Skip to content

Support cosmos predict2 text to image 2B and 14B models. #8517

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
comfyanonymous merged 2 commits into from
Jun 13, 2025
Merged

Support cosmos predict2 text to image 2B and 14B models. #8517

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5355f8e
Support cosmos predict2 text to image 2B and 14B models.
comfyanonymous Jun 13, 2025
eacded5
Fix ruff.
comfyanonymous Jun 13, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 0 additions & 10 deletions comfy/ldm/cosmos/blocks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,16 +26,6 @@
from comfy.ldm.modules.attention import optimized_attention


def apply_rotary_pos_emb(
t: torch.Tensor,
freqs: torch.Tensor,
) -> torch.Tensor:
t_ = t.reshape(*t.shape[:-1], 2, -1).movedim(-2, -1).unsqueeze(-2).float()
t_out = freqs[..., 0] * t_[..., 0] + freqs[..., 1] * t_[..., 1]
t_out = t_out.movedim(-1, -2).reshape(*t.shape).type_as(t)
return t_out


def get_normalization(name: str, channels: int, weight_args={}, operations=None):
if name == "I":
return nn.Identity()
Expand Down
4 changes: 3 additions & 1 deletion comfy/ldm/cosmos/position_embedding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,6 +66,7 @@ def __init__(
h_extrapolation_ratio: float = 1.0,
w_extrapolation_ratio: float = 1.0,
t_extrapolation_ratio: float = 1.0,
enable_fps_modulation: bool = True,
device=None,
**kwargs, # used for compatibility with other positional embeddings; unused in this class
):
Expand All @@ -75,6 +76,7 @@ def __init__(
self.base_fps = base_fps
self.max_h = len_h
self.max_w = len_w
self.enable_fps_modulation = enable_fps_modulation

dim = head_dim
dim_h = dim // 6 * 2
Expand Down Expand Up @@ -143,7 +145,7 @@ def generate_embeddings(
half_emb_w = torch.outer(self.seq[:W].to(device=device), w_spatial_freqs)

# apply sequence scaling in temporal dimension
if fps is None: # image case
if fps is None or self.enable_fps_modulation is False: # image case
half_emb_t = torch.outer(self.seq[:T].to(device=device), temporal_freqs)
else:
half_emb_t = torch.outer(self.seq[:T].to(device=device) / fps * self.base_fps, temporal_freqs)
Expand Down
Loading
Loading