Delete train.py
Browse files
train.py
DELETED
|
@@ -1,43 +0,0 @@
|
|
| 1 |
-
from transformers import Trainer, TrainingArguments, LlamaForCausalLM, LlamaTokenizer, AutoProcessor
|
| 2 |
-
from datasets import Dataset
|
| 3 |
-
import pandas as pd
|
| 4 |
-
import numpy as np
|
| 5 |
-
from tqdm import tqdm
|
| 6 |
-
|
| 7 |
-
def generate_test_candles(n_samples=50):
|
| 8 |
-
timestamps = pd.date_range(start="2024-01-01", periods=n_samples, freq="D")
|
| 9 |
-
data = {
|
| 10 |
-
"time": timestamps,
|
| 11 |
-
"open": np.random.uniform(100, 150, n_samples),
|
| 12 |
-
"high": np.random.uniform(150, 200, n_samples),
|
| 13 |
-
"low": np.random.uniform(50, 100, n_samples),
|
| 14 |
-
"close": np.random.uniform(100, 150, n_samples),
|
| 15 |
-
"volume": np.random.uniform(1000, 5000, n_samples)
|
| 16 |
-
}
|
| 17 |
-
return pd.DataFrame(data)
|
| 18 |
-
|
| 19 |
-
test_data = generate_test_candles()
|
| 20 |
-
dataset = Dataset.from_pandas(test_data)
|
| 21 |
-
|
| 22 |
-
training_args = TrainingArguments(
|
| 23 |
-
output_dir="./results",
|
| 24 |
-
num_train_epochs=3,
|
| 25 |
-
per_device_train_batch_size=4,
|
| 26 |
-
learning_rate=2e-4,
|
| 27 |
-
fp16=True,
|
| 28 |
-
save_steps=100,
|
| 29 |
-
logging_dir="./logs",
|
| 30 |
-
logging_steps=10
|
| 31 |
-
)
|
| 32 |
-
|
| 33 |
-
model = LlamaForCausalLM.from_pretrained("tmmdev2/llama-vision-charts")
|
| 34 |
-
tokenizer = LlamaTokenizer.from_pretrained("tmmdev2/llama-vision-charts")
|
| 35 |
-
|
| 36 |
-
trainer = Trainer(
|
| 37 |
-
model=model,
|
| 38 |
-
args=training_args,
|
| 39 |
-
train_dataset=dataset,
|
| 40 |
-
tokenizer=tokenizer
|
| 41 |
-
)
|
| 42 |
-
|
| 43 |
-
results = trainer.train()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|