Rename testcode.py to train.py
Browse files- testcode.py +0 -3
- train.py +53 -0
testcode.py
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
print("Hello Word")
|
2 |
-
print("test")
|
3 |
-
print("Hello Word")
|
|
|
|
|
|
|
|
train.py
ADDED
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# File 1: Model Repo Code (train.py)
|
2 |
+
# This file contains steps 1 to 4
|
3 |
+
|
4 |
+
from datasets import load_dataset
|
5 |
+
from transformers import AutoTokenizer, AutoModelForQuestionAnswering, TrainingArguments, Trainer
|
6 |
+
|
7 |
+
# Step 1: Load the Dataset
|
8 |
+
dataset = load_dataset("squad")
|
9 |
+
|
10 |
+
# Step 2: Preprocess the Dataset
|
11 |
+
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
|
12 |
+
|
13 |
+
def preprocess_function(examples):
|
14 |
+
return tokenizer(
|
15 |
+
examples["question"],
|
16 |
+
examples["context"],
|
17 |
+
truncation=True,
|
18 |
+
max_length=384,
|
19 |
+
stride=128,
|
20 |
+
return_overflowing_tokens=True,
|
21 |
+
padding="max_length"
|
22 |
+
)
|
23 |
+
|
24 |
+
tokenized_dataset = dataset.map(preprocess_function, batched=True)
|
25 |
+
|
26 |
+
# Step 3: Train the Model
|
27 |
+
model = AutoModelForQuestionAnswering.from_pretrained("bert-base-uncased")
|
28 |
+
|
29 |
+
training_args = TrainingArguments(
|
30 |
+
output_dir="./results",
|
31 |
+
evaluation_strategy="epoch",
|
32 |
+
learning_rate=3e-5,
|
33 |
+
per_device_train_batch_size=16,
|
34 |
+
num_train_epochs=3,
|
35 |
+
weight_decay=0.01,
|
36 |
+
push_to_hub=True, # Automatically push to the Hugging Face Hub
|
37 |
+
hub_model_id="username/qa_model_repo" # Replace with your username and model repo name
|
38 |
+
)
|
39 |
+
|
40 |
+
trainer = Trainer(
|
41 |
+
model=model,
|
42 |
+
args=training_args,
|
43 |
+
train_dataset=tokenized_dataset["train"],
|
44 |
+
eval_dataset=tokenized_dataset["validation"],
|
45 |
+
)
|
46 |
+
|
47 |
+
trainer.train()
|
48 |
+
|
49 |
+
# Step 4: Push the Model and Tokenizer to Hugging Face Hub
|
50 |
+
model.push_to_hub("username/qa_model_repo")
|
51 |
+
tokenizer.push_to_hub("username/qa_model_repo")
|
52 |
+
|
53 |
+
print("Model and tokenizer pushed to Hugging Face Hub successfully!")
|