update

Author: DN6

src/diffusers/models/normalization.py

@@ -243,35 +243,6 @@ def forward(
         return x, gate_msa
 
 
-class AdaLayerNormZeroSinglePruned(nn.Module):
-    r"""
-    Norm layer adaptive layer norm zero (adaLN-Zero).
-
-    Parameters:
-        embedding_dim (`int`): The size of each embedding vector.
-        num_embeddings (`int`): The size of the embeddings dictionary.
-    """
-
-    def __init__(self, embedding_dim: int, norm_type="layer_norm", bias=True):
-        super().__init__()
-
-        if norm_type == "layer_norm":
-            self.norm = nn.LayerNorm(embedding_dim, elementwise_affine=False, eps=1e-6)
-        else:
-            raise ValueError(
-                f"Unsupported `norm_type` ({norm_type}) provided. Supported ones are: 'layer_norm', 'fp32_layer_norm'."
-            )
-
-    def forward(
-        self,
-        x: torch.Tensor,
-        emb: Optional[torch.Tensor] = None,
-    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
-        shift_msa, scale_msa, gate_msa = emb.squeeze(0).chunk(3, dim=0)
-        x = self.norm(x) * (1 + scale_msa[:, None]) + shift_msa[:, None]
-        return x, gate_msa
-
-
 class LuminaRMSNormZero(nn.Module):
     """
     Norm layer adaptive RMS normalization zero.

src/diffusers/models/transformers/transformer_chroma.py

@@ -89,9 +89,7 @@ def __init__(
             persistent=False,
         )
 
-    def forward(
-        self, timestep: torch.Tensor, guidance: Optional[torch.Tensor], pooled_projections: torch.Tensor
-    ) -> torch.Tensor:
+    def forward(self, timestep: torch.Tensor) -> torch.Tensor:
         mod_index_length = self.mod_proj.shape[0]
 
         timesteps_proj = self.time_proj(timestep).to(dtype=timestep.dtype)
@@ -642,7 +640,7 @@ def forward(
         if guidance is not None:
             guidance = guidance.to(hidden_states.dtype) * 1000
 
-        input_vec = self.time_text_embed(timestep, guidance, pooled_projections)
+        input_vec = self.time_text_embed(timestep)
         pooled_temb = self.distilled_guidance_layer(input_vec)
 
         encoder_hidden_states = self.context_embedder(encoder_hidden_states)

src/diffusers/pipelines/flux/pipeline_flux.py

@@ -265,11 +265,6 @@ def _get_t5_prompt_embeds(
         prompt_embeds = self.text_encoder_2(
             text_input_ids.to(device),
             output_hidden_states=False,
-            attention_mask=(
-                self._get_chroma_attn_mask(text_inputs.length, max_sequence_length).to(device)
-                if self.variant == "chroma"
-                else None
-            ),
         )[0]
 
         dtype = self.text_encoder_2.dtype