File size: 1,798 Bytes
a1c81a0 |
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 |
from transformers import T5TokenizerFast, T5ForConditionalGeneration, Trainer, TrainingArguments, DataCollatorForSeq2Seq
from datasets import load_dataset, Dataset
import os
import json
# === Load your custom tokenizer ===
tokenizer = T5TokenizerFast.from_pretrained("minicoderx-tokenizer")
# === Load or create dataset ===
def load_jsonl(path):
with open(path) as f:
data = [json.loads(line) for line in f]
return Dataset.from_dict({
"input": [x["input"] for x in data],
"output": [x["output"] for x in data]
})
dataset = load_jsonl("data/train.jsonl")
# === Tokenize dataset ===
def tokenize(batch):
return tokenizer(batch["input"], padding="max_length", truncation=True, max_length=128)
def tokenize_labels(batch):
labels = tokenizer(batch["output"], padding="max_length", truncation=True, max_length=128)
batch["labels"] = labels["input_ids"]
return batch
dataset = dataset.map(tokenize)
dataset = dataset.map(tokenize_labels)
# === Load pre-trained T5-small ===
model = T5ForConditionalGeneration.from_pretrained("t5-small")
# === Training configuration ===
training_args = TrainingArguments(
output_dir="minicoderx-model",
per_device_train_batch_size=4,
num_train_epochs=3,
logging_steps=10,
save_strategy="epoch",
evaluation_strategy="no",
save_total_limit=2,
fp16=True,
overwrite_output_dir=True,
)
data_collator = DataCollatorForSeq2Seq(tokenizer, model=model)
trainer = Trainer(
model=model,
args=training_args,
train_dataset=dataset,
data_collator=data_collator,
tokenizer=tokenizer
)
# === Train ===
trainer.train()
# === Save model ===
trainer.save_model("minicoderx-model")
tokenizer.save_pretrained("minicoderx-model")
print("Training complete and model saved.") |