Spaces:
Running
on
Zero
Running
on
Zero
Commit
•
aaff709
1
Parent(s):
2ed4418
Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,9 @@ import logging
|
|
5 |
import torch
|
6 |
from PIL import Image
|
7 |
import spaces
|
8 |
-
from diffusers import DiffusionPipeline
|
|
|
|
|
9 |
from huggingface_hub import hf_hub_download, HfFileSystem, ModelCard, snapshot_download
|
10 |
import copy
|
11 |
import random
|
@@ -16,11 +18,18 @@ with open('loras.json', 'r') as f:
|
|
16 |
loras = json.load(f)
|
17 |
|
18 |
# Initialize the base model
|
|
|
|
|
19 |
base_model = "black-forest-labs/FLUX.1-dev"
|
20 |
-
|
|
|
|
|
|
|
21 |
|
22 |
MAX_SEED = 2**32-1
|
23 |
|
|
|
|
|
24 |
class calculateDuration:
|
25 |
def __init__(self, activity_name=""):
|
26 |
self.activity_name = activity_name
|
@@ -61,10 +70,9 @@ def update_selection(evt: gr.SelectData, width, height):
|
|
61 |
def generate_image(prompt_mash, steps, seed, cfg_scale, width, height, lora_scale, progress):
|
62 |
pipe.to("cuda")
|
63 |
generator = torch.Generator(device="cuda").manual_seed(seed)
|
64 |
-
|
65 |
with calculateDuration("Generating image"):
|
66 |
# Generate image
|
67 |
-
|
68 |
prompt=prompt_mash,
|
69 |
num_inference_steps=steps,
|
70 |
guidance_scale=cfg_scale,
|
@@ -72,13 +80,14 @@ def generate_image(prompt_mash, steps, seed, cfg_scale, width, height, lora_scal
|
|
72 |
height=height,
|
73 |
generator=generator,
|
74 |
joint_attention_kwargs={"scale": lora_scale},
|
75 |
-
|
76 |
-
|
|
|
|
|
77 |
|
78 |
def run_lora(prompt, cfg_scale, steps, selected_index, randomize_seed, seed, width, height, lora_scale, progress=gr.Progress(track_tqdm=True)):
|
79 |
if selected_index is None:
|
80 |
raise gr.Error("You must select a LoRA before proceeding.")
|
81 |
-
|
82 |
selected_lora = loras[selected_index]
|
83 |
lora_path = selected_lora["repo"]
|
84 |
trigger_word = selected_lora["trigger_word"]
|
@@ -92,24 +101,31 @@ def run_lora(prompt, cfg_scale, steps, selected_index, randomize_seed, seed, wid
|
|
92 |
prompt_mash = f"{trigger_word} {prompt}"
|
93 |
else:
|
94 |
prompt_mash = prompt
|
|
|
95 |
# Load LoRA weights
|
96 |
with calculateDuration(f"Loading LoRA weights for {selected_lora['title']}"):
|
97 |
if "weights" in selected_lora:
|
98 |
pipe.load_lora_weights(lora_path, weight_name=selected_lora["weights"])
|
99 |
-
#pipe.fuse_lora()
|
100 |
else:
|
101 |
pipe.load_lora_weights(lora_path)
|
102 |
-
|
103 |
# Set random seed for reproducibility
|
104 |
with calculateDuration("Randomizing seed"):
|
105 |
if randomize_seed:
|
106 |
seed = random.randint(0, MAX_SEED)
|
|
|
|
|
107 |
|
108 |
-
|
|
|
|
|
|
|
|
|
|
|
109 |
pipe.to("cpu")
|
110 |
-
#pipe.unfuse_lora()
|
111 |
pipe.unload_lora_weights()
|
112 |
-
|
|
|
113 |
|
114 |
def get_huggingface_safetensors(link):
|
115 |
split_link = link.split("/")
|
|
|
5 |
import torch
|
6 |
from PIL import Image
|
7 |
import spaces
|
8 |
+
from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL
|
9 |
+
from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
|
10 |
+
|
11 |
from huggingface_hub import hf_hub_download, HfFileSystem, ModelCard, snapshot_download
|
12 |
import copy
|
13 |
import random
|
|
|
18 |
loras = json.load(f)
|
19 |
|
20 |
# Initialize the base model
|
21 |
+
dtype = torch.bfloat16
|
22 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
23 |
base_model = "black-forest-labs/FLUX.1-dev"
|
24 |
+
|
25 |
+
taef1 = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=dtype).to(device)
|
26 |
+
good_vae = AutoencoderKL.from_pretrained(base_model, subfolder="vae", torch_dtype=dtype).to(device)
|
27 |
+
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=dtype, vae=taef1).to(device)
|
28 |
|
29 |
MAX_SEED = 2**32-1
|
30 |
|
31 |
+
pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)
|
32 |
+
|
33 |
class calculateDuration:
|
34 |
def __init__(self, activity_name=""):
|
35 |
self.activity_name = activity_name
|
|
|
70 |
def generate_image(prompt_mash, steps, seed, cfg_scale, width, height, lora_scale, progress):
|
71 |
pipe.to("cuda")
|
72 |
generator = torch.Generator(device="cuda").manual_seed(seed)
|
|
|
73 |
with calculateDuration("Generating image"):
|
74 |
# Generate image
|
75 |
+
for img in pipe.flux_pipe_call_that_returns_an_iterable_of_images(
|
76 |
prompt=prompt_mash,
|
77 |
num_inference_steps=steps,
|
78 |
guidance_scale=cfg_scale,
|
|
|
80 |
height=height,
|
81 |
generator=generator,
|
82 |
joint_attention_kwargs={"scale": lora_scale},
|
83 |
+
output_type="pil",
|
84 |
+
good_vae=good_vae,
|
85 |
+
):
|
86 |
+
yield img
|
87 |
|
88 |
def run_lora(prompt, cfg_scale, steps, selected_index, randomize_seed, seed, width, height, lora_scale, progress=gr.Progress(track_tqdm=True)):
|
89 |
if selected_index is None:
|
90 |
raise gr.Error("You must select a LoRA before proceeding.")
|
|
|
91 |
selected_lora = loras[selected_index]
|
92 |
lora_path = selected_lora["repo"]
|
93 |
trigger_word = selected_lora["trigger_word"]
|
|
|
101 |
prompt_mash = f"{trigger_word} {prompt}"
|
102 |
else:
|
103 |
prompt_mash = prompt
|
104 |
+
|
105 |
# Load LoRA weights
|
106 |
with calculateDuration(f"Loading LoRA weights for {selected_lora['title']}"):
|
107 |
if "weights" in selected_lora:
|
108 |
pipe.load_lora_weights(lora_path, weight_name=selected_lora["weights"])
|
|
|
109 |
else:
|
110 |
pipe.load_lora_weights(lora_path)
|
111 |
+
|
112 |
# Set random seed for reproducibility
|
113 |
with calculateDuration("Randomizing seed"):
|
114 |
if randomize_seed:
|
115 |
seed = random.randint(0, MAX_SEED)
|
116 |
+
|
117 |
+
image_generator = generate_image(prompt_mash, steps, seed, cfg_scale, width, height, lora_scale, progress)
|
118 |
|
119 |
+
# Consume the generator to get the final image
|
120 |
+
final_image = None
|
121 |
+
for image in image_generator:
|
122 |
+
final_image = image
|
123 |
+
yield image, seed # Yield intermediate images and seed
|
124 |
+
|
125 |
pipe.to("cpu")
|
|
|
126 |
pipe.unload_lora_weights()
|
127 |
+
|
128 |
+
return final_image, seed # Return the final image and seed
|
129 |
|
130 |
def get_huggingface_safetensors(link):
|
131 |
split_link = link.split("/")
|