Skip to content
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

refactoring: migration to poetry #25

Merged
merged 1 commit into from
Jan 28, 2025
Merged

refactoring: migration to poetry #25

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
12 changes: 6 additions & 6 deletions configs/000_videocrafter/vc1_i2v_512.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
model:
target: src.base.ddpm3d.LatentVisualDiffusionFlow
target: videotuna.base.ddpm3d.LatentVisualDiffusionFlow
params:
linear_start: 0.00085
linear_end: 0.012
Expand All @@ -21,14 +21,14 @@ model:
finegrained: true

diffusion_scheduler_config:
target: src.base.diffusion_schedulers.LDMScheduler
target: videotuna.base.diffusion_schedulers.LDMScheduler
params:
timesteps: 1000
linear_start: 0.00085
linear_end: 0.012

unet_config:
target: src.lvdm.modules.networks.openaimodel3d.UNetModel
target: videotuna.lvdm.modules.networks.openaimodel3d.UNetModel
params:
in_channels: 4
out_channels: 4
Expand Down Expand Up @@ -58,7 +58,7 @@ model:
addition_attention: true
fps_cond: true
first_stage_config:
target: src.lvdm.modules.vae.autoencoder.AutoencoderKL
target: videotuna.lvdm.modules.vae.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
Expand All @@ -80,11 +80,11 @@ model:
lossconfig:
target: torch.nn.Identity
cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
params:
freeze: true
layer: penultimate
img_cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPImageEmbedderV2
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPImageEmbedderV2
params:
freeze: true
10 changes: 5 additions & 5 deletions configs/000_videocrafter/vc1_t2v_1024.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
model:
target: src.base.ddpm3d.LVDMFlow
target: videotuna.base.ddpm3d.LVDMFlow
params:
linear_start: 0.00085
linear_end: 0.012
Expand All @@ -20,14 +20,14 @@ model:
fix_scale_bug: true

diffusion_scheduler_config:
target: src.base.diffusion_schedulers.LDMScheduler
target: videotuna.base.diffusion_schedulers.LDMScheduler
params:
timesteps: 1000
linear_start: 0.00085
linear_end: 0.012

unet_config:
target: src.lvdm.modules.networks.openaimodel3d.UNetModel
target: videotuna.lvdm.modules.networks.openaimodel3d.UNetModel
params:
in_channels: 4
out_channels: 4
Expand Down Expand Up @@ -56,7 +56,7 @@ model:
addition_attention: true
fps_cond: true
first_stage_config:
target: src.lvdm.modules.vae.autoencoder.AutoencoderKL
target: videotuna.lvdm.modules.vae.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
Expand All @@ -78,7 +78,7 @@ model:
lossconfig:
target: torch.nn.Identity
cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
params:
freeze: true
layer: penultimate
18 changes: 9 additions & 9 deletions configs/001_videocrafter2/vc2_t2v_320x512.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ model:
base_learning_rate: 6.0e-06 # 1.5e-04
scale_lr: False
# empty_params_only: True # disable this means finetuning all parameters
target: src.base.ddpm3d.LVDMFlow
target: videotuna.base.ddpm3d.LVDMFlow
params:
log_every_t: 200
first_stage_key: video
Expand All @@ -23,14 +23,14 @@ model:
scale_b: 0.7 # adapt to videocrafter-v2

diffusion_scheduler_config:
target: src.base.diffusion_schedulers.LDMScheduler
target: videotuna.base.diffusion_schedulers.LDMScheduler
params:
timesteps: 1000
linear_start: 0.00085
linear_end: 0.012

unet_config:
target: src.lvdm.modules.networks.openaimodel3d.UNetModel
target: videotuna.lvdm.modules.networks.openaimodel3d.UNetModel
params:
in_channels: 4
out_channels: 4
Expand Down Expand Up @@ -59,7 +59,7 @@ model:
addition_attention: true
fps_cond: true # adapt to videocrafter-v2
first_stage_config:
target: src.lvdm.modules.vae.autoencoder.AutoencoderKL
target: videotuna.lvdm.modules.vae.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
Expand All @@ -81,27 +81,27 @@ model:
lossconfig:
target: torch.nn.Identity
cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
params:
freeze: true
layer: penultimate

data:
target: src.data.lightning_data.DataModuleFromConfig
target: videotuna.data.lightning_data.DataModuleFromConfig
params:
batch_size: 4
num_workers: 16
wrap: false
train:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [320, 512]
video_length: 16
frame_interval: 3
train: True
validation:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [320, 512]
Expand All @@ -118,7 +118,7 @@ lightning:
precision: bf16 # training precision
callbacks:
image_logger:
target: src.utils.callbacks.ImageLogger
target: videotuna.utils.callbacks.ImageLogger
params:
batch_frequency: 500 #2
max_images: 6
Expand Down
22 changes: 11 additions & 11 deletions configs/001_videocrafter2/vc2_t2v_lora.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ model:
base_learning_rate: 6.0e-06 # 1.5e-04
scale_lr: False
# empty_params_only: True # comment this means finetuning all parameters
target: src.base.ddpm3d.LVDMFlow
target: videotuna.base.ddpm3d.LVDMFlow
params:
lora_args:
# lora_ckpt: "/path/to/lora.ckpt" # no need for the first-time training, only used for resume training.
Expand All @@ -29,14 +29,14 @@ model:
scale_b: 0.7 # adapt to videocrafter-v2

diffusion_scheduler_config:
target: src.base.diffusion_schedulers.LDMScheduler
target: videotuna.base.diffusion_schedulers.LDMScheduler
params:
timesteps: 1000
linear_start: 0.00085
linear_end: 0.012

unet_config:
target: src.lvdm.modules.networks.openaimodel3d.UNetModel
target: videotuna.lvdm.modules.networks.openaimodel3d.UNetModel
params:
in_channels: 4
out_channels: 4
Expand Down Expand Up @@ -65,7 +65,7 @@ model:
addition_attention: true
fps_cond: true # adapt to videocrafter-v2
first_stage_config:
target: src.lvdm.modules.vae.autoencoder.AutoencoderKL
target: videotuna.lvdm.modules.vae.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
Expand All @@ -87,27 +87,27 @@ model:
lossconfig:
target: torch.nn.Identity
cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
params:
freeze: true
layer: penultimate

data:
target: src.data.lightning_data.DataModuleFromConfig
target: videotuna.data.lightning_data.DataModuleFromConfig
params:
batch_size: 4
num_workers: 16
wrap: false
train:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [320, 512]
video_length: 16
frame_interval: 3
train: True
validation:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [320, 512]
Expand All @@ -125,7 +125,7 @@ lightning:
precision: bf16 # training precision
callbacks:
image_logger:
target: src.utils.callbacks.ImageLogger
target: videotuna.utils.callbacks.ImageLogger
params:
batch_frequency: 1000
max_images: 2
Expand All @@ -134,12 +134,12 @@ lightning:
unconditional_guidance_scale: 12 # need this, otherwise it is grey
modelcheckpoint:
# target: pytorch_lightning.callbacks.ModelCheckpoint
target: src.utils.callbacks.LoraModelCheckpoint
target: videotuna.utils.callbacks.LoraModelCheckpoint
params:
every_n_epochs: 1
filename: "{epoch:04}-{step:06}"
metrics_over_trainsteps_checkpoint:
target: src.utils.callbacks.LoraModelCheckpoint
target: videotuna.utils.callbacks.LoraModelCheckpoint
params:
filename: "{epoch:06}-{step:09}"
save_weights_only: False
Expand Down
22 changes: 11 additions & 11 deletions configs/002_dynamicrafter/dc_i2v_1024.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
model:
base_learning_rate: 1.0e-05
scale_lr: False
target: src.base.ddpm3d.LatentVisualDiffusionFlow
target: videotuna.base.ddpm3d.LatentVisualDiffusionFlow
params:
parameterization: "v"
log_every_t: 200
Expand All @@ -23,15 +23,15 @@ model:
fps_condition_type: 'fps'

diffusion_scheduler_config:
target: src.base.diffusion_schedulers.LDMScheduler
target: videotuna.base.diffusion_schedulers.LDMScheduler
params:
timesteps: 1000
linear_start: 0.00085
linear_end: 0.012
rescale_betas_zero_snr: True

unet_config:
target: src.lvdm.modules.networks.openaimodel3d_dc.UNetModel
target: videotuna.lvdm.modules.networks.openaimodel3d_dc.UNetModel
params:
in_channels: 8
out_channels: 4
Expand Down Expand Up @@ -64,7 +64,7 @@ model:
fs_condition: true

first_stage_config:
target: src.lvdm.modules.vae.autoencoder.AutoencoderKL
target: videotuna.lvdm.modules.vae.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
Expand All @@ -87,18 +87,18 @@ model:
target: torch.nn.Identity

cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPEmbedder
params:
freeze: true
layer: "penultimate"

img_cond_stage_config:
target: src.lvdm.modules.encoders.condition.FrozenOpenCLIPImageEmbedderV2
target: videotuna.lvdm.modules.encoders.condition.FrozenOpenCLIPImageEmbedderV2
params:
freeze: true

image_proj_stage_config:
target: src.lvdm.modules.encoders.ip_resampler.Resampler
target: videotuna.lvdm.modules.encoders.ip_resampler.Resampler
params:
dim: 1024
depth: 4
Expand All @@ -111,21 +111,21 @@ model:
video_length: 16

data:
target: src.data.lightning_data.DataModuleFromConfig
target: videotuna.data.lightning_data.DataModuleFromConfig
params:
batch_size: 2
num_workers: 16
wrap: false
train:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [576, 1024]
video_length: 16
frame_interval: 3
train: True
validation:
target: src.data.datasets.DatasetFromCSV
target: videotuna.data.datasets.DatasetFromCSV
params:
csv_path: Dataset/ToyDataset/toydataset.csv
resolution: [576, 1024]
Expand All @@ -146,7 +146,7 @@ lightning:
gradient_clip_val: 0.5
callbacks:
image_logger:
target: src.utils.callbacks.ImageLogger
target: videotuna.utils.callbacks.ImageLogger
params:
batch_frequency: 2
save_dir: 'outputs/samples'
Expand Down
Loading