Upload folder using huggingface_hub
Browse files
.gitattributes
CHANGED
|
@@ -541,3 +541,5 @@ normal_2/samples/step_001750_1.mp4 filter=lfs diff=lfs merge=lfs -text
|
|
| 541 |
normal_2/samples/step_001750_2.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 542 |
normal_2/samples/step_002000_1.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 543 |
normal_2/samples/step_002000_2.mp4 filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 541 |
normal_2/samples/step_001750_2.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 542 |
normal_2/samples/step_002000_1.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 543 |
normal_2/samples/step_002000_2.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 544 |
+
normal_lion/samples/step_000250_1.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 545 |
+
normal_lion/samples/step_000250_2.mp4 filter=lfs diff=lfs merge=lfs -text
|
normal_lion/checkpoints/lora_weights_step_00250.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abf48a3b4c81992ac27587b1611cea867620848174857ed0b07467c1f833c0bc
|
| 3 |
+
size 654443392
|
normal_lion/samples/step_000250_1.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1481c1d19940577ecfa51c724811bd8de0fd03b605a99644c2f6b7516b5650b
|
| 3 |
+
size 373484
|
normal_lion/samples/step_000250_2.mp4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1887644ae4a0daaf28e902235a43a57ed108f123fed865a2008c1edbaa113cc3
|
| 3 |
+
size 340339
|
normal_lion/training_config.yaml
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
acceleration:
|
| 2 |
+
load_text_encoder_in_8bit: false
|
| 3 |
+
mixed_precision_mode: bf16
|
| 4 |
+
quantization: null
|
| 5 |
+
checkpoints:
|
| 6 |
+
interval: 250
|
| 7 |
+
keep_last_n: 10
|
| 8 |
+
precision: bfloat16
|
| 9 |
+
data:
|
| 10 |
+
num_dataloader_workers: 2
|
| 11 |
+
preprocessed_data_root: .precomputed
|
| 12 |
+
flow_matching:
|
| 13 |
+
timestep_sampling_mode: shifted_logit_normal
|
| 14 |
+
timestep_sampling_params: {}
|
| 15 |
+
hub:
|
| 16 |
+
hub_model_id: Alissonerdx/BFS-LTX2
|
| 17 |
+
push_to_hub: true
|
| 18 |
+
lora:
|
| 19 |
+
alpha: 64
|
| 20 |
+
dropout: 0.01
|
| 21 |
+
rank: 64
|
| 22 |
+
target_modules:
|
| 23 |
+
- attn1.to_k
|
| 24 |
+
- attn1.to_q
|
| 25 |
+
- attn1.to_v
|
| 26 |
+
- attn1.to_out.0
|
| 27 |
+
- attn2.to_k
|
| 28 |
+
- attn2.to_q
|
| 29 |
+
- attn2.to_v
|
| 30 |
+
- attn2.to_out.0
|
| 31 |
+
- ff.net.0.proj
|
| 32 |
+
- ff.net.2
|
| 33 |
+
model:
|
| 34 |
+
load_checkpoint: /train/LTX-2/packages/ltx-trainer/outputs/ltx2_v2v_ic_lora_head_swap_128_128_normal/checkpoints
|
| 35 |
+
model_path: /train/models/ltx-2-19b-dev.safetensors
|
| 36 |
+
text_encoder_path: /train/models/gemma
|
| 37 |
+
training_mode: lora
|
| 38 |
+
optimization:
|
| 39 |
+
batch_size: 1
|
| 40 |
+
enable_gradient_checkpointing: true
|
| 41 |
+
gradient_accumulation_steps: 1
|
| 42 |
+
learning_rate: 2.0e-05
|
| 43 |
+
max_grad_norm: 1.0
|
| 44 |
+
optimizer_type: lion
|
| 45 |
+
scheduler_params: {}
|
| 46 |
+
scheduler_type: linear
|
| 47 |
+
steps: 1000
|
| 48 |
+
output_dir: /train/LTX-2/packages/ltx-trainer/outputs/ltx2_v2v_ic_lora_head_swap_128_128_normal_lion
|
| 49 |
+
seed: 42
|
| 50 |
+
training_strategy:
|
| 51 |
+
audio_latents_dir: audio_latents
|
| 52 |
+
first_frame_conditioning_p: 1.0
|
| 53 |
+
name: text_to_video
|
| 54 |
+
with_audio: false
|
| 55 |
+
validation:
|
| 56 |
+
frame_rate: 25.0
|
| 57 |
+
generate_audio: false
|
| 58 |
+
guidance_scale: 4.0
|
| 59 |
+
images:
|
| 60 |
+
- dataset/test/351912466498942.png
|
| 61 |
+
- dataset/test/6403624465969299.png
|
| 62 |
+
include_reference_in_output: true
|
| 63 |
+
inference_steps: 30
|
| 64 |
+
interval: 250
|
| 65 |
+
negative_prompt: worst quality, inconsistent motion, blurry, jittery, distorted
|
| 66 |
+
prompts:
|
| 67 |
+
- High-fidelity face swap, strictly maintaining the identity and facial features
|
| 68 |
+
from the first frame throughout the entire video.
|
| 69 |
+
- High-fidelity face swap, strictly maintaining the identity and facial features
|
| 70 |
+
from the first frame throughout the entire video.
|
| 71 |
+
reference_downscale_factor: 1
|
| 72 |
+
reference_videos: null
|
| 73 |
+
seed: 42
|
| 74 |
+
skip_initial_validation: true
|
| 75 |
+
stg_blocks:
|
| 76 |
+
- 29
|
| 77 |
+
stg_mode: stg_v
|
| 78 |
+
stg_scale: 0.0
|
| 79 |
+
video_dims: !!python/tuple
|
| 80 |
+
- 576
|
| 81 |
+
- 1024
|
| 82 |
+
- 49
|
| 83 |
+
videos_per_prompt: 1
|
| 84 |
+
wandb:
|
| 85 |
+
enabled: true
|
| 86 |
+
entity: null
|
| 87 |
+
log_validation_videos: true
|
| 88 |
+
project: ltx-2-trainer-01-02-2026-normal
|
| 89 |
+
tags:
|
| 90 |
+
- ltx2
|
| 91 |
+
- ic-lora
|
| 92 |
+
- video-to-video
|