Skip to content

LTX 0.9.5 #10968

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
merged 13 commits into from
Mar 18, 2025
Merged

LTX 0.9.5 #10968

Show file tree
Hide file tree
Changes from 9 commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
ea436c4
update
a-r-r-o-w Mar 5, 2025
f35b807
update
a-r-r-o-w Mar 5, 2025
d0bdf4b
update
a-r-r-o-w Mar 5, 2025
199e741
update
a-r-r-o-w Mar 5, 2025
3d3ad59
vae fix
a-r-r-o-w Mar 5, 2025
14a2282
update conversion script; todo: update 0.9.1 checkpoint with timestep…
a-r-r-o-w Mar 5, 2025
e98fea2
[WIP] test prepare_latents for ltx0.95 (#10976)
yiyixuxu Mar 14, 2025
1cfd2ee
Merge branch 'main' into integrations/ltx-0.9.5
yiyixuxu Mar 14, 2025
1141cdd
up
yiyixuxu Mar 14, 2025
21502d9
up
yiyixuxu Mar 14, 2025
cf3a77e
revert the timestep_scale_multiplier change
yiyixuxu Mar 18, 2025
6da0a08
style
yiyixuxu Mar 18, 2025
d649fdf
update
yiyixuxu Mar 18, 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
6 changes: 6 additions & 0 deletions docs/source/en/api/pipelines/ltx_video.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -196,6 +196,12 @@ export_to_video(video, "ship.mp4", fps=24)
- all
- __call__

## LTXConditionPipeline

[[autodoc]] LTXConditionPipeline
- all
- __call__

## LTXPipelineOutput

[[autodoc]] pipelines.ltx.pipeline_output.LTXPipelineOutput
106 changes: 91 additions & 15 deletions scripts/convert_ltx_to_diffusers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,15 +74,38 @@ def remove_keys_(key: str, state_dict: Dict[str, Any]):
"last_scale_shift_table": "scale_shift_table",
}

VAE_095_RENAME_DICT = {
# decoder
"up_blocks.0": "mid_block",
"up_blocks.1": "up_blocks.0.upsamplers.0",
"up_blocks.2": "up_blocks.0",
"up_blocks.3": "up_blocks.1.upsamplers.0",
"up_blocks.4": "up_blocks.1",
"up_blocks.5": "up_blocks.2.upsamplers.0",
"up_blocks.6": "up_blocks.2",
"up_blocks.7": "up_blocks.3.upsamplers.0",
"up_blocks.8": "up_blocks.3",
# encoder
"down_blocks.0": "down_blocks.0",
"down_blocks.1": "down_blocks.0.downsamplers.0",
"down_blocks.2": "down_blocks.1",
"down_blocks.3": "down_blocks.1.downsamplers.0",
"down_blocks.4": "down_blocks.2",
"down_blocks.5": "down_blocks.2.downsamplers.0",
"down_blocks.6": "down_blocks.3",
"down_blocks.7": "down_blocks.3.downsamplers.0",
"down_blocks.8": "mid_block",
# common
"last_time_embedder": "time_embedder",
"last_scale_shift_table": "scale_shift_table",
}

VAE_SPECIAL_KEYS_REMAP = {
"per_channel_statistics.channel": remove_keys_,
"per_channel_statistics.mean-of-means": remove_keys_,
"per_channel_statistics.mean-of-stds": remove_keys_,
"model.diffusion_model": remove_keys_,
}

VAE_091_SPECIAL_KEYS_REMAP = {
"timestep_scale_multiplier": remove_keys_,
"decoder.timestep_scale_multiplier": remove_keys_,
}


Expand All @@ -104,12 +127,16 @@ def update_state_dict_inplace(state_dict: Dict[str, Any], old_key: str, new_key:
def convert_transformer(
ckpt_path: str,
dtype: torch.dtype,
version: str = "0.9.0",
):
PREFIX_KEY = "model.diffusion_model."

original_state_dict = get_state_dict(load_file(ckpt_path))
config = {}
if version == "0.9.5":
config["_use_causal_rope_fix"] = True
with init_empty_weights():
transformer = LTXVideoTransformer3DModel()
transformer = LTXVideoTransformer3DModel(**config)

for key in list(original_state_dict.keys()):
new_key = key[:]
Expand Down Expand Up @@ -161,12 +188,19 @@ def get_vae_config(version: str) -> Dict[str, Any]:
"out_channels": 3,
"latent_channels": 128,
"block_out_channels": (128, 256, 512, 512),
"down_block_types": (
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
),
"decoder_block_out_channels": (128, 256, 512, 512),
"layers_per_block": (4, 3, 3, 3, 4),
"decoder_layers_per_block": (4, 3, 3, 3, 4),
"spatio_temporal_scaling": (True, True, True, False),
"decoder_spatio_temporal_scaling": (True, True, True, False),
"decoder_inject_noise": (False, False, False, False, False),
"downsample_type": ("conv", "conv", "conv", "conv"),
"upsample_residual": (False, False, False, False),
"upsample_factor": (1, 1, 1, 1),
"patch_size": 4,
Expand All @@ -183,12 +217,19 @@ def get_vae_config(version: str) -> Dict[str, Any]:
"out_channels": 3,
"latent_channels": 128,
"block_out_channels": (128, 256, 512, 512),
"down_block_types": (
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
"LTXVideoDownBlock3D",
),
"decoder_block_out_channels": (256, 512, 1024),
"layers_per_block": (4, 3, 3, 3, 4),
"decoder_layers_per_block": (5, 6, 7, 8),
"spatio_temporal_scaling": (True, True, True, False),
"decoder_spatio_temporal_scaling": (True, True, True),
"decoder_inject_noise": (True, True, True, False),
"downsample_type": ("conv", "conv", "conv", "conv"),
"upsample_residual": (True, True, True),
"upsample_factor": (2, 2, 2),
"timestep_conditioning": True,
Expand All @@ -200,7 +241,39 @@ def get_vae_config(version: str) -> Dict[str, Any]:
"decoder_causal": False,
}
VAE_KEYS_RENAME_DICT.update(VAE_091_RENAME_DICT)
VAE_SPECIAL_KEYS_REMAP.update(VAE_091_SPECIAL_KEYS_REMAP)
elif version == "0.9.5":
config = {
"in_channels": 3,
"out_channels": 3,
"latent_channels": 128,
"block_out_channels": (128, 256, 512, 1024, 2048),
"down_block_types": (
"LTXVideo095DownBlock3D",
"LTXVideo095DownBlock3D",
"LTXVideo095DownBlock3D",
"LTXVideo095DownBlock3D",
),
"decoder_block_out_channels": (256, 512, 1024),
"layers_per_block": (4, 6, 6, 2, 2),
"decoder_layers_per_block": (5, 5, 5, 5),
"spatio_temporal_scaling": (True, True, True, True),
"decoder_spatio_temporal_scaling": (True, True, True),
"decoder_inject_noise": (False, False, False, False),
"downsample_type": ("spatial", "temporal", "spatiotemporal", "spatiotemporal"),
"upsample_residual": (True, True, True),
"upsample_factor": (2, 2, 2),
"timestep_conditioning": True,
"patch_size": 4,
"patch_size_t": 1,
"resnet_norm_eps": 1e-6,
"scaling_factor": 1.0,
"encoder_causal": True,
"decoder_causal": False,
"spatial_compression_ratio": 32,
"temporal_compression_ratio": 8,
"timestep_scale_multiplier": 1000.0,
}
VAE_KEYS_RENAME_DICT.update(VAE_095_RENAME_DICT)
return config


Expand All @@ -223,7 +296,7 @@ def get_args():
parser.add_argument("--output_path", type=str, required=True, help="Path where converted model should be saved")
parser.add_argument("--dtype", default="fp32", help="Torch dtype to save the model in.")
parser.add_argument(
"--version", type=str, default="0.9.0", choices=["0.9.0", "0.9.1"], help="Version of the LTX model"
"--version", type=str, default="0.9.0", choices=["0.9.0", "0.9.1", "0.9.5"], help="Version of the LTX model"
)
return parser.parse_args()

Expand Down Expand Up @@ -277,14 +350,17 @@ def get_args():
for param in text_encoder.parameters():
param.data = param.data.contiguous()

scheduler = FlowMatchEulerDiscreteScheduler(
use_dynamic_shifting=True,
base_shift=0.95,
max_shift=2.05,
base_image_seq_len=1024,
max_image_seq_len=4096,
shift_terminal=0.1,
)
if args.version == "0.9.5":
scheduler = FlowMatchEulerDiscreteScheduler(use_dynamic_shifting=False)
else:
scheduler = FlowMatchEulerDiscreteScheduler(
use_dynamic_shifting=True,
base_shift=0.95,
max_shift=2.05,
base_image_seq_len=1024,
max_image_seq_len=4096,
shift_terminal=0.1,
)

pipe = LTXPipeline(
scheduler=scheduler,
Expand Down
2 changes: 2 additions & 0 deletions src/diffusers/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -401,6 +401,7 @@
"LDMTextToImagePipeline",
"LEditsPPPipelineStableDiffusion",
"LEditsPPPipelineStableDiffusionXL",
"LTXConditionPipeline",
"LTXImageToVideoPipeline",
"LTXPipeline",
"Lumina2Pipeline",
Expand Down Expand Up @@ -945,6 +946,7 @@
LDMTextToImagePipeline,
LEditsPPPipelineStableDiffusion,
LEditsPPPipelineStableDiffusionXL,
LTXConditionPipeline,
LTXImageToVideoPipeline,
LTXPipeline,
Lumina2Pipeline,
Expand Down
Loading
Loading