File size: 8,165 Bytes
ec3e101
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
01959cc
ec3e101
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d5bca77
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
import gradio as gr
import math
import spacy
from datasets import load_dataset
from sentence_transformers import SentenceTransformer
from sentence_transformers import InputExample
from sentence_transformers import losses
from transformers import AutoTokenizer, AutoModel, AutoModelForSequenceClassification
from transformers import TrainingArguments, Trainer
import torch
import torch.nn.functional as F
from torch.utils.data import DataLoader
import numpy as np
import evaluate
import nltk
from nltk.corpus import stopwords
import subprocess
import sys
from transformers import DataCollatorWithPadding


# !pip install https://huggingface.co/spacy/en_core_web_sm/resolve/main/en_core_web_sm-any-py3-none-any.whl
# subprocess.check_call([sys.executable, '-m', 'pip', 'install', 'https://huggingface.co/spacy/en_core_web_sm/resolve/main/en_core_web_sm-any-py3-none-any.whl'])
# tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
# data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
# nltk.download('stopwords')
# nlp = spacy.load("en_core_web_sm")
# stops = stopwords.words("english")

# answer = "Pizza"
guesses = []
answer = "Pizza"


#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# def normalize(comment, lowercase, remove_stopwords):
#     if lowercase:
#         comment = comment.lower()
#     comment = nlp(comment)
#     lemmatized = list()
#     for word in comment:
#         lemma = word.lemma_.strip()
#         if lemma:
#             if not remove_stopwords or (remove_stopwords and lemma not in stops):
#                 lemmatized.append(lemma)
#     return " ".join(lemmatized)


# def tokenize_function(examples):
#     return tokenizer(examples["text"], truncation=True)


def compute_metrics(eval_pred):
    logits, labels = eval_pred
    predictions = np.argmax(logits, axis=-1)
    metric = evaluate.load("accuracy")
    return metric.compute(predictions=predictions, references=labels)


def training():
    dataset_id = "ag_news"
    
    print("GETTING DATASET")
    dataset = load_dataset(dataset_id)
    # dataset = dataset["train"]
    # tokenized_datasets = dataset.map(tokenize_function, batched=True)
    
    print(f"- The {dataset_id} dataset has {dataset['train'].num_rows} examples.")
    print(f"- Each example is a {type(dataset['train'][0])} with a {type(dataset['train'][0]['text'])} as value.")
    print(f"- Examples look like this: {dataset['train'][0]}")
    
    # small_train_dataset = tokenized_datasets["train"].shuffle(seed=42).select(range(1000))
    # small_eval_dataset = tokenized_datasets["test"].shuffle(seed=42).select(range(1000))
    
    # dataset = dataset["train"].map(tokenize_function, batched=True)
    # dataset.set_format(type="torch", columns=["input_ids", "token_type_ids", "attention_mask", "label"])
    # dataset.format['type']
    
    # tokenized_news = dataset.map(tokenize_function, batched=True)
    
    # model = AutoModelForSequenceClassification.from_pretrained("sentence-transformers/all-MiniLM-L6-v2", num_labels=2)
    
    # print(dataset)
    
    train_examples = []
    train_data = dataset["train"]
    # For agility we only 1/2 of our available data
    n_examples = dataset["train"].num_rows // 2
    
    for i in range(n_examples):
        example = train_data[i]
        # example_opposite = dataset_clean[-(i)]
        # print(example["text"])
        train_examples.append(InputExample(texts=[example['text']], label=example['label']))
    
        
    train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=25)
    
    print("END DATALOADER")
    
    # print(train_examples)
        
    embeddings = finetune(train_dataloader)
    
    return (dataset['train'].num_rows, type(dataset['train'][0]), type(dataset['train'][0]['text']), dataset['train'][0], embeddings)


def finetune(train_dataloader):
    # model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased", num_labels=5)
    model_id = "sentence-transformers/all-MiniLM-L6-v2"
    model = SentenceTransformer(model_id)
    device = torch.device('cuda:0')
    model = model.to(device)
    
    # training_args = TrainingArguments(output_dir="test_trainer")
    
    # USE THIS LINK
    # https://huggingface.co/blog/how-to-train-sentence-transformers
    
    train_loss = losses.BatchHardSoftMarginTripletLoss(model=model)
    
    print("BEGIN FIT")
    
    model.fit(train_objectives=[(train_dataloader, train_loss)], epochs=10)
    
    model.save("ag_news_model")
    
    model.save_to_hub("smhavens/all-MiniLM-agNews")
    # accuracy = compute_metrics(eval, metric)
    
    # training_args = TrainingArguments(output_dir="test_trainer", evaluation_strategy="epoch")
    
    # trainer = Trainer(
    #     model=model,
    #     args=training_args,
    #     train_dataset=train,
    #     eval_dataset=eval,
    #     compute_metrics=compute_metrics,
    # )
    
    # trainer.train()
    
    # sentences = ["This is an example sentence", "Each sentence is converted"]

    # # model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
    # embeddings = model.encode(sentences)
    # print(embeddings)
    
    # # Sentences we want sentence embeddings for
    # sentences = ['This is an example sentence', 'Each sentence is converted']

    # # Load model from HuggingFace Hub
    # # tokenizer = AutoTokenizer.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')
    # # model = AutoModel.from_pretrained('sentence-transformers/all-MiniLM-L6-v2')

    # # Tokenize sentences
    # encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

    # # Compute token embeddings
    # with torch.no_grad():
    #     model_output = model(**encoded_input)

    # # Perform pooling
    # sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

    # # Normalize embeddings
    # sentence_embeddings = F.normalize(sentence_embeddings, p=2, dim=1)

    # print("Sentence embeddings:")
    # print(sentence_embeddings)
    return 0

 

def greet(name):
    return "Hello " + name + "!!"

def check_answer(guess:str):
    global guesses
    global answer
    guesses.append(guess)
    output = ""
    for guess in guesses:
        output += ("- " + guess + "\n")
    output = output[:-1]
    
    if guess.lower() == answer.lower():
        return "Correct!", output
    else:
        return "Try again!", output

def main():
    print("BEGIN")
    word1 = "Black"
    word2 = "White"
    word3 = "Sun"
    global answer
    answer = "Moon"
    global guesses
    
    num_rows, data_type, value, example, embeddings = training()
    
    # prompt = f"{word1} is to {word2} as {word3} is to ____"
    # with gr.Blocks() as iface:
    #     gr.Markdown(prompt)
    #     with gr.Tab("Guess"):
    #         text_input = gr.Textbox()
    #         text_output = gr.Textbox()
    #         text_button = gr.Button("Submit")
    #     with gr.Accordion("Open for previous guesses"):
    #         text_guesses = gr.Textbox()
    #     with gr.Tab("Testing"):
    #         gr.Markdown(f"""Number of rows in dataset is {num_rows}, with each having type {data_type} and value {value}.
    #                     An example is {example}.
    #                     The Embeddings are {embeddings}.""")
    #     text_button.click(check_answer, inputs=[text_input], outputs=[text_output, text_guesses])
    # # iface = gr.Interface(fn=greet, inputs="text", outputs="text")
    # iface.launch()
    
    


    
if __name__ == "__main__":
    main()