Spaces:
Running
on
Zero
Running
on
Zero
Update
Browse files- app.py +65 -53
- live_preview_helpers.py +0 -166
app.py
CHANGED
@@ -1,50 +1,70 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
3 |
-
import
|
4 |
import spaces
|
5 |
import torch
|
6 |
-
from diffusers import
|
7 |
-
from transformers import CLIPTextModel, CLIPTokenizer,T5EncoderModel, T5TokenizerFast
|
8 |
-
from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
|
9 |
|
10 |
dtype = torch.bfloat16
|
11 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
12 |
|
13 |
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
|
14 |
-
good_vae = AutoencoderKL.from_pretrained("black-forest-labs/FLUX.1-dev", subfolder="vae", torch_dtype=dtype).to(device)
|
15 |
pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=dtype, vae=taef1).to(device)
|
16 |
-
torch.cuda.empty_cache()
|
17 |
|
18 |
MAX_SEED = np.iinfo(np.int32).max
|
19 |
MAX_IMAGE_SIZE = 2048
|
20 |
|
21 |
-
pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)
|
22 |
|
23 |
@spaces.GPU(duration=75)
|
24 |
-
def infer(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
if randomize_seed:
|
26 |
-
seed = random.randint(0, MAX_SEED)
|
27 |
generator = torch.Generator().manual_seed(seed)
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
examples = [
|
42 |
"a tiny astronaut hatching from an egg on the moon",
|
43 |
"a cat holding a sign that says hello world",
|
44 |
"an anime illustration of a wiener schnitzel",
|
45 |
]
|
46 |
|
47 |
-
css="""
|
48 |
#col-container {
|
49 |
margin: 0 auto;
|
50 |
max-width: 520px;
|
@@ -52,29 +72,23 @@ css="""
|
|
52 |
"""
|
53 |
|
54 |
with gr.Blocks(css=css) as demo:
|
55 |
-
|
56 |
with gr.Column(elem_id="col-container"):
|
57 |
-
gr.Markdown(
|
58 |
-
12B param rectified flow transformer guidance-distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/)
|
59 |
[[non-commercial license](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)] [[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)] [[model](https://huggingface.co/black-forest-labs/FLUX.1-dev)]
|
60 |
""")
|
61 |
-
|
62 |
with gr.Row():
|
63 |
-
|
64 |
prompt = gr.Text(
|
65 |
label="Prompt",
|
66 |
show_label=False,
|
67 |
max_lines=1,
|
68 |
placeholder="Enter your prompt",
|
69 |
-
|
70 |
)
|
71 |
-
|
72 |
-
run_button = gr.Button("Run", scale=0)
|
73 |
-
|
74 |
result = gr.Image(label="Result", show_label=False)
|
75 |
-
|
76 |
with gr.Accordion("Advanced Settings", open=False):
|
77 |
-
|
78 |
seed = gr.Slider(
|
79 |
label="Seed",
|
80 |
minimum=0,
|
@@ -82,11 +96,9 @@ with gr.Blocks(css=css) as demo:
|
|
82 |
step=1,
|
83 |
value=0,
|
84 |
)
|
85 |
-
|
86 |
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
87 |
-
|
88 |
with gr.Row():
|
89 |
-
|
90 |
width = gr.Slider(
|
91 |
label="Width",
|
92 |
minimum=256,
|
@@ -94,7 +106,7 @@ with gr.Blocks(css=css) as demo:
|
|
94 |
step=32,
|
95 |
value=1024,
|
96 |
)
|
97 |
-
|
98 |
height = gr.Slider(
|
99 |
label="Height",
|
100 |
minimum=256,
|
@@ -102,9 +114,8 @@ with gr.Blocks(css=css) as demo:
|
|
102 |
step=32,
|
103 |
value=1024,
|
104 |
)
|
105 |
-
|
106 |
-
with gr.Row():
|
107 |
|
|
|
108 |
guidance_scale = gr.Slider(
|
109 |
label="Guidance Scale",
|
110 |
minimum=1,
|
@@ -112,7 +123,7 @@ with gr.Blocks(css=css) as demo:
|
|
112 |
step=0.1,
|
113 |
value=3.5,
|
114 |
)
|
115 |
-
|
116 |
num_inference_steps = gr.Slider(
|
117 |
label="Number of inference steps",
|
118 |
minimum=1,
|
@@ -120,20 +131,21 @@ with gr.Blocks(css=css) as demo:
|
|
120 |
step=1,
|
121 |
value=28,
|
122 |
)
|
123 |
-
|
124 |
gr.Examples(
|
125 |
-
examples
|
126 |
-
fn
|
127 |
-
inputs
|
128 |
-
outputs
|
129 |
-
cache_examples=
|
|
|
130 |
)
|
131 |
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
outputs = [result, seed]
|
137 |
)
|
138 |
|
139 |
-
|
|
|
|
1 |
+
import random
|
2 |
+
|
3 |
import gradio as gr
|
4 |
import numpy as np
|
5 |
+
import PIL.Image
|
6 |
import spaces
|
7 |
import torch
|
8 |
+
from diffusers import AutoencoderTiny, DiffusionPipeline
|
|
|
|
|
9 |
|
10 |
dtype = torch.bfloat16
|
11 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
12 |
|
13 |
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
|
|
|
14 |
pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=dtype, vae=taef1).to(device)
|
|
|
15 |
|
16 |
MAX_SEED = np.iinfo(np.int32).max
|
17 |
MAX_IMAGE_SIZE = 2048
|
18 |
|
|
|
19 |
|
20 |
@spaces.GPU(duration=75)
|
21 |
+
def infer(
|
22 |
+
prompt: str,
|
23 |
+
seed: int = 42,
|
24 |
+
randomize_seed: bool = False,
|
25 |
+
width: int = 1024,
|
26 |
+
height: int = 1024,
|
27 |
+
guidance_scale: float = 3.5,
|
28 |
+
num_inference_steps: int = 28,
|
29 |
+
progress: gr.Progress = gr.Progress(track_tqdm=True), # noqa: ARG001, B008
|
30 |
+
) -> tuple[PIL.Image.Image, int]:
|
31 |
+
"""Generate an image from a prompt using the Flux.1 [dev] model.
|
32 |
+
|
33 |
+
Args:
|
34 |
+
prompt: The prompt to generate an image from.
|
35 |
+
seed: The seed to use for the image generation. Defaults to 42.
|
36 |
+
randomize_seed: Whether to randomize the seed. Defaults to False.
|
37 |
+
width: The width of the image. Defaults to 1024.
|
38 |
+
height: The height of the image. Defaults to 1024.
|
39 |
+
guidance_scale: The guidance scale to use for the image generation. Defaults to 3.5.
|
40 |
+
num_inference_steps: The number of inference steps to use for the image generation. Defaults to 28.
|
41 |
+
progress: The progress bar to use for the image generation. Defaults to a progress bar that tracks the tqdm progress.
|
42 |
+
|
43 |
+
Returns:
|
44 |
+
A tuple containing the generated image and the seed.
|
45 |
+
"""
|
46 |
if randomize_seed:
|
47 |
+
seed = random.randint(0, MAX_SEED) # noqa: S311
|
48 |
generator = torch.Generator().manual_seed(seed)
|
49 |
+
|
50 |
+
image = pipe(
|
51 |
+
prompt=prompt,
|
52 |
+
width=width,
|
53 |
+
height=height,
|
54 |
+
num_inference_steps=num_inference_steps,
|
55 |
+
generator=generator,
|
56 |
+
guidance_scale=guidance_scale,
|
57 |
+
).images[0]
|
58 |
+
return image, seed
|
59 |
+
|
60 |
+
|
|
|
61 |
examples = [
|
62 |
"a tiny astronaut hatching from an egg on the moon",
|
63 |
"a cat holding a sign that says hello world",
|
64 |
"an anime illustration of a wiener schnitzel",
|
65 |
]
|
66 |
|
67 |
+
css = """
|
68 |
#col-container {
|
69 |
margin: 0 auto;
|
70 |
max-width: 520px;
|
|
|
72 |
"""
|
73 |
|
74 |
with gr.Blocks(css=css) as demo:
|
|
|
75 |
with gr.Column(elem_id="col-container"):
|
76 |
+
gr.Markdown("""# FLUX.1 [dev]
|
77 |
+
12B param rectified flow transformer guidance-distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/)
|
78 |
[[non-commercial license](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)] [[blog](https://blackforestlabs.ai/announcing-black-forest-labs/)] [[model](https://huggingface.co/black-forest-labs/FLUX.1-dev)]
|
79 |
""")
|
80 |
+
|
81 |
with gr.Row():
|
|
|
82 |
prompt = gr.Text(
|
83 |
label="Prompt",
|
84 |
show_label=False,
|
85 |
max_lines=1,
|
86 |
placeholder="Enter your prompt",
|
87 |
+
submit_btn=True,
|
88 |
)
|
|
|
|
|
|
|
89 |
result = gr.Image(label="Result", show_label=False)
|
90 |
+
|
91 |
with gr.Accordion("Advanced Settings", open=False):
|
|
|
92 |
seed = gr.Slider(
|
93 |
label="Seed",
|
94 |
minimum=0,
|
|
|
96 |
step=1,
|
97 |
value=0,
|
98 |
)
|
|
|
99 |
randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
|
100 |
+
|
101 |
with gr.Row():
|
|
|
102 |
width = gr.Slider(
|
103 |
label="Width",
|
104 |
minimum=256,
|
|
|
106 |
step=32,
|
107 |
value=1024,
|
108 |
)
|
109 |
+
|
110 |
height = gr.Slider(
|
111 |
label="Height",
|
112 |
minimum=256,
|
|
|
114 |
step=32,
|
115 |
value=1024,
|
116 |
)
|
|
|
|
|
117 |
|
118 |
+
with gr.Row():
|
119 |
guidance_scale = gr.Slider(
|
120 |
label="Guidance Scale",
|
121 |
minimum=1,
|
|
|
123 |
step=0.1,
|
124 |
value=3.5,
|
125 |
)
|
126 |
+
|
127 |
num_inference_steps = gr.Slider(
|
128 |
label="Number of inference steps",
|
129 |
minimum=1,
|
|
|
131 |
step=1,
|
132 |
value=28,
|
133 |
)
|
134 |
+
|
135 |
gr.Examples(
|
136 |
+
examples=examples,
|
137 |
+
fn=infer,
|
138 |
+
inputs=prompt,
|
139 |
+
outputs=[result, seed],
|
140 |
+
cache_examples=True,
|
141 |
+
cache_mode="lazy",
|
142 |
)
|
143 |
|
144 |
+
prompt.submit(
|
145 |
+
fn=infer,
|
146 |
+
inputs=[prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
|
147 |
+
outputs=[result, seed],
|
|
|
148 |
)
|
149 |
|
150 |
+
if __name__ == "__main__":
|
151 |
+
demo.launch(mcp_server=True)
|
live_preview_helpers.py
DELETED
@@ -1,166 +0,0 @@
|
|
1 |
-
import torch
|
2 |
-
import numpy as np
|
3 |
-
from diffusers import FluxPipeline, AutoencoderTiny, FlowMatchEulerDiscreteScheduler
|
4 |
-
from typing import Any, Dict, List, Optional, Union
|
5 |
-
|
6 |
-
# Helper functions
|
7 |
-
def calculate_shift(
|
8 |
-
image_seq_len,
|
9 |
-
base_seq_len: int = 256,
|
10 |
-
max_seq_len: int = 4096,
|
11 |
-
base_shift: float = 0.5,
|
12 |
-
max_shift: float = 1.16,
|
13 |
-
):
|
14 |
-
m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
|
15 |
-
b = base_shift - m * base_seq_len
|
16 |
-
mu = image_seq_len * m + b
|
17 |
-
return mu
|
18 |
-
|
19 |
-
def retrieve_timesteps(
|
20 |
-
scheduler,
|
21 |
-
num_inference_steps: Optional[int] = None,
|
22 |
-
device: Optional[Union[str, torch.device]] = None,
|
23 |
-
timesteps: Optional[List[int]] = None,
|
24 |
-
sigmas: Optional[List[float]] = None,
|
25 |
-
**kwargs,
|
26 |
-
):
|
27 |
-
if timesteps is not None and sigmas is not None:
|
28 |
-
raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
|
29 |
-
if timesteps is not None:
|
30 |
-
scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
|
31 |
-
timesteps = scheduler.timesteps
|
32 |
-
num_inference_steps = len(timesteps)
|
33 |
-
elif sigmas is not None:
|
34 |
-
scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
|
35 |
-
timesteps = scheduler.timesteps
|
36 |
-
num_inference_steps = len(timesteps)
|
37 |
-
else:
|
38 |
-
scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
|
39 |
-
timesteps = scheduler.timesteps
|
40 |
-
return timesteps, num_inference_steps
|
41 |
-
|
42 |
-
# FLUX pipeline function
|
43 |
-
@torch.inference_mode()
|
44 |
-
def flux_pipe_call_that_returns_an_iterable_of_images(
|
45 |
-
self,
|
46 |
-
prompt: Union[str, List[str]] = None,
|
47 |
-
prompt_2: Optional[Union[str, List[str]]] = None,
|
48 |
-
height: Optional[int] = None,
|
49 |
-
width: Optional[int] = None,
|
50 |
-
num_inference_steps: int = 28,
|
51 |
-
timesteps: List[int] = None,
|
52 |
-
guidance_scale: float = 3.5,
|
53 |
-
num_images_per_prompt: Optional[int] = 1,
|
54 |
-
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
|
55 |
-
latents: Optional[torch.FloatTensor] = None,
|
56 |
-
prompt_embeds: Optional[torch.FloatTensor] = None,
|
57 |
-
pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
|
58 |
-
output_type: Optional[str] = "pil",
|
59 |
-
return_dict: bool = True,
|
60 |
-
joint_attention_kwargs: Optional[Dict[str, Any]] = None,
|
61 |
-
max_sequence_length: int = 512,
|
62 |
-
good_vae: Optional[Any] = None,
|
63 |
-
):
|
64 |
-
height = height or self.default_sample_size * self.vae_scale_factor
|
65 |
-
width = width or self.default_sample_size * self.vae_scale_factor
|
66 |
-
|
67 |
-
# 1. Check inputs
|
68 |
-
self.check_inputs(
|
69 |
-
prompt,
|
70 |
-
prompt_2,
|
71 |
-
height,
|
72 |
-
width,
|
73 |
-
prompt_embeds=prompt_embeds,
|
74 |
-
pooled_prompt_embeds=pooled_prompt_embeds,
|
75 |
-
max_sequence_length=max_sequence_length,
|
76 |
-
)
|
77 |
-
|
78 |
-
self._guidance_scale = guidance_scale
|
79 |
-
self._joint_attention_kwargs = joint_attention_kwargs
|
80 |
-
self._interrupt = False
|
81 |
-
|
82 |
-
# 2. Define call parameters
|
83 |
-
batch_size = 1 if isinstance(prompt, str) else len(prompt)
|
84 |
-
device = self._execution_device
|
85 |
-
|
86 |
-
# 3. Encode prompt
|
87 |
-
lora_scale = joint_attention_kwargs.get("scale", None) if joint_attention_kwargs is not None else None
|
88 |
-
prompt_embeds, pooled_prompt_embeds, text_ids = self.encode_prompt(
|
89 |
-
prompt=prompt,
|
90 |
-
prompt_2=prompt_2,
|
91 |
-
prompt_embeds=prompt_embeds,
|
92 |
-
pooled_prompt_embeds=pooled_prompt_embeds,
|
93 |
-
device=device,
|
94 |
-
num_images_per_prompt=num_images_per_prompt,
|
95 |
-
max_sequence_length=max_sequence_length,
|
96 |
-
lora_scale=lora_scale,
|
97 |
-
)
|
98 |
-
# 4. Prepare latent variables
|
99 |
-
num_channels_latents = self.transformer.config.in_channels // 4
|
100 |
-
latents, latent_image_ids = self.prepare_latents(
|
101 |
-
batch_size * num_images_per_prompt,
|
102 |
-
num_channels_latents,
|
103 |
-
height,
|
104 |
-
width,
|
105 |
-
prompt_embeds.dtype,
|
106 |
-
device,
|
107 |
-
generator,
|
108 |
-
latents,
|
109 |
-
)
|
110 |
-
# 5. Prepare timesteps
|
111 |
-
sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
|
112 |
-
image_seq_len = latents.shape[1]
|
113 |
-
mu = calculate_shift(
|
114 |
-
image_seq_len,
|
115 |
-
self.scheduler.config.base_image_seq_len,
|
116 |
-
self.scheduler.config.max_image_seq_len,
|
117 |
-
self.scheduler.config.base_shift,
|
118 |
-
self.scheduler.config.max_shift,
|
119 |
-
)
|
120 |
-
timesteps, num_inference_steps = retrieve_timesteps(
|
121 |
-
self.scheduler,
|
122 |
-
num_inference_steps,
|
123 |
-
device,
|
124 |
-
timesteps,
|
125 |
-
sigmas,
|
126 |
-
mu=mu,
|
127 |
-
)
|
128 |
-
self._num_timesteps = len(timesteps)
|
129 |
-
|
130 |
-
# Handle guidance
|
131 |
-
guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32).expand(latents.shape[0]) if self.transformer.config.guidance_embeds else None
|
132 |
-
|
133 |
-
# 6. Denoising loop
|
134 |
-
for i, t in enumerate(timesteps):
|
135 |
-
if self.interrupt:
|
136 |
-
continue
|
137 |
-
|
138 |
-
timestep = t.expand(latents.shape[0]).to(latents.dtype)
|
139 |
-
|
140 |
-
noise_pred = self.transformer(
|
141 |
-
hidden_states=latents,
|
142 |
-
timestep=timestep / 1000,
|
143 |
-
guidance=guidance,
|
144 |
-
pooled_projections=pooled_prompt_embeds,
|
145 |
-
encoder_hidden_states=prompt_embeds,
|
146 |
-
txt_ids=text_ids,
|
147 |
-
img_ids=latent_image_ids,
|
148 |
-
joint_attention_kwargs=self.joint_attention_kwargs,
|
149 |
-
return_dict=False,
|
150 |
-
)[0]
|
151 |
-
# Yield intermediate result
|
152 |
-
latents_for_image = self._unpack_latents(latents, height, width, self.vae_scale_factor)
|
153 |
-
latents_for_image = (latents_for_image / self.vae.config.scaling_factor) + self.vae.config.shift_factor
|
154 |
-
image = self.vae.decode(latents_for_image, return_dict=False)[0]
|
155 |
-
yield self.image_processor.postprocess(image, output_type=output_type)[0]
|
156 |
-
|
157 |
-
latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
|
158 |
-
torch.cuda.empty_cache()
|
159 |
-
|
160 |
-
# Final image using good_vae
|
161 |
-
latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
|
162 |
-
latents = (latents / good_vae.config.scaling_factor) + good_vae.config.shift_factor
|
163 |
-
image = good_vae.decode(latents, return_dict=False)[0]
|
164 |
-
self.maybe_free_model_hooks()
|
165 |
-
torch.cuda.empty_cache()
|
166 |
-
yield self.image_processor.postprocess(image, output_type=output_type)[0]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|