{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5f93b7d1",
   "metadata": {},
   "outputs": [],
   "source": "from transformers import AutoModelForSeq2SeqLM\nfrom peft import get_peft_config, get_peft_model, get_peft_model_state_dict, PrefixTuningConfig, TaskType\nimport torch\nfrom datasets import load_dataset\nimport os\n\nos.environ[\"TOKENIZERS_PARALLELISM\"] = \"false\"\nfrom transformers import AutoTokenizer\nfrom torch.utils.data import DataLoader\nfrom transformers import default_data_collator, get_linear_schedule_with_warmup\nfrom tqdm import tqdm\nfrom datasets import load_dataset\n\ndevice = torch.accelerator.current_accelerator().type if hasattr(torch, \"accelerator\") else \"cuda\"\nmodel_name_or_path = \"t5-large\"\ntokenizer_name_or_path = \"t5-large\"\n\ntext_column = \"text\"\nlabel_column = \"text_label\"\nmax_length = 128\nlr = 1e-2\nnum_epochs = 5\nbatch_size = 8"
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "8d0850ac",
   "metadata": {},
   "outputs": [],
   "source": [
    "# creating model\n",
    "peft_config = PrefixTuningConfig(task_type=TaskType.SEQ_2_SEQ_LM, inference_mode=False, num_virtual_tokens=20)\n",
    "\n",
    "model = AutoModelForSeq2SeqLM.from_pretrained(model_name_or_path)\n",
    "model = get_peft_model(model, peft_config)\n",
    "model.print_trainable_parameters()\n",
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "4ee2babf",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using the latest cached version of the dataset since financial_phrasebank couldn't be found on the Hugging Face Hub\n",
      "Found the latest cached dataset configuration 'sentences_allagree' at /root/.cache/huggingface/datasets/financial_phrasebank/sentences_allagree/1.0.0/550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141 (last modified on Thu Jul 31 06:23:15 2025).\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3b321971d6f942418bd5ef6105a1aa65",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/2037 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5997543529a849bf97719e59a5ec95b2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Map:   0%|          | 0/227 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'sentence': \"Progress Group , QPR 's representative in Saudi Arabia and North Africa , has signed a framework agreement for a long term strategic relationship with ISE .\",\n",
       " 'label': 2,\n",
       " 'text_label': 'positive'}"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# loading dataset\n",
    "dataset = load_dataset(\"zeroshot/twitter-financial-news-sentiment\")\n",
    "dataset = dataset[\"train\"].train_test_split(test_size=0.1)\n",
    "dataset[\"validation\"] = dataset[\"test\"]\n",
    "del dataset[\"test\"]\n",
    "\n",
    "if hasattr(dataset[\"train\"].features[\"label\"], \"names\"):\n",
    "    classes = dataset[\"train\"].features[\"label\"].names\n",
    "else:\n",
    "    classes = [\"Bearish\", \"Bullish\", \"Neutral\"]\n",
    "dataset = dataset.map(\n",
    "    lambda x: {\"text_label\": [classes[label] for label in x[\"label\"]]},\n",
    "    batched=True,\n",
    "    num_proc=1,\n",
    ")\n",
    "\n",
    "dataset[\"train\"][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "adf9608c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ee9bf13a2e3f4812a51a87346a4614f3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "spiece.model:   0%|          | 0.00/792k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "332fcaa33dc343e7a20b24cec7ec97e9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "tokenizer.json:   0%|          | 0.00/1.39M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a2135262c7a44377b35fe32b8d86d6c6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer on dataset:   0%|          | 0/2037 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dc8598b160484939b27c65be001c694c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Running tokenizer on dataset:   0%|          | 0/227 [00:00<?, ? examples/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# data preprocessing\n",
    "tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)\n",
    "\n",
    "\n",
    "def preprocess_function(examples):\n",
    "    inputs = examples[text_column]\n",
    "    targets = examples[label_column]\n",
    "    model_inputs = tokenizer(inputs, max_length=max_length, padding=\"max_length\", truncation=True, return_tensors=\"pt\")\n",
    "    labels = tokenizer(targets, max_length=2, padding=\"max_length\", truncation=True, return_tensors=\"pt\")\n",
    "    labels = labels[\"input_ids\"]\n",
    "    labels[labels == tokenizer.pad_token_id] = -100\n",
    "    model_inputs[\"labels\"] = labels\n",
    "    return model_inputs\n",
    "\n",
    "\n",
    "processed_datasets = dataset.map(\n",
    "    preprocess_function,\n",
    "    batched=True,\n",
    "    num_proc=1,\n",
    "    remove_columns=dataset[\"train\"].column_names,\n",
    "    load_from_cache_file=False,\n",
    "    desc=\"Running tokenizer on dataset\",\n",
    ")\n",
    "\n",
    "train_dataset = processed_datasets[\"train\"]\n",
    "eval_dataset = processed_datasets[\"validation\"]\n",
    "\n",
    "train_dataloader = DataLoader(\n",
    "    train_dataset, shuffle=True, collate_fn=default_data_collator, batch_size=batch_size, pin_memory=True\n",
    ")\n",
    "eval_dataloader = DataLoader(eval_dataset, collate_fn=default_data_collator, batch_size=batch_size, pin_memory=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "f733a3c6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# optimizer and lr scheduler\n",
    "optimizer = torch.optim.AdamW(model.parameters(), lr=lr)\n",
    "lr_scheduler = get_linear_schedule_with_warmup(\n",
    "    optimizer=optimizer,\n",
    "    num_warmup_steps=0,\n",
    "    num_training_steps=(len(train_dataloader) * num_epochs),\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6b3a4090",
   "metadata": {},
   "outputs": [],
   "source": [
    "# training and evaluation\n",
    "model = model.to(device)\n",
    "\n",
    "for epoch in range(num_epochs):\n",
    "    model.train()\n",
    "    total_loss = 0\n",
    "    for step, batch in enumerate(tqdm(train_dataloader)):\n",
    "        batch = {k: v.to(device) for k, v in batch.items()}\n",
    "        outputs = model(**batch)\n",
    "        loss = outputs.loss\n",
    "        total_loss += loss.detach().float()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        lr_scheduler.step()\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "    model.eval()\n",
    "    eval_loss = 0\n",
    "    eval_preds = []\n",
    "    for step, batch in enumerate(tqdm(eval_dataloader)):\n",
    "        batch = {k: v.to(device) for k, v in batch.items()}\n",
    "        with torch.no_grad():\n",
    "            outputs = model(**batch)\n",
    "        loss = outputs.loss\n",
    "        eval_loss += loss.detach().float()\n",
    "        eval_preds.extend(\n",
    "            tokenizer.batch_decode(torch.argmax(outputs.logits, -1).detach().cpu().numpy(), skip_special_tokens=True)\n",
    "        )\n",
    "\n",
    "    eval_epoch_loss = eval_loss / len(eval_dataloader)\n",
    "    eval_ppl = torch.exp(eval_epoch_loss)\n",
    "    train_epoch_loss = total_loss / len(train_dataloader)\n",
    "    train_ppl = torch.exp(train_epoch_loss)\n",
    "    print(f\"{epoch=}: {train_ppl=} {train_epoch_loss=} {eval_ppl=} {eval_epoch_loss=}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6cafa67b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy=96.91629955947137 % on the evaluation dataset\n",
      "eval_preds[:10]=['negative', 'positive', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral']\n",
      "dataset['validation']['text_label'][:10]=['negative', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral', 'neutral']\n"
     ]
    }
   ],
   "source": [
    "# print accuracy\n",
    "correct = 0\n",
    "total = 0\n",
    "for pred, true in zip(eval_preds, dataset[\"validation\"][\"text_label\"]):\n",
    "    if pred.strip() == true.strip():\n",
    "        correct += 1\n",
    "    total += 1\n",
    "accuracy = correct / total * 100\n",
    "print(f\"{accuracy=} % on the evaluation dataset\")\n",
    "print(f\"{eval_preds[:10]=}\")\n",
    "print(f\"{dataset['validation']['text_label'][:10]=}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "a8de6005",
   "metadata": {},
   "outputs": [],
   "source": [
    "# saving model\n",
    "peft_model_id = f\"{model_name_or_path}_{peft_config.peft_type}_{peft_config.task_type}\"\n",
    "model.save_pretrained(peft_model_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bd20cd4c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3,8M\tt5-large_PREFIX_TUNING_SEQ_2_SEQ_LM/adapter_model.safetensors\r\n"
     ]
    }
   ],
   "source": [
    "ckpt = f\"{peft_model_id}/adapter_model.safetensors\"\n",
    "!du -h $ckpt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "76c2fc29",
   "metadata": {},
   "outputs": [],
   "source": [
    "from peft import PeftModel, PeftConfig\n",
    "\n",
    "peft_model_id = f\"{model_name_or_path}_{peft_config.peft_type}_{peft_config.task_type}\"\n",
    "\n",
    "config = PeftConfig.from_pretrained(peft_model_id)\n",
    "model = AutoModelForSeq2SeqLM.from_pretrained(config.base_model_name_or_path)\n",
    "model = PeftModel.from_pretrained(model, peft_model_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "d997f1cc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Acando AB ( ACANB SS ) fell 8.9 percent to 13.35 kronor , the lowest close since Dec. 11 .\n",
      "{'input_ids': tensor([[ 4292,   232,    32,     3,  5359,    41,     3, 22029, 14972,     3,\n",
      "          4256,     3,    61,  4728,  4848,  1298,  1093,    12,  8808,  2469,\n",
      "             3, 22318,    29,   127,     3,     6,     8,  7402,   885,   437,\n",
      "          4451,     5,   850,     3,     5,     1]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
      "         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])}\n",
      "tensor([[   0, 2841,    1]])\n",
      "['negative']\n"
     ]
    }
   ],
   "source": [
    "model.eval()\n",
    "i = 107\n",
    "inputs = tokenizer(dataset[\"validation\"][text_column][i], return_tensors=\"pt\")\n",
    "print(dataset[\"validation\"][text_column][i])\n",
    "print(inputs)\n",
    "\n",
    "with torch.no_grad():\n",
    "    outputs = model.generate(input_ids=inputs[\"input_ids\"], max_new_tokens=10)\n",
    "    print(outputs)\n",
    "    print(tokenizer.batch_decode(outputs.detach().cpu().numpy(), skip_special_tokens=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fb746c1e",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.13"
  },
  "vscode": {
   "interpreter": {
    "hash": "aee8b7b246df8f9039afb4144a1f6fd8d2ca17a180786b69acc140d282b71a49"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}