Update README.md
Browse files
README.md
CHANGED
|
@@ -19,10 +19,13 @@ This is an encoder-decoder model that was trained on various information extract
|
|
| 19 |
First of all, initialize the model:
|
| 20 |
```python
|
| 21 |
from transformers import T5Tokenizer, T5ForConditionalGeneration
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
|
| 24 |
|
| 25 |
-
model = T5ForConditionalGeneration.from_pretrained("knowledgator/t5-for-ie")
|
| 26 |
```
|
| 27 |
|
| 28 |
You need to set a prompt and put it with text to the model, below are examples of how to use it for different tasks:
|
|
@@ -30,7 +33,7 @@ You need to set a prompt and put it with text to the model, below are examples o
|
|
| 30 |
**named entity recognition**
|
| 31 |
```python
|
| 32 |
input_text = "Extract entity types from the text: <e1>Kyiv</e1> is the capital of <e2>Ukraine</e2>."
|
| 33 |
-
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
|
| 34 |
|
| 35 |
outputs = model.generate(input_ids)
|
| 36 |
print(tokenizer.decode(outputs[0]))
|
|
@@ -39,7 +42,7 @@ print(tokenizer.decode(outputs[0]))
|
|
| 39 |
**text classification**
|
| 40 |
```python
|
| 41 |
input_text = "Classify the following text into the most relevant categories: Kyiv is the capital of Ukraine"
|
| 42 |
-
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
|
| 43 |
|
| 44 |
outputs = model.generate(input_ids)
|
| 45 |
print(tokenizer.decode(outputs[0]))
|
|
@@ -48,7 +51,7 @@ print(tokenizer.decode(outputs[0]))
|
|
| 48 |
**relation extraction**
|
| 49 |
```python
|
| 50 |
input_text = "Extract relations between entities in the text: <e1>Kyiv</e1> is the capital of <e2>Ukraine</e2>."
|
| 51 |
-
input_ids = tokenizer(input_text, return_tensors="pt").input_ids
|
| 52 |
|
| 53 |
outputs = model.generate(input_ids)
|
| 54 |
print(tokenizer.decode(outputs[0]))
|
|
@@ -78,7 +81,7 @@ classifier = TextClassifier(
|
|
| 78 |
labels=['default'],
|
| 79 |
model=model,
|
| 80 |
tokenizer=tokenizer,
|
| 81 |
-
device=
|
| 82 |
)
|
| 83 |
classifier.initialize_labels_trie(labels)
|
| 84 |
|
|
|
|
| 19 |
First of all, initialize the model:
|
| 20 |
```python
|
| 21 |
from transformers import T5Tokenizer, T5ForConditionalGeneration
|
| 22 |
+
import torch
|
| 23 |
+
|
| 24 |
+
device = torch.device("cuda") if torch.cuda.is_available() else torch.device('cpu')
|
| 25 |
|
| 26 |
tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-large")
|
| 27 |
|
| 28 |
+
model = T5ForConditionalGeneration.from_pretrained("knowledgator/t5-for-ie").to(device)
|
| 29 |
```
|
| 30 |
|
| 31 |
You need to set a prompt and put it with text to the model, below are examples of how to use it for different tasks:
|
|
|
|
| 33 |
**named entity recognition**
|
| 34 |
```python
|
| 35 |
input_text = "Extract entity types from the text: <e1>Kyiv</e1> is the capital of <e2>Ukraine</e2>."
|
| 36 |
+
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
|
| 37 |
|
| 38 |
outputs = model.generate(input_ids)
|
| 39 |
print(tokenizer.decode(outputs[0]))
|
|
|
|
| 42 |
**text classification**
|
| 43 |
```python
|
| 44 |
input_text = "Classify the following text into the most relevant categories: Kyiv is the capital of Ukraine"
|
| 45 |
+
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
|
| 46 |
|
| 47 |
outputs = model.generate(input_ids)
|
| 48 |
print(tokenizer.decode(outputs[0]))
|
|
|
|
| 51 |
**relation extraction**
|
| 52 |
```python
|
| 53 |
input_text = "Extract relations between entities in the text: <e1>Kyiv</e1> is the capital of <e2>Ukraine</e2>."
|
| 54 |
+
input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
|
| 55 |
|
| 56 |
outputs = model.generate(input_ids)
|
| 57 |
print(tokenizer.decode(outputs[0]))
|
|
|
|
| 81 |
labels=['default'],
|
| 82 |
model=model,
|
| 83 |
tokenizer=tokenizer,
|
| 84 |
+
device=device #if cuda
|
| 85 |
)
|
| 86 |
classifier.initialize_labels_trie(labels)
|
| 87 |
|