john
commited on
Commit
•
730ad9d
1
Parent(s):
b2f107f
learn how to training
Browse files- app.py +45 -8
- requirements.txt +3 -1
app.py
CHANGED
@@ -1,15 +1,52 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
-
pipeline = pipeline(task="image-classification", model="julien-c/hotdog-not-hotdog")
|
5 |
|
6 |
-
def predict(image):
|
7 |
-
predictions = pipeline(image)
|
8 |
-
return {p["label"]: p["score"] for p in predictions}
|
9 |
|
10 |
gr.Interface(
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
title="
|
15 |
).launch()
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline
|
3 |
+
from transformers import AutoTokenizer
|
4 |
+
from datasets import load_dataset
|
5 |
+
from transformers import DataCollatorWithPadding
|
6 |
+
|
7 |
+
raw_datasets = load_dataset("glue", "sst2")
|
8 |
+
raw_datasets
|
9 |
+
checkpoint = "bert-base-uncased"
|
10 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
11 |
+
def tokenize_function(example):
|
12 |
+
return tokenizer(example["sentence"], truncation=True)
|
13 |
+
|
14 |
+
tokenized_datasets = raw_datasets.map(tokenize_function, batched=True,remove_columns=['idx','sentence'])
|
15 |
+
tokenized_datasets
|
16 |
+
|
17 |
+
|
18 |
+
data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
|
19 |
+
|
20 |
+
from transformers import TrainingArguments
|
21 |
+
from transformers import AutoModelForSequenceClassification
|
22 |
+
from datasets import load_metric
|
23 |
+
from transformers import Trainer
|
24 |
+
import numpy as np
|
25 |
+
|
26 |
+
training_args = TrainingArguments("test-trainer", evaluation_strategy="epoch")# ѵ����Ҫ�IJ�����Ĭ�ϵ�
|
27 |
+
model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)
|
28 |
+
|
29 |
+
def compute_metrics(eval_preds):
|
30 |
+
metric = load_metric("glue", "sst2")
|
31 |
+
logits, labels = eval_preds
|
32 |
+
predictions = np.argmax(logits, axis=-1)
|
33 |
+
return metric.compute(predictions=predictions, references=labels)
|
34 |
+
|
35 |
+
trainer = Trainer(
|
36 |
+
model,
|
37 |
+
training_args,
|
38 |
+
train_dataset=tokenized_datasets["train"],
|
39 |
+
eval_dataset=tokenized_datasets["validation"],
|
40 |
+
data_collator=data_collator,
|
41 |
+
tokenizer=tokenizer,
|
42 |
+
compute_metrics=compute_metrics,
|
43 |
+
)
|
44 |
|
|
|
45 |
|
|
|
|
|
|
|
46 |
|
47 |
gr.Interface(
|
48 |
+
fn=trainer.train,
|
49 |
+
NONE,
|
50 |
+
NONE,
|
51 |
+
title="test",
|
52 |
).launch()
|
requirements.txt
CHANGED
@@ -1,2 +1,4 @@
|
|
1 |
transformers
|
2 |
-
|
|
|
|
|
|
1 |
transformers
|
2 |
+
streamlit
|
3 |
+
torch
|
4 |
+
datasets
|