Spaces:
Running
on
A100
Running
on
A100
type hints and Pydantic
Browse files- server/connection_manager.py +13 -8
- server/main.py +88 -33
- server/pipelines/img2imgFlux.py +157 -0
- server/requirements.txt +10 -3
- server/util.py +24 -2
server/connection_manager.py
CHANGED
@@ -1,12 +1,12 @@
|
|
1 |
-
from typing import Dict, Union
|
2 |
from uuid import UUID
|
3 |
import asyncio
|
4 |
from fastapi import WebSocket
|
5 |
from starlette.websockets import WebSocketState
|
6 |
import logging
|
7 |
-
from
|
|
|
8 |
|
9 |
-
Connections =
|
10 |
|
11 |
|
12 |
class ServerFullException(Exception):
|
@@ -44,13 +44,13 @@ class ConnectionManager:
|
|
44 |
def check_user(self, user_id: UUID) -> bool:
|
45 |
return user_id in self.active_connections
|
46 |
|
47 |
-
async def update_data(self, user_id: UUID, new_data:
|
48 |
user_session = self.active_connections.get(user_id)
|
49 |
if user_session:
|
50 |
queue = user_session["queue"]
|
51 |
await queue.put(new_data)
|
52 |
|
53 |
-
async def get_latest_data(self, user_id: UUID) ->
|
54 |
user_session = self.active_connections.get(user_id)
|
55 |
if user_session:
|
56 |
queue = user_session["queue"]
|
@@ -58,6 +58,7 @@ class ConnectionManager:
|
|
58 |
return await queue.get()
|
59 |
except asyncio.QueueEmpty:
|
60 |
return None
|
|
|
61 |
|
62 |
def delete_user(self, user_id: UUID):
|
63 |
user_session = self.active_connections.pop(user_id, None)
|
@@ -86,7 +87,7 @@ class ConnectionManager:
|
|
86 |
await websocket.close()
|
87 |
self.delete_user(user_id)
|
88 |
|
89 |
-
async def send_json(self, user_id: UUID, data:
|
90 |
try:
|
91 |
websocket = self.get_websocket(user_id)
|
92 |
if websocket:
|
@@ -94,18 +95,22 @@ class ConnectionManager:
|
|
94 |
except Exception as e:
|
95 |
logging.error(f"Error: Send json: {e}")
|
96 |
|
97 |
-
async def receive_json(self, user_id: UUID) ->
|
98 |
try:
|
99 |
websocket = self.get_websocket(user_id)
|
100 |
if websocket:
|
101 |
return await websocket.receive_json()
|
|
|
102 |
except Exception as e:
|
103 |
logging.error(f"Error: Receive json: {e}")
|
|
|
104 |
|
105 |
-
async def receive_bytes(self, user_id: UUID) -> bytes:
|
106 |
try:
|
107 |
websocket = self.get_websocket(user_id)
|
108 |
if websocket:
|
109 |
return await websocket.receive_bytes()
|
|
|
110 |
except Exception as e:
|
111 |
logging.error(f"Error: Receive bytes: {e}")
|
|
|
|
|
|
1 |
from uuid import UUID
|
2 |
import asyncio
|
3 |
from fastapi import WebSocket
|
4 |
from starlette.websockets import WebSocketState
|
5 |
import logging
|
6 |
+
from typing import Any, TypeVar
|
7 |
+
from util import ParamsModel
|
8 |
|
9 |
+
Connections = dict[UUID, dict[str, WebSocket | asyncio.Queue]]
|
10 |
|
11 |
|
12 |
class ServerFullException(Exception):
|
|
|
44 |
def check_user(self, user_id: UUID) -> bool:
|
45 |
return user_id in self.active_connections
|
46 |
|
47 |
+
async def update_data(self, user_id: UUID, new_data: ParamsModel):
|
48 |
user_session = self.active_connections.get(user_id)
|
49 |
if user_session:
|
50 |
queue = user_session["queue"]
|
51 |
await queue.put(new_data)
|
52 |
|
53 |
+
async def get_latest_data(self, user_id: UUID) -> ParamsModel | None:
|
54 |
user_session = self.active_connections.get(user_id)
|
55 |
if user_session:
|
56 |
queue = user_session["queue"]
|
|
|
58 |
return await queue.get()
|
59 |
except asyncio.QueueEmpty:
|
60 |
return None
|
61 |
+
return None
|
62 |
|
63 |
def delete_user(self, user_id: UUID):
|
64 |
user_session = self.active_connections.pop(user_id, None)
|
|
|
87 |
await websocket.close()
|
88 |
self.delete_user(user_id)
|
89 |
|
90 |
+
async def send_json(self, user_id: UUID, data: dict):
|
91 |
try:
|
92 |
websocket = self.get_websocket(user_id)
|
93 |
if websocket:
|
|
|
95 |
except Exception as e:
|
96 |
logging.error(f"Error: Send json: {e}")
|
97 |
|
98 |
+
async def receive_json(self, user_id: UUID) -> dict | None:
|
99 |
try:
|
100 |
websocket = self.get_websocket(user_id)
|
101 |
if websocket:
|
102 |
return await websocket.receive_json()
|
103 |
+
return None
|
104 |
except Exception as e:
|
105 |
logging.error(f"Error: Receive json: {e}")
|
106 |
+
return None
|
107 |
|
108 |
+
async def receive_bytes(self, user_id: UUID) -> bytes | None:
|
109 |
try:
|
110 |
websocket = self.get_websocket(user_id)
|
111 |
if websocket:
|
112 |
return await websocket.receive_bytes()
|
113 |
+
return None
|
114 |
except Exception as e:
|
115 |
logging.error(f"Error: Receive bytes: {e}")
|
116 |
+
return None
|
server/main.py
CHANGED
@@ -10,30 +10,53 @@ import logging
|
|
10 |
from config import config, Args
|
11 |
from connection_manager import ConnectionManager, ServerFullException
|
12 |
import uuid
|
|
|
13 |
import time
|
14 |
-
from
|
15 |
-
from util import pil_to_frame, bytes_to_pil, is_firefox, get_pipeline_class
|
16 |
from device import device, torch_dtype
|
17 |
import asyncio
|
18 |
import os
|
19 |
import time
|
20 |
import torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
|
23 |
THROTTLE = 1.0 / 120
|
24 |
|
25 |
|
26 |
class App:
|
27 |
-
def __init__(self, config: Args,
|
28 |
self.args = config
|
29 |
-
self.pipeline =
|
30 |
self.app = FastAPI()
|
31 |
self.conn_manager = ConnectionManager()
|
|
|
32 |
if self.args.safety_checker:
|
33 |
self.safety_checker = SafetyChecker(device=device.type)
|
34 |
self.init_app()
|
35 |
|
36 |
-
def init_app(self):
|
37 |
self.app.add_middleware(
|
38 |
CORSMiddleware,
|
39 |
allow_origins=["*"],
|
@@ -43,7 +66,7 @@ class App:
|
|
43 |
)
|
44 |
|
45 |
@self.app.websocket("/api/ws/{user_id}")
|
46 |
-
async def websocket_endpoint(user_id:
|
47 |
try:
|
48 |
await self.conn_manager.connect(
|
49 |
user_id, websocket, self.args.max_queue_size
|
@@ -55,9 +78,9 @@ class App:
|
|
55 |
await self.conn_manager.disconnect(user_id)
|
56 |
logging.info(f"User disconnected: {user_id}")
|
57 |
|
58 |
-
async def handle_websocket_data(user_id:
|
59 |
if not self.conn_manager.check_user(user_id):
|
60 |
-
|
61 |
last_time = time.time()
|
62 |
try:
|
63 |
while True:
|
@@ -75,19 +98,29 @@ class App:
|
|
75 |
await self.conn_manager.disconnect(user_id)
|
76 |
return
|
77 |
data = await self.conn_manager.receive_json(user_id)
|
|
|
|
|
|
|
78 |
if data["status"] == "next_frame":
|
79 |
-
info = pipeline.Info()
|
80 |
-
|
81 |
-
|
82 |
-
|
|
|
|
|
|
|
83 |
if info.input_mode == "image":
|
84 |
image_data = await self.conn_manager.receive_bytes(user_id)
|
85 |
-
if len(image_data) == 0:
|
86 |
await self.conn_manager.send_json(
|
87 |
user_id, {"status": "send_frame"}
|
88 |
)
|
89 |
continue
|
90 |
-
|
|
|
|
|
|
|
|
|
91 |
|
92 |
await self.conn_manager.update_data(user_id, params)
|
93 |
await self.conn_manager.send_json(user_id, {"status": "wait"})
|
@@ -97,29 +130,32 @@ class App:
|
|
97 |
await self.conn_manager.disconnect(user_id)
|
98 |
|
99 |
@self.app.get("/api/queue")
|
100 |
-
async def get_queue_size():
|
101 |
queue_size = self.conn_manager.get_user_count()
|
102 |
return JSONResponse({"queue_size": queue_size})
|
103 |
|
104 |
@self.app.get("/api/stream/{user_id}")
|
105 |
-
async def stream(user_id:
|
106 |
try:
|
107 |
-
|
108 |
-
|
109 |
-
last_params = SimpleNamespace()
|
110 |
while True:
|
111 |
last_time = time.time()
|
112 |
await self.conn_manager.send_json(
|
113 |
user_id, {"status": "send_frame"}
|
114 |
)
|
115 |
params = await self.conn_manager.get_latest_data(user_id)
|
116 |
-
|
|
|
|
|
|
|
117 |
await asyncio.sleep(THROTTLE)
|
118 |
continue
|
119 |
-
|
120 |
-
|
|
|
121 |
|
122 |
-
if self.args.safety_checker:
|
123 |
image, has_nsfw_concept = self.safety_checker(image)
|
124 |
if has_nsfw_concept:
|
125 |
image = None
|
@@ -141,23 +177,24 @@ class App:
|
|
141 |
)
|
142 |
except Exception as e:
|
143 |
logging.error(f"Streaming Error: {e}, {user_id} ")
|
144 |
-
|
145 |
|
146 |
# route to setup frontend
|
147 |
@self.app.get("/api/settings")
|
148 |
-
async def settings():
|
149 |
-
info_schema = pipeline.Info.schema()
|
150 |
-
info = pipeline.Info()
|
151 |
-
|
|
|
152 |
page_content = markdown2.markdown(info.page_content)
|
153 |
|
154 |
-
input_params = pipeline.InputParams.schema()
|
155 |
return JSONResponse(
|
156 |
{
|
157 |
"info": info_schema,
|
158 |
"input_params": input_params,
|
159 |
"max_queue_size": self.args.max_queue_size,
|
160 |
-
"page_content": page_content
|
161 |
}
|
162 |
)
|
163 |
|
@@ -169,17 +206,35 @@ class App:
|
|
169 |
)
|
170 |
|
171 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
print(f"Device: {device}")
|
173 |
print(f"torch_dtype: {torch_dtype}")
|
|
|
174 |
pipeline_class = get_pipeline_class(config.pipeline)
|
175 |
-
|
176 |
-
app = App(config,
|
|
|
177 |
|
178 |
if __name__ == "__main__":
|
179 |
import uvicorn
|
180 |
|
|
|
|
|
181 |
uvicorn.run(
|
182 |
-
|
183 |
host=config.host,
|
184 |
port=config.port,
|
185 |
reload=config.reload,
|
|
|
10 |
from config import config, Args
|
11 |
from connection_manager import ConnectionManager, ServerFullException
|
12 |
import uuid
|
13 |
+
from uuid import UUID
|
14 |
import time
|
15 |
+
from typing import Any, Protocol, TypeVar, runtime_checkable, cast
|
16 |
+
from util import pil_to_frame, bytes_to_pil, is_firefox, get_pipeline_class, ParamsModel
|
17 |
from device import device, torch_dtype
|
18 |
import asyncio
|
19 |
import os
|
20 |
import time
|
21 |
import torch
|
22 |
+
from pydantic import BaseModel, create_model
|
23 |
+
|
24 |
+
@runtime_checkable
|
25 |
+
class BasePipeline(Protocol):
|
26 |
+
class Info:
|
27 |
+
@classmethod
|
28 |
+
def schema(cls) -> dict[str, Any]:
|
29 |
+
...
|
30 |
+
page_content: str | None
|
31 |
+
input_mode: str
|
32 |
+
|
33 |
+
class InputParams(ParamsModel):
|
34 |
+
@classmethod
|
35 |
+
def schema(cls) -> dict[str, Any]:
|
36 |
+
...
|
37 |
+
|
38 |
+
def dict(self) -> dict[str, Any]:
|
39 |
+
...
|
40 |
+
|
41 |
+
def predict(self, params: ParamsModel) -> Image.Image | None:
|
42 |
+
...
|
43 |
|
44 |
|
45 |
THROTTLE = 1.0 / 120
|
46 |
|
47 |
|
48 |
class App:
|
49 |
+
def __init__(self, config: Args, pipeline_instance: BasePipeline):
|
50 |
self.args = config
|
51 |
+
self.pipeline = pipeline_instance
|
52 |
self.app = FastAPI()
|
53 |
self.conn_manager = ConnectionManager()
|
54 |
+
self.safety_checker: SafetyChecker | None = None
|
55 |
if self.args.safety_checker:
|
56 |
self.safety_checker = SafetyChecker(device=device.type)
|
57 |
self.init_app()
|
58 |
|
59 |
+
def init_app(self) -> None:
|
60 |
self.app.add_middleware(
|
61 |
CORSMiddleware,
|
62 |
allow_origins=["*"],
|
|
|
66 |
)
|
67 |
|
68 |
@self.app.websocket("/api/ws/{user_id}")
|
69 |
+
async def websocket_endpoint(user_id: UUID, websocket: WebSocket) -> None:
|
70 |
try:
|
71 |
await self.conn_manager.connect(
|
72 |
user_id, websocket, self.args.max_queue_size
|
|
|
78 |
await self.conn_manager.disconnect(user_id)
|
79 |
logging.info(f"User disconnected: {user_id}")
|
80 |
|
81 |
+
async def handle_websocket_data(user_id: UUID) -> None:
|
82 |
if not self.conn_manager.check_user(user_id):
|
83 |
+
raise HTTPException(status_code=404, detail="User not found")
|
84 |
last_time = time.time()
|
85 |
try:
|
86 |
while True:
|
|
|
98 |
await self.conn_manager.disconnect(user_id)
|
99 |
return
|
100 |
data = await self.conn_manager.receive_json(user_id)
|
101 |
+
if data is None:
|
102 |
+
continue
|
103 |
+
|
104 |
if data["status"] == "next_frame":
|
105 |
+
info = self.pipeline.Info()
|
106 |
+
params_data = await self.conn_manager.receive_json(user_id)
|
107 |
+
if params_data is None:
|
108 |
+
continue
|
109 |
+
|
110 |
+
params = self.pipeline.InputParams.model_validate(params_data)
|
111 |
+
|
112 |
if info.input_mode == "image":
|
113 |
image_data = await self.conn_manager.receive_bytes(user_id)
|
114 |
+
if image_data is None or len(image_data) == 0:
|
115 |
await self.conn_manager.send_json(
|
116 |
user_id, {"status": "send_frame"}
|
117 |
)
|
118 |
continue
|
119 |
+
|
120 |
+
# Create a new Pydantic model with the image field
|
121 |
+
params_dict = params.model_dump()
|
122 |
+
params_dict["image"] = bytes_to_pil(image_data)
|
123 |
+
params = self.pipeline.InputParams.model_validate(params_dict)
|
124 |
|
125 |
await self.conn_manager.update_data(user_id, params)
|
126 |
await self.conn_manager.send_json(user_id, {"status": "wait"})
|
|
|
130 |
await self.conn_manager.disconnect(user_id)
|
131 |
|
132 |
@self.app.get("/api/queue")
|
133 |
+
async def get_queue_size() -> JSONResponse:
|
134 |
queue_size = self.conn_manager.get_user_count()
|
135 |
return JSONResponse({"queue_size": queue_size})
|
136 |
|
137 |
@self.app.get("/api/stream/{user_id}")
|
138 |
+
async def stream(user_id: UUID, request: Request) -> StreamingResponse:
|
139 |
try:
|
140 |
+
async def generate() -> bytes:
|
141 |
+
last_params: ParamsModel | None = None
|
|
|
142 |
while True:
|
143 |
last_time = time.time()
|
144 |
await self.conn_manager.send_json(
|
145 |
user_id, {"status": "send_frame"}
|
146 |
)
|
147 |
params = await self.conn_manager.get_latest_data(user_id)
|
148 |
+
|
149 |
+
if (params is None or
|
150 |
+
(last_params is not None and
|
151 |
+
params.model_dump() == last_params.model_dump())):
|
152 |
await asyncio.sleep(THROTTLE)
|
153 |
continue
|
154 |
+
|
155 |
+
last_params = params
|
156 |
+
image = self.pipeline.predict(params)
|
157 |
|
158 |
+
if self.args.safety_checker and self.safety_checker is not None and image is not None:
|
159 |
image, has_nsfw_concept = self.safety_checker(image)
|
160 |
if has_nsfw_concept:
|
161 |
image = None
|
|
|
177 |
)
|
178 |
except Exception as e:
|
179 |
logging.error(f"Streaming Error: {e}, {user_id} ")
|
180 |
+
raise HTTPException(status_code=404, detail="User not found")
|
181 |
|
182 |
# route to setup frontend
|
183 |
@self.app.get("/api/settings")
|
184 |
+
async def settings() -> JSONResponse:
|
185 |
+
info_schema = self.pipeline.Info.schema()
|
186 |
+
info = self.pipeline.Info()
|
187 |
+
page_content = ""
|
188 |
+
if hasattr(info, 'page_content') and info.page_content:
|
189 |
page_content = markdown2.markdown(info.page_content)
|
190 |
|
191 |
+
input_params = self.pipeline.InputParams.schema()
|
192 |
return JSONResponse(
|
193 |
{
|
194 |
"info": info_schema,
|
195 |
"input_params": input_params,
|
196 |
"max_queue_size": self.args.max_queue_size,
|
197 |
+
"page_content": page_content,
|
198 |
}
|
199 |
)
|
200 |
|
|
|
206 |
)
|
207 |
|
208 |
|
209 |
+
# def create_app(config):
|
210 |
+
# print(f"Device: {device}")
|
211 |
+
# print(f"torch_dtype: {torch_dtype}")
|
212 |
+
|
213 |
+
# # Create pipeline once
|
214 |
+
# pipeline_class = get_pipeline_class(config.pipeline)
|
215 |
+
# pipeline_instance = pipeline_class(config, device, torch_dtype)
|
216 |
+
|
217 |
+
# # Pass the existing pipeline instance to App
|
218 |
+
# app = App(config, pipeline_instance).app
|
219 |
+
# return app
|
220 |
+
|
221 |
+
|
222 |
+
# Create app instance at module level
|
223 |
print(f"Device: {device}")
|
224 |
print(f"torch_dtype: {torch_dtype}")
|
225 |
+
|
226 |
pipeline_class = get_pipeline_class(config.pipeline)
|
227 |
+
pipeline_instance = pipeline_class(config, device, torch_dtype)
|
228 |
+
app = App(config, pipeline_instance).app # This creates the FastAPI app instance
|
229 |
+
|
230 |
|
231 |
if __name__ == "__main__":
|
232 |
import uvicorn
|
233 |
|
234 |
+
# app = create_app(config) # Create the app once
|
235 |
+
|
236 |
uvicorn.run(
|
237 |
+
app,
|
238 |
host=config.host,
|
239 |
port=config.port,
|
240 |
reload=config.reload,
|
server/pipelines/img2imgFlux.py
ADDED
@@ -0,0 +1,157 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
|
3 |
+
from optimum.quanto import freeze, qfloat8, quantize
|
4 |
+
from transformers.modeling_utils import PreTrainedModel
|
5 |
+
|
6 |
+
from diffusers import (
|
7 |
+
FlowMatchEulerDiscreteScheduler,
|
8 |
+
AutoencoderKL,
|
9 |
+
AutoencoderTiny,
|
10 |
+
FluxImg2ImgPipeline,
|
11 |
+
FluxPipeline,
|
12 |
+
)
|
13 |
+
|
14 |
+
from diffusers import (
|
15 |
+
FluxImg2ImgPipeline,
|
16 |
+
FluxPipeline,
|
17 |
+
FluxTransformer2DModel,
|
18 |
+
GGUFQuantizationConfig,
|
19 |
+
)
|
20 |
+
|
21 |
+
try:
|
22 |
+
import intel_extension_for_pytorch as ipex # type: ignore
|
23 |
+
except:
|
24 |
+
pass
|
25 |
+
|
26 |
+
import psutil
|
27 |
+
from config import Args
|
28 |
+
from pydantic import BaseModel, Field
|
29 |
+
from PIL import Image
|
30 |
+
from pathlib import Path
|
31 |
+
import math
|
32 |
+
import gc
|
33 |
+
|
34 |
+
|
35 |
+
# model_path = "black-forest-labs/FLUX.1-dev"
|
36 |
+
model_path = "black-forest-labs/FLUX.1-schnell"
|
37 |
+
base_model_path = "black-forest-labs/FLUX.1-schnell"
|
38 |
+
taesd_path = "madebyollin/taef1"
|
39 |
+
subfolder = "transformer"
|
40 |
+
transformer_path = model_path
|
41 |
+
models_path = Path("models")
|
42 |
+
|
43 |
+
default_prompt = "close-up photography of old man standing in the rain at night, in a street lit by lamps, leica 35mm summilux"
|
44 |
+
default_negative_prompt = "blurry, low quality, render, 3D, oversaturated"
|
45 |
+
page_content = """
|
46 |
+
<h1 class="text-3xl font-bold">Real-Time FLUX</h1>
|
47 |
+
|
48 |
+
"""
|
49 |
+
|
50 |
+
|
51 |
+
def flush():
|
52 |
+
torch.cuda.empty_cache()
|
53 |
+
gc.collect()
|
54 |
+
|
55 |
+
|
56 |
+
class Pipeline:
|
57 |
+
class Info(BaseModel):
|
58 |
+
name: str = "img2img"
|
59 |
+
title: str = "Image-to-Image SDXL"
|
60 |
+
description: str = "Generates an image from a text prompt"
|
61 |
+
input_mode: str = "image"
|
62 |
+
page_content: str = page_content
|
63 |
+
|
64 |
+
class InputParams(BaseModel):
|
65 |
+
prompt: str = Field(
|
66 |
+
default_prompt,
|
67 |
+
title="Prompt",
|
68 |
+
field="textarea",
|
69 |
+
id="prompt",
|
70 |
+
)
|
71 |
+
seed: int = Field(
|
72 |
+
2159232, min=0, title="Seed", field="seed", hide=True, id="seed"
|
73 |
+
)
|
74 |
+
steps: int = Field(
|
75 |
+
1, min=1, max=15, title="Steps", field="range", hide=True, id="steps"
|
76 |
+
)
|
77 |
+
width: int = Field(
|
78 |
+
256, min=2, max=15, title="Width", disabled=True, hide=True, id="width"
|
79 |
+
)
|
80 |
+
height: int = Field(
|
81 |
+
256, min=2, max=15, title="Height", disabled=True, hide=True, id="height"
|
82 |
+
)
|
83 |
+
strength: float = Field(
|
84 |
+
0.5,
|
85 |
+
min=0.25,
|
86 |
+
max=1.0,
|
87 |
+
step=0.001,
|
88 |
+
title="Strength",
|
89 |
+
field="range",
|
90 |
+
hide=True,
|
91 |
+
id="strength",
|
92 |
+
)
|
93 |
+
guidance: float = Field(
|
94 |
+
3.5,
|
95 |
+
min=0,
|
96 |
+
max=20,
|
97 |
+
step=0.001,
|
98 |
+
title="Guidance",
|
99 |
+
hide=True,
|
100 |
+
field="range",
|
101 |
+
id="guidance",
|
102 |
+
)
|
103 |
+
|
104 |
+
def __init__(self, args: Args, device: torch.device, torch_dtype: torch.dtype):
|
105 |
+
# ckpt_path = (
|
106 |
+
# "https://huggingface.co/city96/FLUX.1-dev-gguf/blob/main/flux1-dev-Q2_K.gguf"
|
107 |
+
# )
|
108 |
+
print("Loading model")
|
109 |
+
# ckpt_path: str = "https://huggingface.co/city96/FLUX.1-schnell-gguf/blob/main/flux1-schnell-Q6_K.gguf"
|
110 |
+
ckpt_path: str = "https://huggingface.co/city96/FLUX.1-schnell-gguf/blob/main/flux1-schnell-Q4_K_S.gguf"
|
111 |
+
transformer = FluxTransformer2DModel.from_single_file(
|
112 |
+
ckpt_path,
|
113 |
+
quantization_config=GGUFQuantizationConfig(compute_dtype=torch.bfloat16),
|
114 |
+
torch_dtype=torch.bfloat16,
|
115 |
+
)
|
116 |
+
|
117 |
+
# else:
|
118 |
+
pipe = FluxImg2ImgPipeline.from_pretrained(
|
119 |
+
# "black-forest-labs/FLUX.1-dev",
|
120 |
+
"black-forest-labs/FLUX.1-Schnell",
|
121 |
+
transformer=transformer,
|
122 |
+
torch_dtype=torch.bfloat16,
|
123 |
+
)
|
124 |
+
if args.taesd:
|
125 |
+
pipe.vae = AutoencoderTiny.from_pretrained(
|
126 |
+
taesd_path, torch_dtype=torch.bfloat16, use_safetensors=True
|
127 |
+
)
|
128 |
+
# pipe.enable_model_cpu_offload()
|
129 |
+
pipe = pipe.to(device)
|
130 |
+
|
131 |
+
# pipe.enable_model_cpu_offload()
|
132 |
+
|
133 |
+
self.pipe = pipe
|
134 |
+
self.pipe.set_progress_bar_config(disable=True)
|
135 |
+
|
136 |
+
# vae = AutoencoderKL.from_pretrained(
|
137 |
+
# base_model_path, subfolder="vae", torch_dtype=torch_dtype
|
138 |
+
# )
|
139 |
+
|
140 |
+
def predict(self, params: "Pipeline.InputParams") -> Image.Image:
|
141 |
+
generator = torch.manual_seed(params.seed)
|
142 |
+
steps = params.steps
|
143 |
+
strength = params.strength
|
144 |
+
prompt = params.prompt
|
145 |
+
guidance = params.guidance
|
146 |
+
|
147 |
+
results = self.pipe(
|
148 |
+
image=params.image,
|
149 |
+
prompt=prompt,
|
150 |
+
generator=generator,
|
151 |
+
strength=strength,
|
152 |
+
num_inference_steps=steps,
|
153 |
+
guidance_scale=guidance,
|
154 |
+
width=params.width,
|
155 |
+
height=params.height,
|
156 |
+
)
|
157 |
+
return results.images[0]
|
server/requirements.txt
CHANGED
@@ -15,9 +15,16 @@ xformers; sys_platform != 'darwin' or platform_machine != 'arm64'
|
|
15 |
markdown2
|
16 |
safetensors
|
17 |
stable_fast @ https://github.com/chengzeyi/stable-fast/releases/download/nightly/stable_fast-1.0.5.dev20241127+torch230cu121-cp310-cp310-manylinux2014_x86_64.whl ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
18 |
-
oneflow @ https://github.com/siliconflow/oneflow_releases/releases/download/community_cu121/oneflow-0.9.1.dev20241114%2Bcu121-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
19 |
-
onediff @ git+https://github.com/siliconflow/onediff.git@main#egg=onediff ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
20 |
setuptools
|
21 |
mpmath==1.3.0
|
22 |
numpy==1.*
|
23 |
-
controlnet-aux
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
markdown2
|
16 |
safetensors
|
17 |
stable_fast @ https://github.com/chengzeyi/stable-fast/releases/download/nightly/stable_fast-1.0.5.dev20241127+torch230cu121-cp310-cp310-manylinux2014_x86_64.whl ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
18 |
+
#oneflow @ https://github.com/siliconflow/oneflow_releases/releases/download/community_cu121/oneflow-0.9.1.dev20241114%2Bcu121-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
19 |
+
#onediff @ git+https://github.com/siliconflow/onediff.git@main#egg=onediff ; sys_platform != 'darwin' or platform_machine != 'arm64'
|
20 |
setuptools
|
21 |
mpmath==1.3.0
|
22 |
numpy==1.*
|
23 |
+
controlnet-aux
|
24 |
+
sentencepiece==0.2.0
|
25 |
+
optimum-quanto
|
26 |
+
gguf==0.13.0
|
27 |
+
pydantic>=2.7.0
|
28 |
+
types-Pillow
|
29 |
+
mypy
|
30 |
+
python-dotenv
|
server/util.py
CHANGED
@@ -1,10 +1,28 @@
|
|
1 |
from importlib import import_module
|
2 |
-
from
|
3 |
from PIL import Image
|
4 |
import io
|
|
|
5 |
|
6 |
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
try:
|
9 |
module = import_module(f"pipelines.{pipeline_name}")
|
10 |
except ModuleNotFoundError:
|
@@ -15,6 +33,10 @@ def get_pipeline_class(pipeline_name: str) -> ModuleType:
|
|
15 |
if pipeline_class is None:
|
16 |
raise ValueError(f"'Pipeline' class not found in module '{pipeline_name}'.")
|
17 |
|
|
|
|
|
|
|
|
|
18 |
return pipeline_class
|
19 |
|
20 |
|
|
|
1 |
from importlib import import_module
|
2 |
+
from typing import Any, TypeVar, Generic, TypeVar
|
3 |
from PIL import Image
|
4 |
import io
|
5 |
+
from pydantic import BaseModel, create_model, Field
|
6 |
|
7 |
|
8 |
+
TPipeline = TypeVar("TPipeline", bound=type[Any])
|
9 |
+
T = TypeVar('T')
|
10 |
+
|
11 |
+
|
12 |
+
class ParamsModel(BaseModel):
|
13 |
+
"""Base model for pipeline parameters."""
|
14 |
+
|
15 |
+
@classmethod
|
16 |
+
def from_dict(cls, data: dict[str, Any]) -> 'ParamsModel':
|
17 |
+
"""Create a model instance from dictionary data."""
|
18 |
+
return cls.model_validate(data)
|
19 |
+
|
20 |
+
def to_dict(self) -> dict[str, Any]:
|
21 |
+
"""Convert model to dictionary."""
|
22 |
+
return self.model_dump()
|
23 |
+
|
24 |
+
|
25 |
+
def get_pipeline_class(pipeline_name: str) -> TPipeline:
|
26 |
try:
|
27 |
module = import_module(f"pipelines.{pipeline_name}")
|
28 |
except ModuleNotFoundError:
|
|
|
33 |
if pipeline_class is None:
|
34 |
raise ValueError(f"'Pipeline' class not found in module '{pipeline_name}'.")
|
35 |
|
36 |
+
# Type check to ensure we're returning a class
|
37 |
+
if not isinstance(pipeline_class, type):
|
38 |
+
raise TypeError(f"'Pipeline' in module '{pipeline_name}' is not a class")
|
39 |
+
|
40 |
return pipeline_class
|
41 |
|
42 |
|