update

Author: DN6

src/diffusers/models/transformers/transformer_mochi.py

@@ -373,8 +373,7 @@ def _get_positions(
         return positions
 
     def _create_rope(self, freqs: torch.Tensor, pos: torch.Tensor) -> torch.Tensor:
-        with torch.autocast("cuda", enabled=False):
-            freqs = torch.einsum("nd,dhf->nhf", pos.to(freqs), freqs)
+        freqs = torch.einsum("nd,dhf->nhf", pos.to(freqs), freqs)
         freqs_cos = torch.cos(freqs)
         freqs_sin = torch.sin(freqs)
         return freqs_cos, freqs_sin

src/diffusers/pipelines/mochi/pipeline_mochi.py

@@ -210,7 +210,6 @@ def __init__(
         self.default_height = 480
         self.default_width = 848
 
-    # Adapted from diffusers.pipelines.cogvideo.pipeline_cogvideox.CogVideoXPipeline._get_t5_prompt_embeds
     def _get_t5_prompt_embeds(
         self,
         prompt: Union[str, List[str]] = None,
@@ -233,6 +232,7 @@ def _get_t5_prompt_embeds(
             add_special_tokens=True,
             return_tensors="pt",
         )
+
         text_input_ids = text_inputs.input_ids
         prompt_attention_mask = text_inputs.attention_mask
         prompt_attention_mask = prompt_attention_mask.bool().to(device)