Fix LoRA Trainer bugs with FP8 models. (#9854)

* Fix adapter weight init

* Fix fp8 model training

* Avoid inference tensor

comfy/ops.py

@@ -365,12 +365,13 @@ class fp8_ops(manual_cast):
             return None
 
         def forward_comfy_cast_weights(self, input):
-            try:
+            if not self.training:
-                out = fp8_linear(self, input)
+                try:
-                if out is not None:
+                    out = fp8_linear(self, input)
-                    return out
+                    if out is not None:
-            except Exception as e:
+                        return out
-                logging.info("Exception during fp8 op: {}".format(e))
+                except Exception as e:
+                    logging.info("Exception during fp8 op: {}".format(e))
 
             weight, bias = cast_bias_weight(self, input)
             return torch.nn.functional.linear(input, weight, bias)

comfy/weight_adapter/loha.py

@@ -130,12 +130,12 @@ class LoHaAdapter(WeightAdapterBase):
     def create_train(cls, weight, rank=1, alpha=1.0):
         out_dim = weight.shape[0]
         in_dim = weight.shape[1:].numel()
-        mat1 = torch.empty(out_dim, rank, device=weight.device, dtype=weight.dtype)
+        mat1 = torch.empty(out_dim, rank, device=weight.device, dtype=torch.float32)
-        mat2 = torch.empty(rank, in_dim, device=weight.device, dtype=weight.dtype)
+        mat2 = torch.empty(rank, in_dim, device=weight.device, dtype=torch.float32)
         torch.nn.init.normal_(mat1, 0.1)
         torch.nn.init.constant_(mat2, 0.0)
-        mat3 = torch.empty(out_dim, rank, device=weight.device, dtype=weight.dtype)
+        mat3 = torch.empty(out_dim, rank, device=weight.device, dtype=torch.float32)
-        mat4 = torch.empty(rank, in_dim, device=weight.device, dtype=weight.dtype)
+        mat4 = torch.empty(rank, in_dim, device=weight.device, dtype=torch.float32)
         torch.nn.init.normal_(mat3, 0.1)
         torch.nn.init.normal_(mat4, 0.01)
         return LohaDiff(

comfy/weight_adapter/lokr.py

@@ -89,8 +89,8 @@ class LoKrAdapter(WeightAdapterBase):
         in_dim = weight.shape[1:].numel()
         out1, out2 = factorization(out_dim, rank)
         in1, in2 = factorization(in_dim, rank)
-        mat1 = torch.empty(out1, in1, device=weight.device, dtype=weight.dtype)
+        mat1 = torch.empty(out1, in1, device=weight.device, dtype=torch.float32)
-        mat2 = torch.empty(out2, in2, device=weight.device, dtype=weight.dtype)
+        mat2 = torch.empty(out2, in2, device=weight.device, dtype=torch.float32)
         torch.nn.init.kaiming_uniform_(mat2, a=5**0.5)
         torch.nn.init.constant_(mat1, 0.0)
         return LokrDiff(

comfy/weight_adapter/lora.py

@@ -66,8 +66,8 @@ class LoRAAdapter(WeightAdapterBase):
     def create_train(cls, weight, rank=1, alpha=1.0):
         out_dim = weight.shape[0]
         in_dim = weight.shape[1:].numel()
-        mat1 = torch.empty(out_dim, rank, device=weight.device, dtype=weight.dtype)
+        mat1 = torch.empty(out_dim, rank, device=weight.device, dtype=torch.float32)
-        mat2 = torch.empty(rank, in_dim, device=weight.device, dtype=weight.dtype)
+        mat2 = torch.empty(rank, in_dim, device=weight.device, dtype=torch.float32)
         torch.nn.init.kaiming_uniform_(mat1, a=5**0.5)
         torch.nn.init.constant_(mat2, 0.0)
         return LoraDiff(

comfy/weight_adapter/oft.py

@@ -68,7 +68,7 @@ class OFTAdapter(WeightAdapterBase):
     def create_train(cls, weight, rank=1, alpha=1.0):
         out_dim = weight.shape[0]
         block_size, block_num = factorization(out_dim, rank)
-        block = torch.zeros(block_num, block_size, block_size, device=weight.device, dtype=weight.dtype)
+        block = torch.zeros(block_num, block_size, block_size, device=weight.device, dtype=torch.float32)
         return OFTDiff(
             (block, None, alpha, None)
         )

comfy_extras/nodes_train.py

@@ -38,6 +38,23 @@ def make_batch_extra_option_dict(d, indicies, full_size=None):
     return new_dict
 
 
+def process_cond_list(d, prefix=""):
+    if hasattr(d, "__iter__") and not hasattr(d, "items"):
+        for index, item in enumerate(d):
+            process_cond_list(item, f"{prefix}.{index}")
+        return d
+    elif hasattr(d, "items"):
+        for k, v in list(d.items()):
+            if isinstance(v, dict):
+                process_cond_list(v, f"{prefix}.{k}")
+            elif isinstance(v, torch.Tensor):
+                d[k] = v.clone()
+            elif isinstance(v, (list, tuple)):
+                for index, item in enumerate(v):
+                    process_cond_list(item, f"{prefix}.{k}.{index}")
+    return d
+
+
 class TrainSampler(comfy.samplers.Sampler):
     def __init__(self, loss_fn, optimizer, loss_callback=None, batch_size=1, grad_acc=1, total_steps=1, seed=0, training_dtype=torch.bfloat16):
         self.loss_fn = loss_fn
@@ -50,6 +67,7 @@ class TrainSampler(comfy.samplers.Sampler):
         self.training_dtype = training_dtype
 
     def sample(self, model_wrap, sigmas, extra_args, callback, noise, latent_image=None, denoise_mask=None, disable_pbar=False):
+        model_wrap.conds = process_cond_list(model_wrap.conds)
         cond = model_wrap.conds["positive"]
         dataset_size = sigmas.size(0)
         torch.cuda.empty_cache()