Thomas Male
commited on
Commit
·
f22523d
1
Parent(s):
58558bd
Upload handler.py
Browse filesCreated custom Handler
- handler.py +80 -0
handler.py
ADDED
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Dict, List, Any
|
2 |
+
import torch
|
3 |
+
from torch import autocast
|
4 |
+
from tqdm.auto import tqdm
|
5 |
+
from point_e.diffusion.configs import DIFFUSION_CONFIGS, diffusion_from_config
|
6 |
+
from point_e.diffusion.sampler import PointCloudSampler
|
7 |
+
from point_e.models.download import load_checkpoint
|
8 |
+
from point_e.models.configs import MODEL_CONFIGS, model_from_config
|
9 |
+
from point_e.util.plotting import plot_point_cloud
|
10 |
+
import base64
|
11 |
+
from io import BytesIO
|
12 |
+
|
13 |
+
|
14 |
+
# set device
|
15 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
16 |
+
|
17 |
+
if device.type != 'cuda':
|
18 |
+
raise ValueError("need to run on GPU")
|
19 |
+
|
20 |
+
class EndpointHandler():
|
21 |
+
def __init__(self, path=""):
|
22 |
+
# load the optimized model
|
23 |
+
print('creating base model...')
|
24 |
+
self.base_name = 'base40M-textvec'
|
25 |
+
self.base_model = model_from_config(MODEL_CONFIGS[self.base_name], device)
|
26 |
+
self.base_model.eval()
|
27 |
+
self.base_diffusion = diffusion_from_config(DIFFUSION_CONFIGS[self.base_name])
|
28 |
+
|
29 |
+
print('creating upsample model...')
|
30 |
+
self.upsampler_model = model_from_config(MODEL_CONFIGS['upsample'], device)
|
31 |
+
self.upsampler_model.eval()
|
32 |
+
self.upsampler_diffusion = diffusion_from_config(DIFFUSION_CONFIGS['upsample'])
|
33 |
+
|
34 |
+
print('downloading base checkpoint...')
|
35 |
+
self.base_model.load_state_dict(load_checkpoint(self.base_name, device))
|
36 |
+
|
37 |
+
print('downloading upsampler checkpoint...')
|
38 |
+
self.upsampler_model.load_state_dict(load_checkpoint('upsample', device))
|
39 |
+
|
40 |
+
def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
|
41 |
+
"""
|
42 |
+
Args:
|
43 |
+
data (:obj:):
|
44 |
+
includes the input data and the parameters for the inference.
|
45 |
+
Return:
|
46 |
+
A :obj:`dict`:. plotly json Data
|
47 |
+
"""
|
48 |
+
inputs = data.pop("inputs", data)
|
49 |
+
|
50 |
+
sampler = PointCloudSampler(
|
51 |
+
device=device,
|
52 |
+
models=[self.base_model,self.upsampler_model],
|
53 |
+
diffusions=[self.base_diffusion, self.upsampler_diffusion],
|
54 |
+
num_points=[1024, 4096 - 1024],
|
55 |
+
aux_channels=['R', 'G', 'B'],
|
56 |
+
guidance_scale=[3.0, 0.0],
|
57 |
+
model_kwargs_key_filter=('texts', ''), # Do not condition the upsampler at all
|
58 |
+
)
|
59 |
+
|
60 |
+
# Set a test prompt to condition on.
|
61 |
+
# prompt = 'A bluebird mid-flight'
|
62 |
+
|
63 |
+
# run inference pipeline
|
64 |
+
with autocast(device.type):
|
65 |
+
samples = None
|
66 |
+
for x in tqdm(sampler.sample_batch_progressive(batch_size=1, model_kwargs=dict(texts=[inputs]))):
|
67 |
+
samples = x
|
68 |
+
#image = self.pipe(inputs, guidance_scale=7.5)["sample"][0]
|
69 |
+
|
70 |
+
pc = sampler.output_to_point_clouds(samples)[0]
|
71 |
+
return {"data": pc}
|
72 |
+
#print(pc)
|
73 |
+
|
74 |
+
# encode image as base 64
|
75 |
+
#buffered = BytesIO()
|
76 |
+
#image.save(buffered, format="JPEG")
|
77 |
+
#img_str = base64.b64encode(buffered.getvalue())
|
78 |
+
|
79 |
+
# postprocess the prediction
|
80 |
+
#return {"image": img_str.decode()}
|