Upload handler.py
Browse files- handler.py +9 -9
handler.py
CHANGED
@@ -5,10 +5,10 @@ from diffusers import FluxPipeline, FluxTransformer2DModel, AutoencoderKL, Torch
|
|
5 |
from PIL.Image import Image
|
6 |
import torch
|
7 |
|
8 |
-
|
9 |
-
|
10 |
|
11 |
-
|
12 |
|
13 |
def compile_pipeline(pipe) -> Any:
|
14 |
pipe.transformer.fuse_qkv_projections()
|
@@ -17,10 +17,10 @@ def compile_pipeline(pipe) -> Any:
|
|
17 |
return pipe
|
18 |
|
19 |
class EndpointHandler:
|
20 |
-
def __init__(self, path="
|
21 |
-
is_compile =
|
22 |
-
repo_id = "camenduru/FLUX.1-dev-diffusers"
|
23 |
-
|
24 |
dtype = torch.bfloat16
|
25 |
quantization_config = TorchAoConfig("int4dq")
|
26 |
vae = AutoencoderKL.from_pretrained(repo_id, subfolder="vae", torch_dtype=dtype)
|
@@ -29,9 +29,9 @@ class EndpointHandler:
|
|
29 |
if is_compile: self.pipeline = compile_pipeline(self.pipeline)
|
30 |
self.pipeline.to("cuda")
|
31 |
|
32 |
-
|
33 |
def __call__(self, data: Dict[str, Any]) -> Image:
|
34 |
-
|
35 |
|
36 |
if "inputs" in data and isinstance(data["inputs"], str):
|
37 |
prompt = data.pop("inputs")
|
|
|
5 |
from PIL.Image import Image
|
6 |
import torch
|
7 |
|
8 |
+
import torch._dynamo
|
9 |
+
torch._dynamo.config.suppress_errors = True
|
10 |
|
11 |
+
from huggingface_inference_toolkit.logging import logger
|
12 |
|
13 |
def compile_pipeline(pipe) -> Any:
|
14 |
pipe.transformer.fuse_qkv_projections()
|
|
|
17 |
return pipe
|
18 |
|
19 |
class EndpointHandler:
|
20 |
+
def __init__(self, path=""):
|
21 |
+
is_compile = True
|
22 |
+
#repo_id = "camenduru/FLUX.1-dev-diffusers"
|
23 |
+
repo_id = "NoMoreCopyright/FLUX.1-dev-test"
|
24 |
dtype = torch.bfloat16
|
25 |
quantization_config = TorchAoConfig("int4dq")
|
26 |
vae = AutoencoderKL.from_pretrained(repo_id, subfolder="vae", torch_dtype=dtype)
|
|
|
29 |
if is_compile: self.pipeline = compile_pipeline(self.pipeline)
|
30 |
self.pipeline.to("cuda")
|
31 |
|
32 |
+
@torch.inference_mode()
|
33 |
def __call__(self, data: Dict[str, Any]) -> Image:
|
34 |
+
logger.info(f"Received incoming request with {data=}")
|
35 |
|
36 |
if "inputs" in data and isinstance(data["inputs"], str):
|
37 |
prompt = data.pop("inputs")
|