Delete configs/generate_distributed.yaml

configs/generate_distributed.yaml

@@ -1,94 +0,0 @@
-__object__:
-  path: generate_distributed
-  name: VINCIEGenerator
-
-dit:
-  model:
-    __object__:
-      path: models.dit.nadit
-      name: NaDiT
-      args: as_params
-    vid_in_channels: 33
-    vid_out_channels: 16
-    vid_dim: 2560
-    vid_out_norm: fusedrms
-    txt_in_dim: 5120
-    txt_in_norm: fusedln
-    txt_dim: ${.vid_dim}
-    emb_dim: ${eval:'6 * ${.vid_dim}'}
-    heads: 20
-    head_dim: 128
-    expand_ratio: 4
-    norm: fusedrms
-    norm_eps: 1e-5
-    ada: single
-    qk_bias: False
-    qk_rope: True
-    qk_norm: fusedrms
-    patch_size: [1, 2, 2]
-    num_layers: 32
-    mm_layers: 10
-    mlp_type: swiglu
-    block_type: mmdit
-    msa_type: mm_full
-    rope_type: mmrope3d
-    rope_dim: 128
-    compile: False
-    gradient_checkpoint: True
-  fsdp:
-    sharding_strategy: _HYBRID_SHARD_ZERO2
-  ema:
-    decay: 0.9998
-
-vae:
-  model:
-    __inherit__: models/vae/s8_c16_t4_inflation_sd3.yaml
-  slicing:
-    split_size:
-    memory_device: same
-  checkpoint: ckpt/VINCIE-3B/vae.pth
-  scaling_factor: 0.9152
-  compile: False
-  grouping: False
-  dtype: bfloat16
-  device: cuda
-
-text:
-  model:
-    __inherit__: models/text/qwen2_vl.yaml
-  dtype: bfloat16
-  device: cuda
-  fsdp:
-    enabled: True
-    sharding_strategy: _HYBRID_SHARD_ZERO2
-
-qwen:
-  path: ckpt/VINCIE-3B/llm14b
-
-diffusion:
-  sampler:
-    type: ddim          # estava euler; DDIM é mais estável no arranque
-  timesteps:
-    schedule: uniform
-    sampling:
-      steps: 40         # 36–40 recomendado
-      shift: 12         # 10–12 recomendado
-
-data:
-  generation:
-    resolution: 512     # melhor avaliação que 256
-    fps: 2
-    batch_size: 1
-    sequence_parallel: 1
-  positive_prompt:
-    image_path: []
-    prompts: []
-  negative_prompt:
-    prompts: []
-  output_dir: outputs
-
-inference:
-  save:
-    save_image: True
-    save_video: True
-  seed: -1