weight.dtype), self.weight.dtype, seed=seed)
-                if return_weight:
-                    return weight
-                if inplace_update:
-                    self.weight.data.copy_(weight)
-                else:
-                    self.weight = torch.nn.Parameter(weight, requires_grad=False)
-
-    return scaled_fp8_op
-
 CUBLAS_IS_AVAILABLE = False
 try:
     from cublas_ops import CublasLinear
@@ -550,9 +488,9 @@ if CUBLAS_IS_AVAILABLE:
 from .quant_ops import QuantizedTensor, QUANT_ALGOS
 
 
-def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, full_precision_mm=False):
+def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_precision_mm=False):
     class MixedPrecisionOps(manual_cast):
-        _layer_quant_config = layer_quant_config
+        _quant_config = quant_config
         _compute_dtype = compute_dtype
         _full_precision_mm = full_precision_mm
 
@@ -595,27 +533,36 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
 
                 manually_loaded_keys = [weight_key]
 
-                if layer_name not in MixedPrecisionOps._layer_quant_config:
+                layer_conf = state_dict.pop(f"{prefix}comfy_quant", None)
+                if layer_conf is not None:
+                    layer_conf = json.loads(layer_conf.numpy().tobytes())
+
+                if layer_conf is None:
                     self.weight = torch.nn.Parameter(weight.to(device=device, dtype=MixedPrecisionOps._compute_dtype), requires_grad=False)
                 else:
-                    quant_format = MixedPrecisionOps._layer_quant_config[layer_name].get("format", None)
-                    if quant_format is None:
+                    self.quant_format = layer_conf.get("format", None)
+                    if not self._full_precision_mm:
+                        self._full_precision_mm = layer_conf.get("full_precision_matrix_mult", False)
+
+                    if self.quant_format is None:
                         raise ValueError(f"Unknown quantization format for layer {layer_name}")
 
-                    qconfig = QUANT_ALGOS[quant_format]
+                    qconfig = QUANT_ALGOS[self.quant_format]
                     self.layout_type = qconfig["comfy_tensor_layout"]
 
                     weight_scale_key = f"{prefix}weight_scale"
+                    scale = state_dict.pop(weight_scale_key, None)
                     layout_params = {
-                        'scale': state_dict.pop(weight_scale_key, None),
+                        'scale': scale,
                         'orig_dtype': MixedPrecisionOps._compute_dtype,
                         'block_size': qconfig.get("group_size", None),
                     }
-                    if layout_params['scale'] is not None:
+
+                    if scale is not None:
                         manually_loaded_keys.append(weight_scale_key)
 
                     self.weight = torch.nn.Parameter(
-                        QuantizedTensor(weight.to(device=device), self.layout_type, layout_params),
+                        QuantizedTensor(weight.to(device=device, dtype=qconfig.get("storage_t", None)), self.layout_type, layout_params),
                         requires_grad=False
                     )
 
@@ -624,7 +571,7 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
                         _v = state_dict.pop(param_key, None)
                         if _v is None:
                             continue
-                        setattr(self, param_name, torch.nn.Parameter(_v.to(device=device), requires_grad=False))
+                        self.register_parameter(param_name, torch.nn.Parameter(_v.to(device=device), requires_grad=False))
                         manually_loaded_keys.append(param_key)
 
                 super()._load_from_state_dict(state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs)
@@ -633,6 +580,16 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
                     if key in missing_keys:
                         missing_keys.remove(key)
 
+            def state_dict(self, *args, destination=None, prefix="", **kwargs):
+                sd = super().state_dict(*args, destination=destination, prefix=prefix, **kwargs)
+                if isinstance(self.weight, QuantizedTensor):
+                    sd["{}weight_scale".format(prefix)] = self.weight._layout_params['scale']
+                    quant_conf = {"format": self.quant_format}
+                    if self._full_precision_mm:
+                        quant_conf["full_precision_matrix_mult"] = True
+                    sd["{}comfy_quant".format(prefix)] = torch.frombuffer(json.dumps(quant_conf).encode('utf-8'), dtype=torch.uint8)
+                return sd
+
             def _forward(self, input, weight, bias):
                 return torch.nn.functional.linear(input, weight, bias)
 
@@ -648,9 +605,8 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
                 if self._full_precision_mm or self.comfy_cast_weights or len(self.weight_function) > 0 or len(self.bias_function) > 0:
                     return self.forward_comfy_cast_weights(input, *args, **kwargs)
                 if (getattr(self, 'layout_type', None) is not None and
-                    getattr(self, 'input_scale', None) is not None and
                     not isinstance(input, QuantizedTensor)):
-                    input = QuantizedTensor.from_float(input, self.layout_type, scale=self.input_scale, dtype=self.weight.dtype)
+                    input = QuantizedTensor.from_float(input, self.layout_type, scale=getattr(self, 'input_scale', None), dtype=self.weight.dtype)
                 return self._forward(input, self.weight, self.bias)
 
             def convert_weight(self, weight, inplace=False, **kwargs):
@@ -661,7 +617,7 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
 
             def set_weight(self, weight, inplace_update=False, seed=None, return_weight=False, **kwargs):
                 if getattr(self, 'layout_type', None) is not None:
-                    weight = QuantizedTensor.from_float(weight, self.layout_type, scale=None, dtype=self.weight.dtype, stochastic_rounding=seed, inplace_ops=True)
+                    weight = QuantizedTensor.from_float(weight, self.layout_type, scale="recalculate", dtype=self.weight.dtype, stochastic_rounding=seed, inplace_ops=True)
                 else:
                     weight = weight.to(self.weight.dtype)
                 if return_weight:
@@ -670,17 +626,28 @@ def mixed_precision_ops(layer_quant_config={}, compute_dtype=torch.bfloat16, ful
                 assert inplace_update is False  # TODO: eventually remove the inplace_update stuff
                 self.weight = torch.nn.Parameter(weight, requires_grad=False)
 
+            def _apply(self, fn, recurse=True):  # This is to get torch.compile + moving weights to another device working
+                if recurse:
+                    for module in self.children():
+                        module._apply(fn)
+
+                for key, param in self._parameters.items():
+                    if param is None:
+                        continue
+                    self.register_parameter(key, torch.nn.Parameter(fn(param), requires_grad=False))
+                for key, buf in self._buffers.items():
+                    if buf is not None:
+                        self._buffers[key] = fn(buf)
+                return self
+
     return MixedPrecisionOps
 
-def pick_operations(weight_dtype, compute_dtype, load_device=None, disable_fast_fp8=False, fp8_optimizations=False, scaled_fp8=None, model_config=None):
+def pick_operations(weight_dtype, compute_dtype, load_device=None, disable_fast_fp8=False, fp8_optimizations=False, model_config=None):
     fp8_compute = comfy.model_management.supports_fp8_compute(load_device) # TODO: if we support more ops this needs to be more granular
 
-    if model_config and hasattr(model_config, 'layer_quant_config') and model_config.layer_quant_config:
-        logging.info(f"Using mixed precision operations: {len(model_config.layer_quant_config)} quantized layers")
-        return mixed_precision_ops(model_config.layer_quant_config, compute_dtype, full_precision_mm=not fp8_compute)
-
-    if scaled_fp8 is not None:
-        return scaled_fp8_ops(fp8_matrix_mult=fp8_compute and fp8_optimizations, scale_input=fp8_optimizations, override_dtype=scaled_fp8)
+    if model_config and hasattr(model_config, 'quant_config') and model_config.quant_config:
+        logging.info("Using mixed precision operations")
+        return mixed_precision_ops(model_config.quant_config, compute_dtype, full_precision_mm=not fp8_compute)
 
     if (
         fp8_compute and