Create flan_t5_finetuned_model

MODEL_NAME: google/flan-t5-base

flan_t5_finetuned_model

@@ -0,0 +1,92 @@
+# -*- coding: utf-8 -*-
+"""Untitled4.ipynb
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/19SAJcA_N4eQVyeNjT1iFdgpyLvvtSSEw
+"""
+
+!pip install transformers datasets accelerate -q
+
+from google.colab import files
+uploaded = files.upload()
+
+from datasets import Dataset
+from transformers import T5Tokenizer, T5ForConditionalGeneration, Trainer, TrainingArguments, DataCollatorForSeq2Seq, Seq2SeqTrainingArguments
+import pandas as pd
+import torch
+
+# Load CSV file (adjust filename if needed)
+df = pd.read_csv("flan_t5_true_false_dataset.csv")
+
+# Convert to Hugging Face Dataset
+dataset = Dataset.from_pandas(df)
+
+# Load tokenizer and model
+model_name = "google/flan-t5-base"
+tokenizer = T5Tokenizer.from_pretrained(model_name)
+model = T5ForConditionalGeneration.from_pretrained(model_name)
+
+# Preprocessing
+def preprocess(example):
+    inputs = tokenizer(example["input"], padding="max_length", truncation=True, max_length=256)
+    with tokenizer.as_target_tokenizer():
+        labels = tokenizer(example["output"], padding="max_length", truncation=True, max_length=64)
+    inputs["labels"] = labels["input_ids"]
+    return inputs
+
+# Tokenize dataset
+tokenized_dataset = dataset.map(preprocess, batched=True)
+
+# Define training arguments
+training_args = Seq2SeqTrainingArguments(
+    output_dir="./flan_t5_finetuned_model",
+    per_device_train_batch_size=4,
+    per_device_eval_batch_size=4, # Added evaluation batch size
+    num_train_epochs=3,
+    save_steps=500,
+    logging_steps=100,
+    save_total_limit=1,
+    fp16=torch.cuda.is_available()
+)
+
+# Trainer setup
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=tokenized_dataset,
+    tokenizer=tokenizer,
+    data_collator=DataCollatorForSeq2Seq(tokenizer, model)
+)
+
+# Start training
+trainer.train()
+
+!zip -r flan_t5_finetuned_model.zip flan_t5_finetuned_model
+files.download("flan_t5_finetuned_model.zip")
+
+import pandas as pd
+
+data = [
+    {
+        "input": f"Convert this fact into a true/false question: The moon is made of cheese {i}.",
+        "output": f"The moon is made of cheese {i}. True or False?"
+    }
+    for i in range(150)
+]
+
+df = pd.DataFrame(data)
+df.to_csv("flan_t5_eval.csv", index=False)
+
+from google.colab import files
+files.download('flan_t5_eval.csv')
+
+!pip install transformers datasets bert-score sentence-transformers -q
+
+from google.colab import files
+uploaded = files.upload()
+
+EVAL_CSV = "/content/flan_t5_eval.csv"
+
+!ls -l ./flan_t5_finetuned