prompt, from image working
Browse files
app.py
CHANGED
@@ -33,50 +33,21 @@ embedding_base64s = [None for i in range(max_tabs)]
|
|
33 |
|
34 |
|
35 |
def image_to_embedding(input_im):
|
36 |
-
|
37 |
-
|
38 |
-
transforms.ToTensor(),
|
39 |
-
transforms.Resize(
|
40 |
-
(336, 336),
|
41 |
-
interpolation=transforms.InterpolationMode.BICUBIC,
|
42 |
-
antialias=False,
|
43 |
-
),
|
44 |
-
transforms.Normalize(
|
45 |
-
[0.48145466, 0.4578275, 0.40821073],
|
46 |
-
[0.26862954, 0.26130258, 0.27577711]),
|
47 |
-
])
|
48 |
-
input = tform(input_im).to(device)
|
49 |
-
|
50 |
-
# approch B: convert input_im to torch
|
51 |
-
# inp = torch.from_numpy(np.array(input_im)).to(device)
|
52 |
-
# inp = torch.from_numpy(np.array(input_im)).permute(2, 0, 1).to(device)
|
53 |
-
|
54 |
-
# dtype = torch.float32
|
55 |
-
# input = input.to(device=device, dtype=dtype)
|
56 |
-
input = input.unsqueeze(0)
|
57 |
-
# image_embeddings = pipe.image_encoder(image).image_embeds
|
58 |
-
# image_embeddings = image_embeddings[0]
|
59 |
-
|
60 |
with torch.no_grad():
|
61 |
-
|
62 |
-
image_embeddings = model.get_image_features(input)
|
63 |
-
|
64 |
# image_embeddings /= image_embeddings.norm(dim=-1, keepdim=True)
|
65 |
-
image_embeddings_np = image_embeddings.cpu().detach().numpy()
|
66 |
return image_embeddings_np
|
67 |
|
68 |
def prompt_to_embedding(prompt):
|
69 |
-
|
70 |
-
inputs = processor(prompt, return_tensors="pt", padding='max_length', max_length=77)
|
71 |
-
# labels = torch.tensor(labels)
|
72 |
-
# prompt_tokens = inputs.input_ids[0]
|
73 |
-
prompt_tokens = inputs.input_ids
|
74 |
-
# image = inputs.pixel_values
|
75 |
with torch.no_grad():
|
76 |
-
prompt_embededdings = model.
|
77 |
# prompt_embededdings /= prompt_embededdings.norm(dim=-1, keepdim=True)
|
78 |
-
|
79 |
-
return
|
80 |
|
81 |
def embedding_to_image(embeddings):
|
82 |
size = math.ceil(math.sqrt(embeddings.shape[0]))
|
@@ -87,15 +58,15 @@ def embedding_to_image(embeddings):
|
|
87 |
|
88 |
def embedding_to_base64(embeddings):
|
89 |
import base64
|
90 |
-
# ensure
|
91 |
-
embeddings = embeddings.astype(np.
|
92 |
embeddings_b64 = base64.urlsafe_b64encode(embeddings).decode()
|
93 |
return embeddings_b64
|
94 |
|
95 |
def base64_to_embedding(embeddings_b64):
|
96 |
import base64
|
97 |
embeddings = base64.urlsafe_b64decode(embeddings_b64)
|
98 |
-
embeddings = np.frombuffer(embeddings, dtype=np.
|
99 |
# embeddings = torch.tensor(embeddings)
|
100 |
return embeddings
|
101 |
|
@@ -177,6 +148,9 @@ def update_average_embeddings(embedding_base64s_state, embedding_powers):
|
|
177 |
# TODO toggle this to support average or sum
|
178 |
final_embedding = final_embedding / num_embeddings
|
179 |
|
|
|
|
|
|
|
180 |
embeddings_b64 = embedding_to_base64(final_embedding)
|
181 |
return embeddings_b64
|
182 |
|
@@ -229,35 +203,12 @@ def on_example_image_click_set_image(input_image, image_url):
|
|
229 |
|
230 |
# device = torch.device("mps" if torch.backends.mps.is_available() else "cuda:0" if torch.cuda.is_available() else "cpu")
|
231 |
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
232 |
-
torch_size = torch.float16 if device == ('cuda') else torch.float32
|
233 |
-
# torch_size = torch.float32
|
234 |
-
# pipe = StableDiffusionPipeline.from_pretrained(
|
235 |
-
# model_id,
|
236 |
-
# custom_pipeline="pipeline.py",
|
237 |
-
# torch_dtype=torch_size,
|
238 |
-
# # , revision="fp16",
|
239 |
-
# requires_safety_checker = False, safety_checker=None,
|
240 |
-
# text_encoder = CLIPTextModel,
|
241 |
-
# tokenizer = CLIPTokenizer,
|
242 |
-
# )
|
243 |
-
# pipe = pipe.to(device)
|
244 |
-
|
245 |
-
from transformers import AutoProcessor, AutoModel
|
246 |
-
# processor = AutoProcessor.from_pretrained(clip_model_id)
|
247 |
-
# model = AutoModel.from_pretrained(clip_model_id)
|
248 |
-
# model = model.to(device)
|
249 |
|
250 |
from clip_retrieval.load_clip import load_clip, get_tokenizer
|
251 |
# model, preprocess = load_clip(clip_model, use_jit=True, device=device)
|
252 |
model, preprocess = load_clip(clip_model, use_jit=True, device=device)
|
253 |
tokenizer = get_tokenizer(clip_model)
|
254 |
|
255 |
-
test_url = "https://placekitten.com/400/600"
|
256 |
-
test_caption = "an image of a cat"
|
257 |
-
test_image_1 = "tests/test_clip_inference/test_images/123_456.jpg"
|
258 |
-
test_image_2 = "tests/test_clip_inference/test_images/416_264.jpg"
|
259 |
-
|
260 |
-
# clip_retrieval_service_url = "https://knn.laion.ai/knn-service"
|
261 |
clip_retrieval_client = ClipClient(
|
262 |
url=clip_retrieval_service_url,
|
263 |
indice_name=clip_model_id,
|
|
|
33 |
|
34 |
|
35 |
def image_to_embedding(input_im):
|
36 |
+
input_im = Image.fromarray(input_im)
|
37 |
+
prepro = preprocess(input_im).unsqueeze(0).to(device)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
with torch.no_grad():
|
39 |
+
image_embeddings = model.encode_image(prepro)
|
|
|
|
|
40 |
# image_embeddings /= image_embeddings.norm(dim=-1, keepdim=True)
|
41 |
+
image_embeddings_np = image_embeddings.cpu().to(torch.float32).detach().numpy()
|
42 |
return image_embeddings_np
|
43 |
|
44 |
def prompt_to_embedding(prompt):
|
45 |
+
text = tokenizer([prompt]).to(device)
|
|
|
|
|
|
|
|
|
|
|
46 |
with torch.no_grad():
|
47 |
+
prompt_embededdings = model.encode_text(text)
|
48 |
# prompt_embededdings /= prompt_embededdings.norm(dim=-1, keepdim=True)
|
49 |
+
prompt_embededdings_np = prompt_embededdings.cpu().to(torch.float32).detach().numpy()
|
50 |
+
return prompt_embededdings_np
|
51 |
|
52 |
def embedding_to_image(embeddings):
|
53 |
size = math.ceil(math.sqrt(embeddings.shape[0]))
|
|
|
58 |
|
59 |
def embedding_to_base64(embeddings):
|
60 |
import base64
|
61 |
+
# ensure float32
|
62 |
+
embeddings = embeddings.astype(np.float32)
|
63 |
embeddings_b64 = base64.urlsafe_b64encode(embeddings).decode()
|
64 |
return embeddings_b64
|
65 |
|
66 |
def base64_to_embedding(embeddings_b64):
|
67 |
import base64
|
68 |
embeddings = base64.urlsafe_b64decode(embeddings_b64)
|
69 |
+
embeddings = np.frombuffer(embeddings, dtype=np.float32)
|
70 |
# embeddings = torch.tensor(embeddings)
|
71 |
return embeddings
|
72 |
|
|
|
148 |
# TODO toggle this to support average or sum
|
149 |
final_embedding = final_embedding / num_embeddings
|
150 |
|
151 |
+
# normalize embeddings in numpy
|
152 |
+
final_embedding /= np.linalg.norm(final_embedding)
|
153 |
+
|
154 |
embeddings_b64 = embedding_to_base64(final_embedding)
|
155 |
return embeddings_b64
|
156 |
|
|
|
203 |
|
204 |
# device = torch.device("mps" if torch.backends.mps.is_available() else "cuda:0" if torch.cuda.is_available() else "cpu")
|
205 |
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
206 |
|
207 |
from clip_retrieval.load_clip import load_clip, get_tokenizer
|
208 |
# model, preprocess = load_clip(clip_model, use_jit=True, device=device)
|
209 |
model, preprocess = load_clip(clip_model, use_jit=True, device=device)
|
210 |
tokenizer = get_tokenizer(clip_model)
|
211 |
|
|
|
|
|
|
|
|
|
|
|
|
|
212 |
clip_retrieval_client = ClipClient(
|
213 |
url=clip_retrieval_service_url,
|
214 |
indice_name=clip_model_id,
|