upload checkpoints

results/car_turn/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8df8248bd60072f838fe9ecb16aa483fd3d84a459f83de0f7155f0f178df361f
+size 54192856

results/car_turn/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0221e70227a25ae58f1767fb37d48b4721240aa962aad85f06eff95dbed4d72c
+size 295737352

results/car_turn/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/02
+  validation_steps: 2000
+  checkpointing_steps: 50
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 500
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/5048253f30b40909d274ab1248fa8a0bf881ea4976ceab7f4957313b7d15be67/hybrid.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 576
+  height: 320
+  n_sample_frames: 24
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 576
+  height: 320
+  num_frames: 24
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/dolly_zoom/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d20dc41f61c7d58ad6609566e1ed6eac5d0e52bce494e84e2609b114ddd78c4e
+size 31362776

results/dolly_zoom/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:450e0f53ab4872ba2ade2f3bbe7c651464e39a2cfad9336f4de3d9a1cc8028fd
+size 277714952

results/dolly_zoom/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/04
+  validation_steps: 2000
+  checkpointing_steps: 100
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 400
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/a31b567c8f08f688bd0a79e5f3f5af57a7967ba4d054e0f4c5b71f192af241e9/dolly_zoom_out.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 400
+  height: 400
+  n_sample_frames: 16
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 400
+  height: 400
+  num_frames: 16
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/orbit_shot/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20561bf4fc68685f5e1339a9b4c9cc749266a85984ba2284ca309cfc769803b8
+size 46164696

results/orbit_shot/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3006dbf459be771c60d6d4334f2dac63bd845905e61bddfbd98186bf5f2565f
+size 394041352

results/orbit_shot/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/07
+  validation_steps: 2000
+  checkpointing_steps: 100
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 500
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/5353948117dc7b8683fe28682a334e8076ba7e9ec45763cc8860b1758fb29b94/orbit_shot.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 640
+  height: 368
+  n_sample_frames: 16
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 640
+  height: 368
+  num_frames: 16
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/pan_up/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c02007a490e09f7f835e48a078e46b5204df7f3eb5ccfbe8000b4867c094ecb
+size 36129496

results/pan_up/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf9b7c679cc3c028f52a688f983de3a9e74a81e86f33423eceaf99189ae67035
+size 295737352

results/pan_up/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/03
+  validation_steps: 2000
+  checkpointing_steps: 50
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 500
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/f794a1692f7472c03b0ad36870908f81af7ce5ee7cc61a73ccda287da11f5dfa/camera.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 576
+  height: 320
+  n_sample_frames: 24
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 576
+  height: 320
+  num_frames: 24
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/santa_dance/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b4347b062704cb98a54d53ac42d63c01e5e37f2b96bdc38a6726aa4f35c1c47
+size 46164696

results/santa_dance/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d499bd32fdfd0b07e61f5c52e4a76ad43fb7c849f245f5a1bb3e33f765fba6aa
+size 394041352

results/santa_dance/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/07
+  validation_steps: 2000
+  checkpointing_steps: 100
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 500
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/e9c3bf4f837c93c3246204098c8139985534a94dfef880786ca6fd89c9ff89d1/santa_dance.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 640
+  height: 368
+  n_sample_frames: 16
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 640
+  height: 368
+  num_frames: 16
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/train_ride/cached_latents/cached_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52600e72a6e4f5dbfaa02dddfafd68744b59d209398c0b8964ffa797036b8db7
+size 32115416

results/train_ride/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec2ad3186c036c32f25b07c3c978c91a75872b33145c85b1e26a618eec825988
+size 262969352

results/train_ride/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/06
+  validation_steps: 2000
+  checkpointing_steps: 100
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 400
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/91dd9b06d279d3b24bced706161c8fe3a911d3e853ed9bb6d92fbd69127e07ac/train_ride.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 512
+  height: 320
+  n_sample_frames: 16
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 512
+  height: 320
+  num_frames: 16
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false

results/walk/checkpoint/motion_embed.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ed1d3c74c0e870c764550e7b5dea07fa6392e5e9c6851591db24459531f81a4
+size 262969352

results/walk/config.yaml

@@ -0,0 +1,61 @@
+model:
+  type: unet
+  pretrained_model_path: cerspense/zeroscope_v2_576w
+  motion_embeddings:
+    combinations:
+    - - down
+      - 1280
+    - - up
+      - 1280
+  unet: videoCrafter2
+train:
+  output_dir: results/05
+  validation_steps: 2000
+  checkpointing_steps: 100
+  checkpointing_start: 200
+  train_batch_size: 1
+  max_train_steps: 200
+  gradient_accumulation_steps: 1
+  cache_latents: true
+  cached_latent_dir: null
+  logger_type: tensorboard
+  mixed_precision: fp16
+  use_8bit_adam: false
+  resume_from_checkpoint: null
+  resume_step: null
+dataset:
+  type:
+  - single_video
+  single_video_path: /tmp/gradio/399a0cb64057460050f4db313c05d150200f8a54bbbd783d56f078e74ee64037/walk.mp4
+  single_video_prompt: A car turnaround in a city street
+  width: 512
+  height: 320
+  n_sample_frames: 16
+  fps: 8
+  sample_start_idx: 1
+  frame_step: 1
+  use_bucketing: false
+  use_caption: false
+loss:
+  type: BaseLoss
+  learning_rate: 0.02
+  lr_scheduler: constant
+  lr_warmup_steps: 0
+noise_init:
+  type: BlendInit
+  noise_prior: 0.5
+val:
+  prompt:
+  - A skateboard slides along a city lane
+  negative_prompt: ''
+  sample_preview: true
+  width: 512
+  height: 320
+  num_frames: 16
+  num_inference_steps: 30
+  guidance_scale: 12.0
+  seeds:
+  - 0
+strategy:
+  vSpatial_frameSubtraction: false
+  removeMFromV: false