wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-01-10 06:10:50 +08:00
Code Issues Actions 10 Packages Projects Releases Wiki Activity

Lower ltxv text encoder vram use.

Browse Source
This commit is contained in:
comfyanonymous 2026-01-07 19:06:16 -05:00
parent 1c705f7bfb
commit 76aa79384f
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
comfy/text_encoders/lt.py
View File

@ -98,10 +98,13 @@ class LTXAVTEModel(torch.nn.Module):
out, pooled, extra = self.gemma3_12b.encode_token_weights(token_weight_pairs)
out_device = out.device
if comfy.model_management.should_use_bf16(self.execution_device):
out = out.to(device=self.execution_device, dtype=torch.bfloat16)
out = out.movedim(1, -1).to(self.execution_device)
out = 8.0 * (out - out.mean(dim=(1, 2), keepdim=True)) / (out.amax(dim=(1, 2), keepdim=True) - out.amin(dim=(1, 2), keepdim=True) + 1e-6)
out = out.reshape((out.shape[0], out.shape[1], -1))
out = self.text_embedding_projection(out)
out = out.float()
out_vid = self.video_embeddings_connector(out)[0]
out_audio = self.audio_embeddings_connector(out)[0]
out = torch.concat((out_vid, out_audio), dim=-1)