MCPcopy Create free account

hub / github.com/character-ai/Ovi / functions

Functions478 in github.com/character-ai/Ovi

⨍ Functions 478◇ Types & classes 112
↓ 11 callersMethod__init__
( self, in_channels: int, out_channels: int, factor_t, factor_s=1,
ovi/modules/vae2_2.py:372
↓ 10 callersMethod_sigma_to_alpha_sigma_t
(self, sigma)
ovi/utils/fm_solvers.py:333
↓ 10 callersFunctionall_to_all_4D
( input_: torch.Tensor, scatter_dim: int = 2, gather_dim: int = 1, )
ovi/distributed_comms/communications.py:132
↓ 10 callersFunctionflash_attention
q: [B, Lq, Nq, C1]. k: [B, Lk, Nk, C1]. v: [B, Lk, Nk, C2]. Nq must be divisible by Nk. q_lens
ovi/modules/attention.py:25
↓ 9 callersMethod__init__
(self, dim, out_dim, patch_size, eps=1e-6)
ovi/modules/model.py:476
↓ 9 callersMethoddevice
(self)
ovi/modules/mmaudio/ext/autoencoder/vae.py:156
↓ 8 callersMethod__init__
(self, vocab_size, dim, dim_attn, dim_ffn,
ovi/modules/t5.py:374
↓ 8 callersMethod__init__
(self, z_dim=16, vae_pth='cache/vae_step_411000.pth', dtype
ovi/modules/vae.py:645
↓ 8 callersMethod_sigma_to_alpha_sigma_t
(self, sigma)
ovi/utils/fm_solvers_unipc.py:272
↓ 8 callersFunctionget_padding
(kernel_size, dilation=1)
ovi/modules/mmaudio/ext/bigvgan/utils.py:22
↓ 8 callersMethodoffload_to_cpu
(self, model)
ovi/ovi_fusion_engine.py:357
↓ 6 callersMethod__init__
(self, dim, mid_dim)
ovi/modules/clip.py:97
↓ 6 callersFunctionhalf
(x)
ovi/modules/attention.py:60
↓ 5 callersFunctionfp16_clamp
(x)
ovi/modules/t5.py:20
↓ 5 callersFunctionget_sequence_parallel_state
()
ovi/distributed_comms/parallel_states.py:37
↓ 4 callersFunctionall_to_all
( input_: torch.Tensor, scatter_dim: int = 2, gather_dim: int = 1, )
ovi/distributed_comms/communications.py:194
↓ 4 callersFunctionceil_by_factor
Returns the smallest integer greater than or equal to 'number' that is divisible by 'factor'.
ovi/utils/qwen_vl_utils.py:44
↓ 4 callersMethodclear_cache
(self)
ovi/modules/vae.py:606
↓ 4 callersMethodclear_cache
(self)
ovi/modules/vae2_2.py:853
↓ 4 callersMethodencode
(self, x, scale)
ovi/modules/vae2_2.py:783
↓ 4 callersMethodgenerate
(self, text_prompt, image_path=None, video_frame_
ovi/ovi_fusion_engine.py:144
↓ 4 callersFunctionnonlinearity
(x)
ovi/modules/mmaudio/ext/autoencoder/vae_modules.py:9
↓ 4 callersFunctionnormalize
(x, dim=None, eps=1e-4)
ovi/modules/mmaudio/ext/autoencoder/edm2_utils.py:55
↓ 4 callersMethodnormalize
(self, x: torch.Tensor)
ovi/modules/mmaudio/ext/autoencoder/vae.py:129
↓ 4 callersFunctionrope_apply
(x, grid_sizes, freqs)
ovi/modules/model.py:103
↓ 4 callersFunctionrope_params
(max_seq_len, dim, theta=10000, freqs_scaling=1.0)
ovi/modules/model.py:38
↓ 4 callersFunctionround_by_factor
Returns the closest integer to 'number' that is divisible by 'factor'.
ovi/utils/qwen_vl_utils.py:39
↓ 4 callersMethodset_timesteps
Sets the discrete timesteps used for the diffusion chain (to be run before inference). Args: num_inference_steps (`int`):
ovi/utils/fm_solvers.py:226
↓ 4 callersFunctionsmart_resize
Rescales the image so that the following conditions are met: 1. Both dimensions (height and width) are divisible by 'factor'. 2. The to
ovi/utils/qwen_vl_utils.py:54
↓ 3 callersMethod__init__
(self, in_channels, with_conv)
ovi/modules/mmaudio/ext/autoencoder/vae_modules.py:84
↓ 3 callersFunctionbasic_clean
(text)
ovi/modules/tokenizers.py:12
↓ 3 callersMethoddecode
(self, z, scale)
ovi/modules/vae2_2.py:812
↓ 3 callersFunctionfloor_by_factor
Returns the largest integer less than or equal to 'number' that is divisible by 'factor'.
ovi/utils/qwen_vl_utils.py:49
↓ 3 callersFunctionget_global_rank
Get the global rank, the global index of the GPU.
ovi/distributed_comms/util.py:7
↓ 3 callersFunctionget_padding
(kernel_size, dilation=1)
ovi/modules/mmaudio/ext/bigvgan_v2/utils.py:22
↓ 3 callersFunctionget_world_size
Get the world size, the total amount of GPUs.
ovi/distributed_comms/util.py:21
↓ 3 callersFunctionhalf
(x)
ovi/distributed_comms/distributed/xdit_context_parallel.py:158
↓ 3 callersMethodpost_transformer_block_out
(self, x, grid_sizes, e)
ovi/modules/model.py:806
↓ 3 callersMethodprepare_transformer_block_kwargs
( self, x, t, context, seq_len, clip_fea=None, y=None,
ovi/modules/model.py:692
↓ 3 callersMethodqkv_fn
(self, x)
ovi/modules/model.py:231
↓ 3 callersFunctiontimed_download
Download files from HF repo and log time + destination.
download_weights.py:20
↓ 2 callersMethod__init__
(self, vocab_size=250002, max_seq_len=514, type_size=1,
ovi/modules/xlm_roberta.py:81
↓ 2 callersMethod__init__
( self, *, data_dim: int, embed_dim: int, hidden_dim: int, )
ovi/modules/mmaudio/ext/autoencoder/vae.py:68
↓ 2 callersMethod__init__
(self, h: AttrDict, use_cuda_kernel: bool = False)
ovi/modules/mmaudio/ext/bigvgan_v2/bigvgan.py:232
↓ 2 callersMethod__init__
(self, h, channels, kernel_size=3, dilation=(1, 3, 5), activation=None)
ovi/modules/mmaudio/ext/bigvgan/models.py:22
↓ 2 callersMethod__init__
(self, model_name, is_vl=False, device=0, **kwargs)
ovi/utils/prompt_extend.py:114
↓ 2 callersFunction_all_to_all
( input_: torch.Tensor, world_size: int, group: dist.ProcessGroup, scatter_dim: int, gathe
ovi/distributed_comms/communications.py:141
↓ 2 callersFunction_n32
(x: int)
ovi/utils/processing_utils.py:165
↓ 2 callersMethod_threshold_sample
"Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the prediction of x_0 at t
ovi/utils/fm_solvers.py:292
↓ 2 callersMethod_threshold_sample
"Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the prediction of x_0 at t
ovi/utils/fm_solvers_unipc.py:230
↓ 2 callersFunctionclean_text
Remove all text between <S>...</E> and <AUDCAP>...</ENDAUDCAP> tags, as well as any 'Audio: ...' lines. Also trims excess whitespace.
ovi/utils/processing_utils.py:290
↓ 2 callersFunctioncount_conv3d
(model)
ovi/modules/vae.py:475
↓ 2 callersFunctioncount_conv3d
(model)
ovi/modules/vae2_2.py:726
↓ 2 callersMethoddecode
(self, zs)
ovi/modules/vae.py:681
↓ 2 callersMethodencode
(self, x, scale)
ovi/modules/vae.py:516
↓ 2 callersMethodencode
(self, x: torch.Tensor, normalize: bool = True)
ovi/modules/mmaudio/ext/autoencoder/vae.py:116
↓ 2 callersMethodextend
(self, prompt, system_prompt, seed=-1, *args, **kwargs)
ovi/utils/prompt_extend.py:128
↓ 2 callersFunctionfetch_image
(ele: dict[str, str | Image.Image], size_factor: int = IMAGE_FACTOR)
ovi/utils/qwen_vl_utils.py:85
↓ 2 callersMethodforward
(self, x)
ovi/modules/clip.py:107
↓ 2 callersMethodforward
(self, x)
ovi/modules/vae.py:510
↓ 2 callersMethodforward
(self, x: torch.Tensor, first_chunk=False)
ovi/modules/vae2_2.py:390
↓ 2 callersMethodforward
r""" Args: x(Tensor): Shape [B, L1, C] e(Tensor): Shape [B, L, C]
ovi/modules/model.py:491
↓ 2 callersFunctionget_local_rank
Get the local rank, the local index of the GPU.
ovi/distributed_comms/util.py:14
↓ 2 callersFunctionget_my_vae
(name: str, **kwargs)
ovi/modules/mmaudio/ext/autoencoder/vae.py:356
↓ 2 callersMethodget_scheduler_time_steps
(self, sampling_steps, solver_name='unipc', device=0, shift=5.0)
ovi/ovi_fusion_engine.py:365
↓ 2 callersFunctiongradient_checkpointing
(module: nn.Module, *args, enabled: bool, **kwargs)
ovi/modules/model.py:17
↓ 2 callersMethodindex_for_timestep
(self, timestep, schedule_timesteps=None)
ovi/utils/fm_solvers.py:679
↓ 2 callersMethodindex_for_timestep
(self, timestep, schedule_timesteps=None)
ovi/utils/fm_solvers_unipc.py:628
↓ 2 callersFunctionkaiser_sinc_filter1d
( cutoff, half_width, kernel_size )
ovi/modules/mmaudio/ext/bigvgan_v2/alias_free_activation/torch/filter.py:30
↓ 2 callersFunctionkaiser_sinc_filter1d
(cutoff, half_width, kernel_size)
ovi/modules/mmaudio/ext/bigvgan/alias_free_torch/filter.py:28
↓ 2 callersFunctionmp_silu
(x)
ovi/modules/mmaudio/ext/autoencoder/edm2_utils.py:106
↓ 2 callersFunctionmp_sum
(a, b, t=0.5)
ovi/modules/mmaudio/ext/autoencoder/edm2_utils.py:120
↓ 2 callersFunctionpreprocess_image_tensor
Preprocess video data into standardized tensor format and (optionally) resize area.
ovi/utils/processing_utils.py:15
↓ 2 callersMethodremove_weight_norm
(self)
ovi/modules/mmaudio/ext/autoencoder/vae.py:162
↓ 2 callersFunctionretrieve_timesteps
( scheduler, num_inference_steps=None, device=None, timesteps=None, sigmas=None, **kwa
ovi/utils/fm_solvers.py:29
↓ 2 callersFunctionrope_apply
x: [B, L, N, C]. grid_sizes: [B, 3]. freqs: [M, C // 2].
ovi/distributed_comms/distributed/xdit_context_parallel.py:26
↓ 2 callersMethodsample
(self, rng: Optional[torch.Generator] = None)
ovi/modules/mmaudio/ext/autoencoder/distributions.py:18
↓ 2 callersFunctionsave_video
Combine a sequence of video frames with an optional audio track and save as an MP4. Args: output_path (str): Path to the output MP4
ovi/utils/io_utils.py:9
↓ 2 callersFunctionscale_hw_to_area_divisible
Scale (h, w) so that area ≈ A, while keeping aspect ratio, and then round so both are divisible by n. Args: h (int): origina
ovi/utils/processing_utils.py:169
↓ 2 callersMethodset_rope_params
(self)
ovi/modules/model.py:669
↓ 2 callersMethodsingle_fusion_cross_attention_ffn_forward
(self, attn_block, src
ovi/modules/fusion.py:132
↓ 2 callersFunctionsinusoidal_embedding_1d
(dim, position)
ovi/modules/model.py:24
↓ 2 callersFunctionsmart_nframes
calculate the number of frames for video used for model inputs. Args: ele (dict): a dict contains the configuration of video.
ovi/utils/qwen_vl_utils.py:133
↓ 2 callersMethodstep
Predict the sample from the previous timestep by reversing the SDE. This function propagates the sample with the multistep DPMSolver.
ovi/utils/fm_solvers.py:706
↓ 2 callersMethodunpatchify
r""" Reconstruct video tensors from patch embeddings. Args: x (List[Tensor]): List of patchified features
ovi/modules/model.py:880
↓ 2 callersFunctionwhitespace_clean
(text)
ovi/modules/tokenizers.py:18
↓ 2 callersMethodwrapped_decode
(self, z)
ovi/modules/vae.py:690
↓ 2 callersMethodwrapped_encode
(self, audio)
ovi/modules/mmaudio/features_utils.py:95
↓ 1 callersFunctionVAE_16k
(**kwargs)
ovi/modules/mmaudio/ext/autoencoder/vae.py:348
↓ 1 callersFunctionVAE_44k
(**kwargs)
ovi/modules/mmaudio/ext/autoencoder/vae.py:352
↓ 1 callersMethod__init__
(self, in_channels, out_channels, kernel_size)
ovi/modules/mmaudio/ext/autoencoder/edm2_utils.py:144
↓ 1 callersMethod__init__
Initialization. INPUT: - in_features: shape of the input - alpha: trainable parameter alpha is in
ovi/modules/mmaudio/ext/bigvgan_v2/activations.py:26
↓ 1 callersMethod__init__
(self, ratio=2, kernel_size=None)
ovi/modules/mmaudio/ext/bigvgan_v2/alias_free_activation/torch/resample.py:12
↓ 1 callersMethod__init__
Initialization. INPUT: - in_features: shape of the input - alpha: trainable parameter alpha is in
ovi/modules/mmaudio/ext/bigvgan/activations.py:25
↓ 1 callersMethod__init__
(self, ratio=2, kernel_size=None)
ovi/modules/mmaudio/ext/bigvgan/alias_free_torch/resample.py:11
↓ 1 callersFunction_all_to_all_4D
all-to-all for QKV Args: input (torch.tensor): a tensor sharded along dim scatter dim scatter_idx (int): default 1 g
ovi/distributed_comms/communications.py:18
↓ 1 callersFunction_best_hw_for_area
(h, w, area_target, multiple)
ovi/utils/processing_utils.py:32
↓ 1 callersMethod_clean
(self, text)
ovi/modules/tokenizers.py:75
↓ 1 callersFunction_clip
(pretrained=False, pretrained_name=None, model_cls=XLMRobertaCLIP, return_transf
ovi/modules/clip.py:437
↓ 1 callersFunction_cpp_extention_load_helper
(name, sources, extra_cuda_flags)
ovi/modules/mmaudio/ext/bigvgan_v2/alias_free_activation/cuda/load.py:31
next →1–100 of 478, ranked by callers