Spaces:
Paused
Paused
Delete configs/generate_distributed.yaml
Browse files
configs/generate_distributed.yaml
DELETED
|
@@ -1,94 +0,0 @@
|
|
| 1 |
-
__object__:
|
| 2 |
-
path: generate_distributed
|
| 3 |
-
name: VINCIEGenerator
|
| 4 |
-
|
| 5 |
-
dit:
|
| 6 |
-
model:
|
| 7 |
-
__object__:
|
| 8 |
-
path: models.dit.nadit
|
| 9 |
-
name: NaDiT
|
| 10 |
-
args: as_params
|
| 11 |
-
vid_in_channels: 33
|
| 12 |
-
vid_out_channels: 16
|
| 13 |
-
vid_dim: 2560
|
| 14 |
-
vid_out_norm: fusedrms
|
| 15 |
-
txt_in_dim: 5120
|
| 16 |
-
txt_in_norm: fusedln
|
| 17 |
-
txt_dim: ${.vid_dim}
|
| 18 |
-
emb_dim: ${eval:'6 * ${.vid_dim}'}
|
| 19 |
-
heads: 20
|
| 20 |
-
head_dim: 128
|
| 21 |
-
expand_ratio: 4
|
| 22 |
-
norm: fusedrms
|
| 23 |
-
norm_eps: 1e-5
|
| 24 |
-
ada: single
|
| 25 |
-
qk_bias: False
|
| 26 |
-
qk_rope: True
|
| 27 |
-
qk_norm: fusedrms
|
| 28 |
-
patch_size: [1, 2, 2]
|
| 29 |
-
num_layers: 32
|
| 30 |
-
mm_layers: 10
|
| 31 |
-
mlp_type: swiglu
|
| 32 |
-
block_type: mmdit
|
| 33 |
-
msa_type: mm_full
|
| 34 |
-
rope_type: mmrope3d
|
| 35 |
-
rope_dim: 128
|
| 36 |
-
compile: False
|
| 37 |
-
gradient_checkpoint: True
|
| 38 |
-
fsdp:
|
| 39 |
-
sharding_strategy: _HYBRID_SHARD_ZERO2
|
| 40 |
-
ema:
|
| 41 |
-
decay: 0.9998
|
| 42 |
-
|
| 43 |
-
vae:
|
| 44 |
-
model:
|
| 45 |
-
__inherit__: models/vae/s8_c16_t4_inflation_sd3.yaml
|
| 46 |
-
slicing:
|
| 47 |
-
split_size:
|
| 48 |
-
memory_device: same
|
| 49 |
-
checkpoint: ckpt/VINCIE-3B/vae.pth
|
| 50 |
-
scaling_factor: 0.9152
|
| 51 |
-
compile: False
|
| 52 |
-
grouping: False
|
| 53 |
-
dtype: bfloat16
|
| 54 |
-
device: cuda
|
| 55 |
-
|
| 56 |
-
text:
|
| 57 |
-
model:
|
| 58 |
-
__inherit__: models/text/qwen2_vl.yaml
|
| 59 |
-
dtype: bfloat16
|
| 60 |
-
device: cuda
|
| 61 |
-
fsdp:
|
| 62 |
-
enabled: True
|
| 63 |
-
sharding_strategy: _HYBRID_SHARD_ZERO2
|
| 64 |
-
|
| 65 |
-
qwen:
|
| 66 |
-
path: ckpt/VINCIE-3B/llm14b
|
| 67 |
-
|
| 68 |
-
diffusion:
|
| 69 |
-
sampler:
|
| 70 |
-
type: ddim # estava euler; DDIM é mais estável no arranque
|
| 71 |
-
timesteps:
|
| 72 |
-
schedule: uniform
|
| 73 |
-
sampling:
|
| 74 |
-
steps: 40 # 36–40 recomendado
|
| 75 |
-
shift: 12 # 10–12 recomendado
|
| 76 |
-
|
| 77 |
-
data:
|
| 78 |
-
generation:
|
| 79 |
-
resolution: 512 # melhor avaliação que 256
|
| 80 |
-
fps: 2
|
| 81 |
-
batch_size: 1
|
| 82 |
-
sequence_parallel: 1
|
| 83 |
-
positive_prompt:
|
| 84 |
-
image_path: []
|
| 85 |
-
prompts: []
|
| 86 |
-
negative_prompt:
|
| 87 |
-
prompts: []
|
| 88 |
-
output_dir: outputs
|
| 89 |
-
|
| 90 |
-
inference:
|
| 91 |
-
save:
|
| 92 |
-
save_image: True
|
| 93 |
-
save_video: True
|
| 94 |
-
seed: -1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|