Skip to content

Commit 69ab9f9

Browse files
ovalle15ovalle15
authored
added how to create data rows with embeddings (#1582)
- Deleted the datasets.ipynb notebook - Updated custom embeddings notebook - update langchain notebook (removed duplicate imports)
1 parent 3793a9d commit 69ab9f9

File tree

4 files changed

+164
-269
lines changed

4 files changed

+164
-269
lines changed

‎examples/README.md

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,6 @@
1414
| Data Rows | [Github](basics/data_rows.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/data_rows.ipynb) |
1515
| Data Row Metadata | [Github](basics/data_row_metadata.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/data_row_metadata.ipynb) |
1616
| Custom Embeddings | [Github](basics/custom_embeddings_sdk.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/custom_embeddings_sdk.ipynb) |
17-
| Datasets | [Github](basics/datasets.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/datasets.ipynb) |
1817
| Export data | [Github](exports/export_data.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/exports/export_data.ipynb) |
1918
| Ontologies | [Github](basics/ontologies.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/ontologies.ipynb) |
2019
| Projects | [Github](basics/projects.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/Labelbox/labelbox-python/blob/develop/examples/basics/projects.ipynb) |

‎examples/basics/custom_embeddings_sdk.ipynb

Lines changed: 159 additions & 50 deletions
Original file line numberDiff line numberDiff line change
@@ -39,14 +39,14 @@
3939
{
4040
"metadata": {},
4141
"source": [
42-
"# Setup"
42+
"# Set up "
4343
],
4444
"cell_type": "markdown"
4545
},
4646
{
4747
"metadata": {},
4848
"source": [
49-
"!pip3 install -q \"labelbox\""
49+
"!pip3 install -q \"labelbox[data]\""
5050
],
5151
"cell_type": "code",
5252
"outputs": [],
@@ -57,12 +57,21 @@
5757
"source": [
5858
"import labelbox as lb\n",
5959
"import numpy as np\n",
60-
"import json"
60+
"import json\n",
61+
"import uuid\n",
62+
"import random"
6163
],
6264
"cell_type": "code",
6365
"outputs": [],
6466
"execution_count": null
6567
},
68+
{
69+
"metadata": {},
70+
"source": [
71+
"# Replace with your API key"
72+
],
73+
"cell_type": "markdown"
74+
},
6675
{
6776
"metadata": {},
6877
"source": [
@@ -76,19 +85,31 @@
7685
{
7786
"metadata": {},
7887
"source": [
79-
"# Select data rows in Labelbox for custom embeddings"
88+
"# Select data rows"
8089
],
8190
"cell_type": "markdown"
8291
},
8392
{
8493
"metadata": {},
8594
"source": [
86-
"client.enable_experimental = True\n",
87-
"\n",
88-
"# get images from a Labelbox dataset\n",
89-
"# Our systems start to process data after 1000 embeddings of each type, for this demo make sure your dataset is over 1000 data rows\n",
90-
"dataset = client.get_dataset(\"<DATASET-ID>\")\n",
91-
"\n",
95+
"- Get images from a Labelbox dataset\n",
96+
"- To improve similarity search, you need to upload custom embeddings to at least 1,000 data rows.\n"
97+
],
98+
"cell_type": "markdown"
99+
},
100+
{
101+
"metadata": {},
102+
"source": [
103+
"DATASET_ID = \"\""
104+
],
105+
"cell_type": "code",
106+
"outputs": [],
107+
"execution_count": null
108+
},
109+
{
110+
"metadata": {},
111+
"source": [
112+
"dataset = client.get_dataset(dataset_id=DATASET_ID)\n",
92113
"export_task = dataset.export()\n",
93114
"export_task.wait_till_done()"
94115
],
@@ -124,8 +145,8 @@
124145
{
125146
"metadata": {},
126147
"source": [
127-
"data_row_ids = [dr[\"data_row\"][\"id\"] for dr in data_rows]\n",
128-
"data_row_ids = data_row_ids[:1000] # keep the first 1000 examples for the sake of this demo"
148+
"data_row_dict = [{\"data_row_id\": dr[\"data_row\"][\"id\"], \"row_data\": dr[\"data_row\"][\"row_data\"]} for dr in data_rows]\n",
149+
"data_row_dict = data_row_dict[:1000] # keep the first 1000 examples for the sake of this demo"
129150
],
130151
"cell_type": "code",
131152
"outputs": [],
@@ -134,19 +155,22 @@
134155
{
135156
"metadata": {},
136157
"source": [
137-
"# Create the payload for custom embeddings\n",
138-
"-- It should be a .ndjson file. \n",
139-
"-- Every line is a json file that finishes with a \\n character. \n",
140-
"-- It does not have to be created through Python. "
158+
"# Create custom embedding payload "
141159
],
142160
"cell_type": "markdown"
143161
},
144162
{
145163
"metadata": {},
146164
"source": [
147-
"nb_data_rows = len(data_row_ids)\n",
165+
"Generate random vectors for embeddings (max : 2048 dimensions)"
166+
],
167+
"cell_type": "markdown"
168+
},
169+
{
170+
"metadata": {},
171+
"source": [
172+
"nb_data_rows = len(data_row_dict)\n",
148173
"print(\"Number of data rows: \", nb_data_rows)\n",
149-
"# Generate random vectors, of dimension 2048 each\n",
150174
"# Labelbox supports custom embedding vectors of dimension up to 2048\n",
151175
"custom_embeddings = [list(np.random.random(2048)) for _ in range(nb_data_rows)]"
152176
],
@@ -157,12 +181,14 @@
157181
{
158182
"metadata": {},
159183
"source": [
160-
"# Create the payload for custom embeddings\n",
161-
"payload = []\n",
162-
"for data_row_id,custom_embedding in zip(data_row_ids,custom_embeddings):\n",
163-
" payload.append({\"id\": data_row_id, \"vector\": custom_embedding})\n",
164-
"\n",
165-
"print('payload', len(payload),payload[:1])"
184+
"List all custom embeddings available in your Labelbox workspace"
185+
],
186+
"cell_type": "markdown"
187+
},
188+
{
189+
"metadata": {},
190+
"source": [
191+
"embeddings = client.get_embeddings()"
166192
],
167193
"cell_type": "code",
168194
"outputs": [],
@@ -171,16 +197,15 @@
171197
{
172198
"metadata": {},
173199
"source": [
174-
"# Delete any pre-existing file\n",
175-
"import os\n",
176-
"if os.path.exists(\"payload.ndjson\"):\n",
177-
" os.remove(\"payload.ndjson\")\n",
178-
"\n",
179-
"# Convert the payload to a JSON file\n",
180-
"with open('payload.ndjson', 'w') as f:\n",
181-
" for p in payload:\n",
182-
" f.write(json.dumps(p) + \"\\n\")\n",
183-
" # sanity_check_payload = json.dump(payload, f)"
200+
"Choose an existing embedding type or create a new one"
201+
],
202+
"cell_type": "markdown"
203+
},
204+
{
205+
"metadata": {},
206+
"source": [
207+
"# Name of the custom embedding must be unique\n",
208+
"embedding = client.create_embedding(\"my_custom_embedding_2048_dimensions\", 2048)"
184209
],
185210
"cell_type": "code",
186211
"outputs": [],
@@ -189,10 +214,26 @@
189214
{
190215
"metadata": {},
191216
"source": [
192-
"# Sanity check that you can read/load the file and the payload is correct\n",
193-
"with open('payload.ndjson') as f:\n",
194-
" sanity_check_payload = [json.loads(l) for l in f.readlines()]\n",
195-
"print(\"Nb of custom embedding vectors in sanity_check_payload: \", len(sanity_check_payload))"
217+
"Create payload"
218+
],
219+
"cell_type": "markdown"
220+
},
221+
{
222+
"metadata": {},
223+
"source": [
224+
"- The payload should encompass all the data you wish to retain, along with the new embeddings vector data.\n",
225+
"- `row_data` and `key` is required when using `dataset.upsert_data_rows()` "
226+
],
227+
"cell_type": "markdown"
228+
},
229+
{
230+
"metadata": {},
231+
"source": [
232+
"payload = []\n",
233+
"for data_row_dict, custom_embedding in zip(data_row_dict,custom_embeddings):\n",
234+
" payload.append({\"key\": lb.UniqueId(data_row_dict['data_row_id']), \"row_data\": data_row_dict['row_data'], \"embeddings\": [{\"embedding_id\": embedding.id, \"vector\": custom_embedding}]})\n",
235+
"\n",
236+
"print('payload', len(payload),payload[:1])"
196237
],
197238
"cell_type": "code",
198239
"outputs": [],
@@ -201,8 +242,24 @@
201242
{
202243
"metadata": {},
203244
"source": [
204-
"# See all custom embeddings available in your Labelbox workspace\n",
205-
"embeddings = client.get_embeddings()"
245+
"# Upload payload"
246+
],
247+
"cell_type": "markdown"
248+
},
249+
{
250+
"metadata": {},
251+
"source": [
252+
"Upsert data rows with custom embeddings"
253+
],
254+
"cell_type": "markdown"
255+
},
256+
{
257+
"metadata": {},
258+
"source": [
259+
"task = dataset.upsert_data_rows(payload)\n",
260+
"task.wait_till_done()\n",
261+
"print(task.errors)\n",
262+
"print(task.status)"
206263
],
207264
"cell_type": "code",
208265
"outputs": [],
@@ -211,9 +268,16 @@
211268
{
212269
"metadata": {},
213270
"source": [
214-
"# Create a new custom embedding, unless you want to re-use one\n",
215-
"# Name of the custom embedding must be unique\n",
216-
"embedding = client.create_embedding(\"my_custom_embedding_2048_dimensions\", 2048)"
271+
"Get the count of imported vectors for a custom embedding"
272+
],
273+
"cell_type": "markdown"
274+
},
275+
{
276+
"metadata": {},
277+
"source": [
278+
"# Count how many data rows have a specific custom embedding (this can take a couple of minutes)\n",
279+
"count = embedding.get_imported_vector_count()\n",
280+
"print(count)"
217281
],
218282
"cell_type": "code",
219283
"outputs": [],
@@ -222,7 +286,13 @@
222286
{
223287
"metadata": {},
224288
"source": [
225-
"# Delete a custom embedding\n",
289+
"Delete custom embedding type"
290+
],
291+
"cell_type": "markdown"
292+
},
293+
{
294+
"metadata": {},
295+
"source": [
226296
"#embedding.delete()"
227297
],
228298
"cell_type": "code",
@@ -232,15 +302,22 @@
232302
{
233303
"metadata": {},
234304
"source": [
235-
"# Upload the payload to Labelbox"
305+
"# Upload custom embeddings during data row creation"
306+
],
307+
"cell_type": "markdown"
308+
},
309+
{
310+
"metadata": {},
311+
"source": [
312+
"Create a dataset"
236313
],
237314
"cell_type": "markdown"
238315
},
239316
{
240317
"metadata": {},
241318
"source": [
242-
"# Replace the current id with the newly generated id from the previous step, or any existing custom embedding id\n",
243-
"embedding.import_vectors_from_file(\"./payload.ndjson\")"
319+
"# Create a dataset\n",
320+
"dataset_new = client.create_dataset(name=\"data_rows_with_embeddings\")"
244321
],
245322
"cell_type": "code",
246323
"outputs": [],
@@ -249,15 +326,47 @@
249326
{
250327
"metadata": {},
251328
"source": [
252-
"# Get the count of imported vectors for a custom embedding"
329+
"Fetch an embedding (2048 dimension)"
253330
],
254331
"cell_type": "markdown"
255332
},
256333
{
257334
"metadata": {},
258335
"source": [
259-
"# Count how many data rows have a specific custom embedding (this can take a couple of minutes)\n",
260-
"count = embedding.get_imported_vector_count()"
336+
"embedding = client.get_embedding_by_name(\"my_custom_embedding_2048_dimensions\")\n",
337+
"vector = [random.uniform(1.0, 2.0) for _ in range(embedding.dims)]"
338+
],
339+
"cell_type": "code",
340+
"outputs": [],
341+
"execution_count": null
342+
},
343+
{
344+
"metadata": {},
345+
"source": [
346+
"Upload data rows with embeddings"
347+
],
348+
"cell_type": "markdown"
349+
},
350+
{
351+
"metadata": {},
352+
"source": [
353+
"\n",
354+
"uploads = []\n",
355+
"# Generate data rows\n",
356+
"for i in range(1,9):\n",
357+
" uploads.append({\n",
358+
" \"row_data\": f\"https://storage.googleapis.com/labelbox-datasets/People_Clothing_Segmentation/jpeg_images/IMAGES/img_000{i}.jpeg\",\n",
359+
" \"global_key\": \"TEST-ID-%id\" % uuid.uuid1(),\n",
360+
" \"embeddings\": [{\n",
361+
" \"embedding_id\": embedding.id,\n",
362+
" \"vector\": vector\n",
363+
" }]\n",
364+
" })\n",
365+
"\n",
366+
"task1 = dataset_new.create_data_rows(uploads)\n",
367+
"task1.wait_till_done()\n",
368+
"print(\"ERRORS: \" , task1.errors)\n",
369+
"print(\"RESULTS:\" , task1.result)"
261370
],
262371
"cell_type": "code",
263372
"outputs": [],

0 commit comments

Comments
 (0)