Functions — github.com/black-forest-labs/flux

hub / github.com/black-forest-labs/flux / functions

Functions217 in github.com/black-forest-labs/flux

⨍ Functions 217◇ Types & classes 54

↓ 97 callersMethodto

(self, device: str | torch.device)

src/flux/trt/engine/base_engine.py:182

↓ 43 callersMethodcpu

(self)

src/flux/trt/engine/base_engine.py:160

↓ 8 callersMethod__init__

(self, dim: int)

src/flux/modules/layers.py:76

↓ 8 callersMethoddecode

(self, z: torch.Tensor)

src/flux/trt/engine/vae_engine.py:64

↓ 8 callersFunctiondenoise

( model: Flux, # model input img: Tensor, img_ids: Tensor, txt: Tensor, txt_ids: Tenso

src/flux/sampling.py:308

↓ 8 callersFunctionget_noise

( num_samples: int, height: int, width: int, device: torch.device, dtype: torch.dtype,

src/flux/sampling.py:17

↓ 8 callersFunctionget_schedule

( num_steps: int, image_seq_len: int, base_shift: float = 0.5, max_shift: float = 1.15, sh

src/flux/sampling.py:289

↓ 8 callersFunctionload_ae

(name: str, device: str | torch.device = "cuda")

src/flux/util.py:698

↓ 8 callersFunctionload_clip

(device: str | torch.device = "cuda")

src/flux/util.py:694

↓ 8 callersFunctionload_flow_model

(name: str, device: str | torch.device = "cuda", verbose: bool = True)

src/flux/util.py:657

↓ 8 callersFunctionload_t5

(device: str | torch.device = "cuda", max_length: int = 512)

src/flux/util.py:689

↓ 8 callersMethodresize

(self, name: str, size: int)

src/flux/trt/engine/base_engine.py:48

↓ 8 callersFunctionunpack

(x: Tensor, height: int, width: int)

src/flux/sampling.py:356

↓ 7 callersMethod__init__

( self, resolution: int, in_channels: int, ch: int, ch_mult: list[int]

src/flux/modules/autoencoder.py:110

↓ 7 callersMethod_get_context_dim

( image_height: int, image_width: int, compression_factor: int, )

src/flux/trt/trt_config/transformer_trt_config.py:103

↓ 7 callersMethodencode

(self, x: torch.Tensor)

src/flux/trt/engine/vae_engine.py:67

↓ 6 callersMethod_get_latent_dim

(self, image_dim: int)

src/flux/trt/trt_config/vae_trt_config.py:38

↓ 6 callersFunctionprepare

(t5: HFEmbedder, clip: HFEmbedder, img: Tensor, prompt: str | list[str])

src/flux/sampling.py:36

↓ 6 callersMethodtrt_datatype_to_torch

(datatype)

src/flux/trt/engine/base_engine.py:101

↓ 5 callersFunctionclear_canvas_state

Clear all canvas-related state

demo_st_fill.py:116

↓ 5 callersMethodget_dtype

(self, tensor_name: str)

src/flux/trt/engine/base_engine.py:220

↓ 5 callersMethodinfer

( self, feed_dict: dict[str, torch.Tensor], )

src/flux/trt/engine/base_engine.py:280

↓ 5 callersFunctionparse_img_cond_path

(options: SamplingOptions | None)

src/flux/cli_kontext.py:108

↓ 5 callersFunctionparse_prompt

(options: SamplingOptions)

src/flux/cli_kontext.py:34

↓ 5 callersFunctionsave_image

( nsfw_classifier, name: str, output_name: str, idx: int, x: torch.Tensor, add_samplin

src/flux/util.py:243

↓ 4 callersMethod_check_batch

(self, batch_size)

src/flux/trt/trt_config/base_trt_config.py:213

↓ 4 callersFunctionget_checkpoint_path

Get the local path for a checkpoint file, downloading if necessary.

src/flux/util.py:64

↓ 4 callersFunctionprint_load_warning

(missing: list[str], unexpected: list[str])

src/flux/util.py:646

↓ 4 callersMethodset_scale

(self, scale: float)

src/flux/modules/lora.py:84

↓ 4 callersFunctionswish

(x: Tensor)

src/flux/modules/autoencoder.py:21

↓ 4 callersFunctiontrack_usage_via_api

Track usage of licensed models via the BFL API for commercial licensing compliance. For more information on licensing BFL's models for comme

src/flux/util.py:206

↓ 3 callersMethod_clean_memory

()

src/flux/trt/trt_manager.py:279

↓ 3 callersMethod_get_repo_id

(model_name: str)

src/flux/trt/trt_config/base_trt_config.py:240

↓ 3 callersFunctionattention

(q: Tensor, k: Tensor, v: Tensor, pe: Tensor)

src/flux/math.py:6

↓ 3 callersMethodload_engines

( self, model_name: str, module_names: set[ModuleName], engine_dir: str,

src/flux/trt/trt_manager.py:211

↓ 3 callersFunctionset_new_image

Safely set a new image and clear relevant state

demo_st_fill.py:124

↓ 2 callersMethod__init__

( self, decoder: VAEDecoder, encoder: VAEEncoder | None = None, )

src/flux/trt/engine/vae_engine.py:55

↓ 2 callersMethod_get_latent

(image_dim: int, compression_factor: int)

src/flux/trt/trt_config/transformer_trt_config.py:99

↓ 2 callersFunctionaspect_ratio_to_height_width

(aspect_ratio: str, area: int = 1024**2)

src/flux/util.py:637

↓ 2 callersMethodcalculate_input_hash

(feed_dict: Dict[str, torch.Tensor])

src/flux/trt/engine/base_engine.py:267

↓ 2 callersMethodcheck_dims

( self, batch_size: int, image_height: int, image_width: int, )

src/flux/trt/trt_config/vae_trt_config.py:46

↓ 2 callersFunctioncheck_onnx_access_for_trt

Check ONNX access and download models for TRT - returns ONNX directory path

src/flux/util.py:201

↓ 2 callersFunctionparse_img_cond_path

(options: SamplingOptions | None)

src/flux/cli_fill.py:73

↓ 2 callersFunctionparse_img_cond_path

(options: SamplingOptions | None)

src/flux/cli_control.py:95

↓ 2 callersFunctionparse_img_cond_path

(options: SamplingOptions | None)

src/flux/cli_redux.py:99

↓ 2 callersFunctionparse_img_mask_path

(options: SamplingOptions | None)

src/flux/cli_fill.py:119

↓ 2 callersFunctionparse_lora_scale

(options: SamplingOptions | None)

src/flux/cli_control.py:134

↓ 2 callersFunctionparse_prompt

(options: SamplingOptions)

src/flux/cli.py:35

↓ 2 callersFunctionparse_prompt

(options: SamplingOptions)

src/flux/cli_fill.py:28

↓ 2 callersFunctionparse_prompt

(options: SamplingOptions)

src/flux/cli_control.py:28

↓ 2 callersFunctionparse_prompt

(options: SamplingOptions)

src/flux/cli_redux.py:34

↓ 2 callersFunctionprepare_fill

( t5: HFEmbedder, clip: HFEmbedder, img: Tensor, prompt: str | list[str], ae: AutoEncoder,

src/flux/sampling.py:107

↓ 2 callersMethodstop_runtime

(self)

src/flux/trt/trt_manager.py:290

↓ 2 callersMethodtest_image

(self, image: Image.Image | str | torch.Tensor)

src/flux/content_filters.py:92

↓ 2 callersFunctiontimestep_embedding

Create sinusoidal timestep embeddings. :param t: a 1-D Tensor of N indices, one per batch element. These may be fractio

src/flux/modules/layers.py:28

↓ 2 callersMethodto

(self, device)

src/flux/trt/engine/vae_engine.py:83

↓ 1 callersMethod__init__

(self, params: FluxParams)

src/flux/model.py:39

↓ 1 callersMethod__init__

( self, trt_config: TRTBaseConfig, stream: torch.cuda.Stream, context_memory:

src/flux/trt/engine/base_engine.py:131

↓ 1 callersMethod__init__

(self, device)

src/flux/modules/image_embedders.py:16

↓ 1 callersMethod_build_engine

( trt_config: TRTBaseConfig, batch_size: int, image_height: int | None, image_

src/flux/trt/trt_manager.py:179

↓ 1 callersMethod_create_directories

(engine_dir: str)

src/flux/trt/trt_manager.py:128

↓ 1 callersMethod_get_engine_path

(self)

src/flux/trt/trt_config/base_trt_config.py:232

↓ 1 callersMethod_get_onnx_path

(self)

src/flux/trt/trt_config/base_trt_config.py:223

↓ 1 callersMethod_get_trt_configs

( self, model_name: str, module_names: set[ModuleName], engine_dir: str,

src/flux/trt/trt_manager.py:132

↓ 1 callersMethod_parse_custom_onnx_path

Parse a string of comma-separated key-value pairs into a dictionary. Args: string (str): A string of comma-separated key-value pa

src/flux/trt/trt_manager.py:99

↓ 1 callersMethod_parse_models_precisions

( trt_transformer_precision: str, trt_t5_precision: str )

src/flux/trt/trt_manager.py:76

↓ 1 callersFunctionadd_border_and_mask

Adds a black border around the image with individual side control and mask overlap

demo_st_fill.py:31

↓ 1 callersFunctionapp

()

demo_st.py:292

↓ 1 callersFunctionapp

()

demo_st_fill.py:497

↓ 1 callersFunctionapply_rope

(xq: Tensor, xk: Tensor, freqs_cis: Tensor)

src/flux/math.py:25

↓ 1 callersMethodattention

(self, h_: Tensor)

src/flux/modules/autoencoder.py:37

↓ 1 callersMethodbuild_trt_engine

Metod used to build a TRT engine from a given set of flags or configurations using polygraphy. Args: engine_path (str):

src/flux/trt/trt_config/base_trt_config.py:69

↓ 1 callersMethodcheck_dims

(self, batch_size: int)

src/flux/trt/trt_config/clip_trt_config.py:48

↓ 1 callersMethodcheck_dims

( self, batch_size: int, image_height: int, image_width: int, )

src/flux/trt/trt_config/transformer_trt_config.py:194

↓ 1 callersMethodcheck_dims

(self, batch_size: int)

src/flux/trt/trt_config/t5_trt_config.py:53

↓ 1 callersFunctioncreate_demo

( model_name: str, device: str = "cuda" if torch.cuda.is_available() else "cpu", offload: bool = F

demo_gr.py:175

↓ 1 callersMethodcuda

(self)

src/flux/trt/engine/base_engine.py:171

↓ 1 callersMethoddeactivate

(self)

src/flux/trt/engine/base_engine.py:194

↓ 1 callersMethoddecode

(self, z: Tensor)

src/flux/modules/autoencoder.py:313

↓ 1 callersFunctiondownload_onnx_models_for_trt

Download ONNX models for TRT to our checkpoints directory

src/flux/util.py:108

↓ 1 callersFunctiondownscale_image

Downscale image by a given factor while maintaining 32-pixel multiple dimensions

demo_st_fill.py:131

↓ 1 callersMethodencode

(self, x: Tensor)

src/flux/modules/autoencoder.py:308

↓ 1 callersFunctionensure_hf_auth

()

src/flux/util.py:27

↓ 1 callersMethodfrom_args

( cls, model_name: str, **kwargs, )

src/flux/trt/trt_config/t5_trt_config.py:40

↓ 1 callersFunctionget_config

Retrieve the appropriate configuration instance based on current flags.

src/flux/trt/trt_config/base_trt_config.py:266

↓ 1 callersFunctionget_image

()

demo_st.py:41

↓ 1 callersMethodget_input_profile

( self, batch_size: int, image_height=None, image_width=None, )

src/flux/trt/trt_config/t5_trt_config.py:56

↓ 1 callersFunctionget_lin_function

( x1: float = 256, y1: float = 0.5, x2: float = 4096, y2: float = 1.15 )

src/flux/sampling.py:281

↓ 1 callersMethodget_minmax_dims

( self, batch_size: int, image_height: int, image_width: int, )

src/flux/trt/trt_config/vae_trt_config.py:102

↓ 1 callersMethodget_minmax_dims

( self, batch_size: int, image_height: int, image_width: int, )

src/flux/trt/trt_config/vae_trt_config.py:206

↓ 1 callersMethodget_minmax_dims

( self, batch_size: int, image_height: int, image_width: int, )

src/flux/trt/trt_config/transformer_trt_config.py:160

↓ 1 callersFunctionget_models

(name: str, device: torch.device, offload: bool, is_schnell: bool)

demo_st.py:32

↓ 1 callersFunctionget_models

(name: str, device: torch.device, offload: bool, is_schnell: bool)

demo_gr.py:27

↓ 1 callersFunctionget_models

(name: str, device: torch.device, offload: bool)

demo_st_fill.py:83

↓ 1 callersMethodinit_runtime

(self)

src/flux/trt/trt_manager.py:283

↓ 1 callersFunctionoptionally_expand_state_dict

Optionally expand the state dict to match the model's parameters shapes.

src/flux/util.py:714

↓ 1 callersMethodoverride_shapes

(self, feed_dict: Dict[str, torch.Tensor])

src/flux/trt/engine/base_engine.py:223

↓ 1 callersFunctionprepare_control

( t5: HFEmbedder, clip: HFEmbedder, img: Tensor, prompt: str | list[str], ae: AutoEncoder,

src/flux/sampling.py:70

↓ 1 callersFunctionprepare_kontext

( t5: HFEmbedder, clip: HFEmbedder, prompt: str | list[str], ae: AutoEncoder, img_cond_pat

src/flux/sampling.py:210

↓ 1 callersFunctionprepare_redux

( t5: HFEmbedder, clip: HFEmbedder, img: Tensor, prompt: str | list[str], encoder: ReduxIm

src/flux/sampling.py:160

next →1–100 of 217, ranked by callers