Upload 33 files

configs/dit/inference/16x256x256.py

@@ -0,0 +1,31 @@
+num_frames = 16
+fps = 8
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="DiT-XL/2",
+    condition="text",
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=4.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/ucf101_labels.txt"
+save_dir = "./samples/samples/"

configs/dit/inference/1x256x256-class.py

@@ -0,0 +1,31 @@
+num_frames = 1
+fps = 1
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="DiT-XL/2",
+    no_temporal_pos_emb=True,
+    condition="label_1000",
+    from_pretrained="DiT-XL-2-256x256.pt",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="classes",
+    num_classes=1000,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=4.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/imagenet_id.txt"
+save_dir = "./samples/samples/"

configs/dit/inference/1x256x256.py

@@ -0,0 +1,32 @@
+num_frames = 1
+fps = 1
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="DiT-XL/2",
+    no_temporal_pos_emb=True,
+    condition="text",
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=4.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/imagenet_labels.txt"
+save_dir = "./samples/samples/"

configs/dit/train/16x256x256.py

@@ -0,0 +1,50 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="DiT-XL/2",
+    from_pretrained="DiT-XL-2-256x256.pt",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/dit/train/1x256x256.py

@@ -0,0 +1,51 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=1,
+    frame_interval=1,
+    image_size=(256, 256),
+    transform_name="center",
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = False
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="DiT-XL/2",
+    no_temporal_pos_emb=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 128
+lr = 1e-4  # according to DiT repo
+grad_clip = 1.0

configs/latte/inference/16x256x256-class.py

@@ -0,0 +1,30 @@
+num_frames = 16
+fps = 8
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="Latte-XL/2",
+    condition="label_101",
+    from_pretrained="Latte-XL-2-256x256-ucf101.pt",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="classes",
+    num_classes=101,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=4.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/ucf101_id.txt"
+save_dir = "./samples/samples/"

configs/latte/inference/16x256x256.py

@@ -0,0 +1,31 @@
+num_frames = 16
+fps = 8
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="Latte-XL/2",
+    condition="text",
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=4.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/ucf101_labels.txt"
+save_dir = "./samples/samples/"

configs/latte/train/16x256x256.py

@@ -0,0 +1,49 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="Latte-XL/2",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="clip",
+    from_pretrained="openai/clip-vit-base-patch32",
+    model_max_length=77,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/inference/sample-ref.py

@@ -0,0 +1,62 @@
+num_frames = 16
+frame_interval = 3
+fps = 24
+image_size = (240, 426)
+multi_resolution = "STDiT2"
+
+# Condition
+prompt_path = None
+prompt = [
+    "A car driving on the ocean.",
+    'Drone view of waves crashing against the rugged cliffs along Big Sur\'s garay point beach. The crashing blue waters create white-tipped waves, while the golden light of the setting sun illuminates the rocky shore. A small island with a lighthouse sits in the distance, and green shrubbery covers the cliff\'s edge. The steep drop from the road down to the beach is a dramatic feat, with the cliff\'s edges jutting out over the sea. This is a view that captures the raw beauty of the coast and the rugged landscape of the Pacific Coast Highway.{"reference_path": "assets/images/condition/cliff.png", "mask_strategy": "0"}',
+    "In an ornate, historical hall, a massive tidal wave peaks and begins to crash. Two surfers, seizing the moment, skillfully navigate the face of the wave.",
+]
+
+loop = 2
+condition_frame_length = 4
+reference_path = [
+    "https://cdn.openai.com/tmp/s/interp/d0.mp4",
+    None,
+    "assets/images/condition/wave.png",
+]
+# valid when reference_path is not None
+# (loop id, ref id, ref start, length, target start)
+mask_strategy = [
+    "0,0,0,0,8,0.3",
+    None,
+    "0",
+]
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    cache_dir=None,  # "/mnt/hdd/cached_models",
+    micro_batch_size=4,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    cache_dir=None,  # "/mnt/hdd/cached_models",
+    model_max_length=200,
+)
+scheduler = dict(
+    type="iddpm",
+    num_sampling_steps=100,
+    cfg_scale=7.0,
+    cfg_channel=3,  # or None
+)
+dtype = "bf16"
+
+# Others
+batch_size = 1
+seed = 42
+save_dir = "./samples/samples/"

configs/opensora-v1-1/inference/sample.py

@@ -0,0 +1,43 @@
+num_frames = 16
+frame_interval = 3
+fps = 24
+image_size = (240, 426)
+multi_resolution = "STDiT2"
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    cache_dir=None,  # "/mnt/hdd/cached_models",
+    micro_batch_size=4,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    cache_dir=None,  # "/mnt/hdd/cached_models",
+    model_max_length=200,
+)
+scheduler = dict(
+    type="iddpm",
+    num_sampling_steps=100,
+    cfg_scale=7.0,
+    cfg_channel=3,  # or None
+)
+dtype = "bf16"
+
+# Condition
+prompt_path = "./assets/texts/t2v_samples.txt"
+prompt = None  # prompt has higher priority than prompt_path
+
+# Others
+batch_size = 1
+seed = 42
+save_dir = "./samples/samples/"

configs/opensora-v1-1/train/benchmark.py

@@ -0,0 +1,101 @@
+# this file is only for batch size search and is not used for training
+
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+
+# bucket config format:
+# 1. { resolution: {num_frames: (prob, batch_size)} }, in this case batch_size is ignored when searching
+# 2. { resolution: {num_frames: (prob, (max_batch_size, ))} }, batch_size is searched in the range [batch_size_start, max_batch_size), batch_size_start is configured via CLI
+# 3. { resolution: {num_frames: (prob, (min_batch_size, max_batch_size))} }, batch_size is searched in the range [min_batch_size, max_batch_size)
+# 4. { resolution: {num_frames: (prob, (min_batch_size, max_batch_size, step_size))} }, batch_size is searched in the range [min_batch_size, max_batch_size) with step_size (grid search)
+# 5. { resolution: {num_frames: (0.0, None)} }, this bucket will not be used
+
+bucket_config = {
+    # == manual search ==
+    # "240p": {128: (1.0, 2)}, # 4.28s/it
+    # "240p": {64: (1.0, 4)},
+    # "240p": {32: (1.0, 8)},  # 4.6s/it
+    # "240p": {16: (1.0, 16)},  # 4.6s/it
+    # "480p": {16: (1.0, 4)},  # 4.6s/it
+    # "720p": {16: (1.0, 2)},  # 5.89s/it
+    # "256": {1: (1.0, 256)},  # 4.5s/it
+    # "512": {1: (1.0, 96)}, # 4.7s/it
+    # "512": {1: (1.0, 128)}, # 6.3s/it
+    # "480p": {1: (1.0, 50)},  # 4.0s/it
+    # "1024": {1: (1.0, 32)},  # 6.8s/it
+    # "1024": {1: (1.0, 20)}, # 4.3s/it
+    # "1080p": {1: (1.0, 16)}, # 8.6s/it
+    # "1080p": {1: (1.0, 8)},  # 4.4s/it
+    # == stage 2 ==
+    # "240p": {
+    #     16: (1.0, (2, 32)),
+    #     32: (1.0, (2, 16)),
+    #     64: (1.0, (2, 8)),
+    #     128: (1.0, (2, 6)),
+    # },
+    # "256": {1: (1.0, (128, 300))},
+    # "512": {1: (0.5, (64, 128))},
+    # "480p": {1: (0.4, (32, 128)), 16: (0.4, (2, 32)), 32: (0.0, None)},
+    # "720p": {16: (0.1, (2, 16)), 32: (0.0, None)},  # No examples now
+    # "1024": {1: (0.3, (8, 64))},
+    # "1080p": {1: (0.3, (2, 32))},
+    # == stage 3 ==
+    "720p": {1: (20, 40), 32: (0.5, (2, 4)), 64: (0.5, (1, 1))},
+}
+
+
+# Define acceleration
+num_workers = 4
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = None
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/train/image.py

@@ -0,0 +1,65 @@
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+bucket_config = {  # 6s/it
+    "256": {1: (1.0, 256)},
+    "512": {1: (1.0, 80)},
+    "480p": {1: (1.0, 52)},
+    "1024": {1: (1.0, 20)},
+    "1080p": {1: (1.0, 8)},
+}
+
+# Define acceleration
+num_workers = 4
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = 10  # only for logging
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/train/stage1.py

@@ -0,0 +1,77 @@
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+# IMG: 1024 (20%) 512 (30%) 256 (50%) drop (50%)
+bucket_config = {  # 1s/it
+    "144p": {1: (0.5, 48), 16: (1.0, 6), 32: (1.0, 3), 96: (1.0, 1)},
+    "256": {1: (0.5, 24), 16: (0.5, 3), 48: (0.5, 1), 64: (0.0, None)},
+    "240p": {16: (0.3, 2), 32: (0.3, 1), 64: (0.0, None)},
+    "512": {1: (0.4, 12)},
+    "1024": {1: (0.3, 3)},
+}
+mask_ratios = {
+    "mask_no": 0.75,
+    "mask_quarter_random": 0.025,
+    "mask_quarter_head": 0.025,
+    "mask_quarter_tail": 0.025,
+    "mask_quarter_head_tail": 0.05,
+    "mask_image_random": 0.025,
+    "mask_image_head": 0.025,
+    "mask_image_tail": 0.025,
+    "mask_image_head_tail": 0.05,
+}
+
+# Define acceleration
+num_workers = 8
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = False
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = None
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/train/stage2.py

@@ -0,0 +1,79 @@
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+bucket_config = {  # 7s/it
+    "144p": {1: (1.0, 48), 16: (1.0, 17), 32: (1.0, 9), 64: (1.0, 4), 128: (1.0, 1)},
+    "256": {1: (0.8, 254), 16: (0.5, 17), 32: (0.5, 9), 64: (0.5, 4), 128: (0.5, 1)},
+    "240p": {1: (0.1, 20), 16: (0.9, 17), 32: (0.8, 9), 64: (0.8, 4), 128: (0.8, 2)},
+    "512": {1: (0.5, 86), 16: (0.2, 4), 32: (0.2, 2), 64: (0.2, 1), 128: (0.0, None)},
+    "480p": {1: (0.4, 54), 16: (0.4, 4), 32: (0.0, None)},
+    "720p": {1: (0.1, 20), 16: (0.1, 2), 32: (0.0, None)},
+    "1024": {1: (0.3, 20)},
+    "1080p": {1: (0.4, 8)},
+}
+mask_ratios = {
+    "mask_no": 0.75,
+    "mask_quarter_random": 0.025,
+    "mask_quarter_head": 0.025,
+    "mask_quarter_tail": 0.025,
+    "mask_quarter_head_tail": 0.05,
+    "mask_image_random": 0.025,
+    "mask_image_head": 0.025,
+    "mask_image_tail": 0.025,
+    "mask_image_head_tail": 0.05,
+}
+
+# Define acceleration
+num_workers = 8
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = None
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/train/stage3.py

@@ -0,0 +1,79 @@
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+bucket_config = {  # 13s/it
+    "144p": {1: (1.0, 200), 16: (1.0, 36), 32: (1.0, 18), 64: (1.0, 9), 128: (1.0, 4)},
+    "256": {1: (0.8, 200), 16: (0.5, 22), 32: (0.5, 11), 64: (0.5, 6), 128: (0.8, 4)},
+    "240p": {1: (0.8, 200), 16: (0.5, 22), 32: (0.5, 10), 64: (0.5, 6), 128: (0.5, 3)},
+    "360p": {1: (0.5, 120), 16: (0.5, 9), 32: (0.5, 4), 64: (0.5, 2), 128: (0.5, 1)},
+    "512": {1: (0.5, 120), 16: (0.5, 9), 32: (0.5, 4), 64: (0.5, 2), 128: (0.8, 1)},
+    "480p": {1: (0.4, 80), 16: (0.6, 6), 32: (0.6, 3), 64: (0.6, 1), 128: (0.0, None)},
+    "720p": {1: (0.4, 40), 16: (0.6, 3), 32: (0.6, 1), 96: (0.0, None)},
+    "1024": {1: (0.3, 40)},
+}
+mask_ratios = {
+    "mask_no": 0.75,
+    "mask_quarter_random": 0.025,
+    "mask_quarter_head": 0.025,
+    "mask_quarter_tail": 0.025,
+    "mask_quarter_head_tail": 0.05,
+    "mask_image_random": 0.025,
+    "mask_image_head": 0.025,
+    "mask_image_tail": 0.025,
+    "mask_image_head_tail": 0.05,
+}
+
+# Define acceleration
+num_workers = 8
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = None
+lr = 2e-5
+grad_clip = 1.0

configs/opensora-v1-1/train/video.py

@@ -0,0 +1,67 @@
+# Define dataset
+dataset = dict(
+    type="VariableVideoTextDataset",
+    data_path=None,
+    num_frames=None,
+    frame_interval=3,
+    image_size=(None, None),
+    transform_name="resize_crop",
+)
+bucket_config = {  # 6s/it
+    "240p": {16: (1.0, 16), 32: (1.0, 8), 64: (1.0, 4), 128: (1.0, 2)},
+    "256": {1: (1.0, 256)},
+    "512": {1: (0.5, 80)},
+    "480p": {1: (0.4, 52), 16: (0.4, 4), 32: (0.0, None)},
+    "720p": {16: (0.1, 2), 32: (0.0, None)},  # No examples now
+    "1024": {1: (0.3, 20)},
+    "1080p": {1: (0.3, 8)},
+}
+
+# Define acceleration
+num_workers = 4
+num_bucket_build_workers = 16
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT2-XL/2",
+    from_pretrained=None,
+    input_sq_size=512,  # pretrained model is trained on 512x512
+    qk_norm=True,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+    local_files_only=True,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=200,
+    shardformer=True,
+    local_files_only=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = 10  # only for logging
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/inference/16x256x256.py

@@ -0,0 +1,39 @@
+num_frames = 16
+fps = 24 // 3
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=4,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="iddpm",
+    num_sampling_steps=100,
+    cfg_scale=7.0,
+    cfg_channel=3,  # or None
+)
+dtype = "bf16"
+
+# Condition
+prompt_path = "./assets/texts/t2v_samples.txt"
+prompt = None  # prompt has higher priority than prompt_path
+
+# Others
+batch_size = 1
+seed = 42
+save_dir = "./samples/samples/"

configs/opensora/inference/16x512x512.py

@@ -0,0 +1,35 @@
+num_frames = 16
+fps = 24 // 3
+image_size = (512, 512)
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=1.0,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=2,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="iddpm",
+    num_sampling_steps=100,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/t2v_samples.txt"
+save_dir = "./samples/samples/"

configs/opensora/inference/64x512x512.py

@@ -0,0 +1,35 @@
+num_frames = 64
+fps = 24 // 2
+image_size = (512, 512)
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=2 / 3,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+    from_pretrained="PRETRAINED_MODEL",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=128,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="iddpm",
+    num_sampling_steps=100,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 1
+seed = 42
+prompt_path = "./assets/texts/t2v_samples.txt"
+save_dir = "./samples/samples/"

configs/opensora/train/16x256x256-mask.py

@@ -0,0 +1,60 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+mask_ratios = {
+    "mask_no": 0.7,
+    "mask_random": 0.15,
+    "mask_head": 0.05,
+    "mask_tail": 0.05,
+    "mask_head_tail": 0.05,
+}
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/16x256x256-spee.py

@@ -0,0 +1,60 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+mask_ratios = {
+    "mask_no": 0.5,
+    "mask_random": 0.29,
+    "mask_head": 0.07,
+    "mask_tail": 0.07,
+    "mask_head_tail": 0.07,
+}
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm-speed",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/16x256x256.py

@@ -0,0 +1,53 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/16x512x512.py

@@ -0,0 +1,54 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(512, 512),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=1.0,
+    from_pretrained=None,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=128,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 500
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/360x512x512.py

@@ -0,0 +1,61 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=360,
+    frame_interval=3,
+    image_size=(512, 512),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define acceleration
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2-seq"
+sp_size = 2
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=2 / 3,
+    from_pretrained=None,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+    enable_sequence_parallelism=True,  # enable sq here
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=128,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 250
+load = None
+
+batch_size = 1
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/64x512x512-sp.py

@@ -0,0 +1,54 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(512, 512),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 2
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=2 / 3,
+    from_pretrained=None,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+    enable_sequence_parallelism=True,  # enable sq here
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 1
+lr = 2e-5
+grad_clip = 1.0

configs/opensora/train/64x512x512.py

@@ -0,0 +1,54 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=64,
+    frame_interval=3,
+    image_size=(512, 512),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="STDiT-XL/2",
+    space_scale=1.0,
+    time_scale=2 / 3,
+    from_pretrained=None,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=64,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 250
+load = None
+
+batch_size = 4
+lr = 2e-5
+grad_clip = 1.0

configs/pixart/inference/16x256x256.py

@@ -0,0 +1,32 @@
+num_frames = 16
+fps = 8
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    from_pretrained="outputs/098-F16S3-PixArt-XL-2/epoch7-global_step30000/model_ckpt.pt",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/t2v_samples.txt"
+save_dir = "./samples/samples/"

configs/pixart/inference/1x1024MS.py

@@ -0,0 +1,34 @@
+num_frames = 1
+fps = 1
+image_size = (1920, 512)
+multi_resolution = "PixArtMS"
+
+# Define model
+model = dict(
+    type="PixArtMS-XL/2",
+    space_scale=2.0,
+    time_scale=1.0,
+    no_temporal_pos_emb=True,
+    from_pretrained="PixArt-XL-2-1024-MS.pth",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/t2i_samples.txt"
+save_dir = "./samples/samples/"

configs/pixart/inference/1x256x256.py

@@ -0,0 +1,33 @@
+num_frames = 1
+fps = 1
+image_size = (256, 256)
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=1.0,
+    time_scale=1.0,
+    no_temporal_pos_emb=True,
+    from_pretrained="PixArt-XL-2-256x256.pth",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# Others
+batch_size = 2
+seed = 42
+prompt_path = "./assets/texts/t2i_samples.txt"
+save_dir = "./samples/samples/"

configs/pixart/inference/1x512x512.py

@@ -0,0 +1,39 @@
+num_frames = 1
+fps = 1
+image_size = (512, 512)
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=1.0,
+    time_scale=1.0,
+    no_temporal_pos_emb=True,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+)
+scheduler = dict(
+    type="dpm-solver",
+    num_sampling_steps=20,
+    cfg_scale=7.0,
+)
+dtype = "bf16"
+
+# prompt_path = "./assets/texts/t2i_samples.txt"
+prompt = [
+    "Pirate ship trapped in a cosmic maelstrom nebula.",
+    "A small cactus with a happy face in the Sahara desert.",
+    "A small cactus with a sad face in the Sahara desert.",
+]
+
+# Others
+batch_size = 2
+seed = 42
+save_dir = "./samples/samples/"

configs/pixart/train/16x256x256.py

@@ -0,0 +1,53 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=16,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=0.5,
+    time_scale=1.0,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 8
+lr = 2e-5
+grad_clip = 1.0

configs/pixart/train/1x512x512.py

@@ -0,0 +1,54 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=1,
+    frame_interval=3,
+    image_size=(512, 512),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=1.0,
+    time_scale=1.0,
+    no_temporal_pos_emb=True,
+    from_pretrained="PixArt-XL-2-512x512.pth",
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 1000
+load = None
+
+batch_size = 32
+lr = 2e-5
+grad_clip = 1.0

configs/pixart/train/64x512x512.py

@@ -0,0 +1,55 @@
+# Define dataset
+dataset = dict(
+    type="VideoTextDataset",
+    data_path=None,
+    num_frames=64,
+    frame_interval=3,
+    image_size=(256, 256),
+)
+
+# Define acceleration
+num_workers = 4
+dtype = "bf16"
+grad_checkpoint = True
+plugin = "zero2"
+sp_size = 1
+
+
+# Define model
+model = dict(
+    type="PixArt-XL/2",
+    space_scale=1.0,
+    time_scale=2 / 3,
+    from_pretrained=None,
+    enable_flashattn=True,
+    enable_layernorm_kernel=True,
+)
+vae = dict(
+    type="VideoAutoencoderKL",
+    from_pretrained="stabilityai/sd-vae-ft-ema",
+    micro_batch_size=128,
+)
+text_encoder = dict(
+    type="t5",
+    from_pretrained="DeepFloyd/t5-v1_1-xxl",
+    model_max_length=120,
+    shardformer=True,
+)
+scheduler = dict(
+    type="iddpm",
+    timestep_respacing="",
+)
+
+# Others
+seed = 42
+outputs = "outputs"
+wandb = False
+
+epochs = 1000
+log_every = 10
+ckpt_every = 250
+load = None
+
+batch_size = 4
+lr = 2e-5
+grad_clip = 1.0