ably.do/allegro.py

import os
os.environ["TOKENIZERS_PARALLELISM"] = "false"

import torch
import weaviate
from weaviate.classes.config import Property, DataType, Configure
from weaviate.classes.query import Query
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, TrainingArguments, Trainer, DataCollatorForSeq2Seq
from datasets import Dataset

# 1️⃣ Połączenie z bazą Weaviate
client = weaviate.WeaviateClient(
    connection_params=weaviate.ConnectionParams.from_params(
        http_host="weaviate",
        http_port=8080,
        http_secure=False,
        grpc_host="weaviate",
        grpc_port=50051,
        grpc_secure=False,
    )
)

# 2️⃣ Pobranie dokumentów z bazy Weaviate
collection_name = "Documents"
query = Query(collection_name).limit(1000)
result = client.query.run(query)

documents = []
file_names = []

for item in result[collection_name]['objects']:
    documents.append(item['properties']['content'])
    file_names.append(item['properties']['fileName'])

# 3️⃣ Tworzenie datasetu
training_data = {
    "text": documents,
    "file_name": file_names
}
dataset = Dataset.from_dict(training_data)

# Podział na zestaw treningowy i ewaluacyjny
split_dataset = dataset.train_test_split(test_size=0.25)
train_dataset = split_dataset["train"]
eval_dataset = split_dataset["test"]

# 4️⃣ Ładowanie modelu Multislav
model_name = "allegro/multislav-5lang"
device = "cuda" if torch.cuda.is_available() else "cpu"
model = AutoModelForSeq2SeqLM.from_pretrained(model_name).to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# 5️⃣ Tokenizacja
max_length = 512
def tokenize_function(examples):
    return tokenizer(
        examples["text"],
        padding="max_length",
        truncation=True,
        max_length=max_length
    )

tokenized_train = train_dataset.map(tokenize_function, batched=True)
tokenized_eval = eval_dataset.map(tokenize_function, batched=True)

# 6️⃣ Parametry treningu
training_args = TrainingArguments(
    output_dir="./results",
    evaluation_strategy="steps",
    eval_steps=500,
    save_strategy="steps",
    save_steps=500,
    learning_rate=2e-5,
    per_device_train_batch_size=4,
    per_device_eval_batch_size=4,
    num_train_epochs=10,
    weight_decay=0.01,
    load_best_model_at_end=True,
    metric_for_best_model="loss",
    greater_is_better=False,
)

# 7️⃣ Data Collator
data_collator = DataCollatorForSeq2Seq(tokenizer=tokenizer, model=model)

# 8️⃣ Trening
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_train,
    eval_dataset=tokenized_eval,
    data_collator=data_collator,
)

trainer.train()

# 9️⃣ Zapis modelu
model.save_pretrained("./trained_model/multislav")
tokenizer.save_pretrained("./trained_model/multislav")

print("✅ Model został wytrenowany i zapisany!")