Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -7,6 +7,31 @@ from PIL import Image
|
|
7 |
from huggingface_hub import InferenceApi, InferenceClient
|
8 |
from datasets import load_dataset
|
9 |
import pandas as pd
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
'''
|
12 |
dataset = load_dataset("Gustavosta/Stable-Diffusion-Prompts")
|
@@ -21,26 +46,25 @@ def get_samples():
|
|
21 |
prompt_list = prompt_df.sample(n = 10)["Prompt"].map(lambda x: x).values.tolist()
|
22 |
return prompt_list
|
23 |
|
24 |
-
def update_models():
|
25 |
client = InferenceClient()
|
26 |
models = client.list_deployed_models()
|
27 |
list_models = models["text-to-image"]
|
|
|
|
|
|
|
|
|
28 |
return gr.Dropdown.update(choices=list_models)
|
29 |
|
30 |
def update_prompts():
|
31 |
return gr.Dropdown.update(choices=get_samples())
|
32 |
|
33 |
-
def get_params(request: gr.Request):
|
34 |
params = request.query_params
|
35 |
ip = request.client.host
|
36 |
req = {"params": params,
|
37 |
"ip": ip}
|
38 |
-
return update_models(), update_prompts()
|
39 |
-
|
40 |
-
client = InferenceClient()
|
41 |
-
models = client.list_deployed_models()
|
42 |
-
list_models = models["text-to-image"]
|
43 |
-
list_prompts = get_samples()
|
44 |
|
45 |
'''
|
46 |
list_models = [
|
@@ -239,6 +263,12 @@ with gr.Blocks(css=css) as demo:
|
|
239 |
with gr.Row(elem_id="prompt-container"):
|
240 |
with gr.Column():
|
241 |
btn_refresh = gr.Button(value="Click to get current deployed models and newly Prompt candidates")
|
|
|
|
|
|
|
|
|
|
|
|
|
242 |
#btn_refresh.click(None, js="window.location.reload()")
|
243 |
current_model = gr.Dropdown(label="Current Model", choices=list_models, value=DEFAULT_MODEL,
|
244 |
info = "default model: {}".format(DEFAULT_MODEL)
|
@@ -276,9 +306,11 @@ with gr.Blocks(css=css) as demo:
|
|
276 |
|
277 |
text_button.click(generate_txt2img, inputs=[current_model, text_prompt, negative_prompt, image_style], outputs=image_output)
|
278 |
select_button.click(generate_txt2img, inputs=[current_model, select_prompt, negative_prompt, image_style], outputs=image_output)
|
279 |
-
btn_refresh.click(update_models,
|
280 |
btn_refresh.click(update_prompts, None, select_prompt)
|
|
|
|
|
281 |
|
282 |
-
demo.load(get_params,
|
283 |
|
284 |
demo.launch(show_api=False)
|
|
|
7 |
from huggingface_hub import InferenceApi, InferenceClient
|
8 |
from datasets import load_dataset
|
9 |
import pandas as pd
|
10 |
+
import re
|
11 |
+
|
12 |
+
def rank_score(repo_str):
|
13 |
+
p_list = re.findall(r"[-_xlv\d]+" ,repo_str.split("/")[-1])
|
14 |
+
xl_in_str = any(map(lambda x: "xl" in x, p_list))
|
15 |
+
v_in_str = any(map(lambda x: "v" in x and
|
16 |
+
any(map(lambda y:
|
17 |
+
any(map(lambda z: y.startswith(z), "0123456789"))
|
18 |
+
,x.split("v")))
|
19 |
+
, p_list))
|
20 |
+
stable_in_str = repo_str.split("/")[-1].lower().startswith("stable")
|
21 |
+
score = sum(map(lambda t2: t2[0] * t2[1] ,(zip(*[[stable_in_str, xl_in_str, v_in_str], [1000, 100, 10]]))))
|
22 |
+
#return p_list, xl_in_str, v_in_str, stable_in_str, score
|
23 |
+
return score
|
24 |
+
|
25 |
+
def shorten_by(repo_list, by = None):
|
26 |
+
if by == "user":
|
27 |
+
return sorted(
|
28 |
+
pd.DataFrame(pd.Series(repo_list).map(lambda x: (x.split("/")[0], x)).values.tolist()).groupby(0)[1].apply(list).map(lambda x:
|
29 |
+
sorted(x, key = rank_score, reverse = True)[0]).values.tolist(),
|
30 |
+
key = rank_score, reverse = True
|
31 |
+
)
|
32 |
+
if by == "model":
|
33 |
+
return sorted(repo_list, key = lambda x: rank_score(x), reverse = True)
|
34 |
+
return repo_list
|
35 |
|
36 |
'''
|
37 |
dataset = load_dataset("Gustavosta/Stable-Diffusion-Prompts")
|
|
|
46 |
prompt_list = prompt_df.sample(n = 10)["Prompt"].map(lambda x: x).values.tolist()
|
47 |
return prompt_list
|
48 |
|
49 |
+
def update_models(models_rank_by = "model"):
|
50 |
client = InferenceClient()
|
51 |
models = client.list_deployed_models()
|
52 |
list_models = models["text-to-image"]
|
53 |
+
if hasattr(models_rank_by, "value"):
|
54 |
+
list_models = shorten_by(list_models, models_rank_by.value)
|
55 |
+
else:
|
56 |
+
list_models = shorten_by(list_models, models_rank_by)
|
57 |
return gr.Dropdown.update(choices=list_models)
|
58 |
|
59 |
def update_prompts():
|
60 |
return gr.Dropdown.update(choices=get_samples())
|
61 |
|
62 |
+
def get_params(request: gr.Request, models_rank_by):
|
63 |
params = request.query_params
|
64 |
ip = request.client.host
|
65 |
req = {"params": params,
|
66 |
"ip": ip}
|
67 |
+
return update_models(models_rank_by), update_prompts()
|
|
|
|
|
|
|
|
|
|
|
68 |
|
69 |
'''
|
70 |
list_models = [
|
|
|
263 |
with gr.Row(elem_id="prompt-container"):
|
264 |
with gr.Column():
|
265 |
btn_refresh = gr.Button(value="Click to get current deployed models and newly Prompt candidates")
|
266 |
+
models_rank_by = gr.Radio(choices=["model", "user"],
|
267 |
+
value="model", label="Models ranked by", elem_id="rank_radio")
|
268 |
+
|
269 |
+
list_models = update_models(models_rank_by)
|
270 |
+
list_prompts = get_samples()
|
271 |
+
|
272 |
#btn_refresh.click(None, js="window.location.reload()")
|
273 |
current_model = gr.Dropdown(label="Current Model", choices=list_models, value=DEFAULT_MODEL,
|
274 |
info = "default model: {}".format(DEFAULT_MODEL)
|
|
|
306 |
|
307 |
text_button.click(generate_txt2img, inputs=[current_model, text_prompt, negative_prompt, image_style], outputs=image_output)
|
308 |
select_button.click(generate_txt2img, inputs=[current_model, select_prompt, negative_prompt, image_style], outputs=image_output)
|
309 |
+
btn_refresh.click(update_models, models_rank_by, current_model)
|
310 |
btn_refresh.click(update_prompts, None, select_prompt)
|
311 |
+
|
312 |
+
models_rank_by.change(update_models, models_rank_by, current_model)
|
313 |
|
314 |
+
demo.load(get_params, models_rank_by, [current_model, select_prompt])
|
315 |
|
316 |
demo.launch(show_api=False)
|