Upload 5iarwqz6/config.py with huggingface_hub

5iarwqz6/config.py

@@ -0,0 +1,86 @@
+"""Sweep Phase 1 (dual aug): ttvidt x adaptive_ar (tt1d)"""
+
+import random
+from kohakuengine import Config
+
+SEED = 42
+RUN_ID = "".join(random.choices("abcdefghijklmnopqrstuvwxyz0123456789", k=8))
+RUN_NAME = "H100_ttvidt_adaptive_ar_tt1d_phase1_dual_aug"
+BASE_MODEL_NAME = "facebook/dinov3-vitb16-pretrain-lvd1689m"
+ENCODER_AE_PATH = None
+ENCODER_AE_BACKEND = "latentmaid"
+ENCODER_AE_MODE = "video"
+DECODER_AE_PATH = "latent_models/latentmaid_image_vae/latentmaid_2222_eq.pt"
+DECODER_AE_BACKEND = "image_vae"
+DECODER_AE_MODE = "image"
+LATENT_MEAN = [-0.69, -0.48, -0.60, 0.28]
+LATENT_STD = [12.38, 11.22, 7.93, 21.22]
+T_PATCH = 1
+TOKEN_COUNTS = [1]
+BACKBONE_ARCH = "ttvidt"
+BACKBONE_CONFIG = None
+
+MOTION_ENCODER_CONFIG = {
+    "temporal_patch_size": 1,
+    "temporal_patch_overlap": 0,
+    "motion_layers_period": 1,
+    "num_motion_tokens": 8,
+    "motion_ffn_type": "gelu",
+    "tt_mode": "1d",
+}
+MOTION_DECODER_CONFIG = {
+    "image_dim": 4,
+    "patch_size": 2,
+    "decoder_style": "dit",
+    "num_layers": 12,
+    "hidden_size": 768,
+    "intermediate_size": 3072,
+    "num_heads": 12,
+    "decode_mode": "diffusion",
+    "qk_norm": True,
+    "attn_bias": False,
+    "use_final_norm": False,
+}
+FRAME_COUNT = 8
+SAMPLE_FPS = 6
+DATASET_FOLDERS = [
+    "/mnt/shared/p03/kohaku/openvid384/OpenVid384px-tar",
+    "/mnt/shared/p03/kohaku/moments_in_time/Moments_in_Time_Raw/training-tar",
+]
+TRAIN_MODE = "adaptive_ar"
+AR_SHIFT_RANGE = (1, 3)
+AR_SHIFT_SAMPLING = "gamma"
+BUFFER_FRAMES = 3
+UNFREEZE_BACKBONE = True
+GRAD_CKPT = False
+GRAD_CLIP_VAL = 0.1
+DECODER_PRETRAINED = "KBlueLeaf/dit-testing/pretrain_imgnet_S_qknorm_nofinal"
+
+LEARNING_RATE = 5e-4
+BASE_DIM = 256
+WEIGHT_DECAY = 0.01
+BETAS = (0.9, 0.98)
+BASE_SCHEDULER_CONFIG = {
+    "lr": {
+        "mode": "cosine",
+        "end": -1,
+        "min_value": 0.01,
+        "warmup": 10000,
+    }
+}
+
+GPUS = [0, 1, 2, 3]
+EPOCH = 8
+BATCH_SIZE = 8
+GRAD_ACC = 1
+NUM_WORKERS = 24
+DATASET_TYPE = "tar"
+DUAL_AUG = True
+
+LOG_INTERVAL = 2500
+CKPT_INTERVAL = 5000
+
+
+def config_gen():
+    print("Run with run_id:", RUN_ID)
+    return Config.from_globals()