NEARCoderEval.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "53584db8",
   "metadata": {},
   "source": [
    "# NEARCoder evaluation\n",
    "\n",
    "NEARCoder, the Web3 Code LLM for the NEAR Protocol, is built on StarCoder2-3B. In this notebook, we are evaluating NEARCoder and its intermediary building steps in a structured testing protocol. Our primary hypothesis is that our final model doesn't lose performance to general coding tasks compared to the base model, while it increases its performance on specialized coding tasks related to the NEAR Protocol.\n",
    "\n",
    "#### 1. Evaluation framework\n",
    "\n",
    "We will evaluate the following four models:\n",
    "- [StarCoder2-3B](https://huggingface.co/bigcode/starcoder2-3b): The base model.\n",
    "- [NEAR-preTrainedStarCoder2](https://huggingface.co/jcarbonnell/NEAR-preTrainedStarCoder2): StarCoder2-3B pre-trained on NEAR-specific data.\n",
    "- [NEAR-structTunedStarCoder2](https://huggingface.co/jcarbonnell/NEAR-structTunedStarcoder2): Structurally tuned version of the pre-trained model.\n",
    "- [NEARCoder-3b](https://huggingface.co/jcarbonnell/NEARCoder-3b): The final model.\n",
    "\n",
    "On the following four tasks:\n",
    "- [HumanEvalPack](https://huggingface.co/datasets/bigcode/humanevalpack): an extension of OpenAI's general code challenges HumanEval with Javascript and Rust programming languages.\n",
    "- NEAR Protocol Questions: Specific NEAR-related questions.\n",
    "- dApp Trees Generation: Generating NEAR dApp structures with comments.\n",
    "- Basic NEAR dApp Development: Building a simple NEAR dApp (e.g. an event calendar on the NEAR blockchain).\n",
    "\n",
    "#### 2. Evaluation Metrics\n",
    "\n",
    "Each task will have specific metrics to ensure a comprehensive evaluation:\n",
    "\n",
    "- HumanEvalPack: BLEU score.\n",
    "- NEAR Protocol Questions: BLEU, ROUGE-1.\n",
    "- dApp Trees Generation: Structural similarity, comment quality.\n",
    "- Basic NEAR dApp Development: Functionality, usability, and code quality.\n",
    "\n",
    "#### 3. Comparative Analysis\n",
    "\n",
    "After evaluating each model on each tasks, we will perform a comparative analysis to identify:\n",
    "\n",
    "- Performance trends across models.\n",
    "- Areas where the NEARCoder excels or needs improvement.\n",
    "- Any trade-offs between general coding performance and NEAR-specific capabilities."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "f0f199f7",
   "metadata": {},
   "outputs": [],
   "source": [
    "#!pip install datasets transformers evaluate rouge_score sentence-transformers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "914f0df3",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to\n",
      "[nltk_data]     /Users/juliencarbonnell/nltk_data...\n",
      "[nltk_data]   Package punkt is already up-to-date!\n",
      "[nltk_data] Downloading package averaged_perceptron_tagger to\n",
      "[nltk_data]     /Users/juliencarbonnell/nltk_data...\n",
      "[nltk_data]   Package averaged_perceptron_tagger is already up-to-\n",
      "[nltk_data]       date!\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using MPS device: mps\n"
     ]
    }
   ],
   "source": [
    "# import libraries\n",
    "import os\n",
    "import glob\n",
    "import difflib\n",
    "import re\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import torch\n",
    "import random\n",
    "import evaluate\n",
    "import gc\n",
    "import psutil\n",
    "import json\n",
    "import nltk\n",
    "from nltk.tokenize import word_tokenize\n",
    "nltk.download('punkt')\n",
    "nltk.download('averaged_perceptron_tagger')\n",
    "\n",
    "from datasets import load_dataset, concatenate_datasets\n",
    "from sentence_transformers import SentenceTransformer, util\n",
    "from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline\n",
    "\n",
    "from evaluate.visualization import radar_plot\n",
    "\n",
    "device = torch.device(\"mps\") if torch.backends.mps.is_available() else torch.device(\"cpu\")\n",
    "print (\"Using MPS device:\", device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b0542aa",
   "metadata": {},
   "source": [
    "## 1. Set up the evaluation environment\n",
    "\n",
    "- Prepare the evaluation datasets\n",
    "- Load the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b47e5c33",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['task_id', 'prompt', 'declaration', 'canonical_solution',\n",
      "       'buggy_solution', 'bug_type', 'failure_symptoms', 'entry_point',\n",
      "       'import', 'test_setup', 'test', 'example_test', 'signature',\n",
      "       'docstring', 'instruction'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "# Load general coding challenges for JavaScript (js) and Rust (rust)\n",
    "js_dataset = load_dataset(\"bigcode/humanevalpack\", \"js\", trust_remote_code=True)[\"test\"]\n",
    "rs_dataset = load_dataset(\"bigcode/humanevalpack\", \"rust\", trust_remote_code=True)[\"test\"]\n",
    "# select first 5 rows\n",
    "js_dataset = js_dataset[:5]\n",
    "rs_dataset = rs_dataset[:5]\n",
    "js_df = pd.DataFrame(js_dataset)\n",
    "rs_df = pd.DataFrame(rs_dataset)\n",
    "#concatenate them\n",
    "human_eval_ds = pd.concat([js_df, rs_df], ignore_index=True)\n",
    "\n",
    "# NEAR Protocol questions\n",
    "near_protocol_questions = [\n",
    "    \"What is the Nightshade Sharding in NEAR Protocol?\",\n",
    "    \"Explain the role of validators in the NEAR Protocol.\",\n",
    "    \"How does NEAR handle transactions and smart contract execution?\",\n",
    "    \"What are the advantages of NEAR's Proof of Stake consensus mechanism?\",\n",
    "    \"Describe how NEAR's Rainbow Bridge works for cross-chain communication.\",\n",
    "    \"What is the Aurora Engine on NEAR Protocol?\"\n",
    "]\n",
    "\n",
    "# extract 2 examples of dApp trees from the original dataset\n",
    "dapp_trees = load_dataset('jcarbonnell/NEARdAppsPrompts')['validation']\n",
    "# Define a function to filter trees by length (number of lines and characters)\n",
    "def filter_by_length(dataset, max_lines=20, max_chars=1000):\n",
    "    filtered_entries = []\n",
    "    for entry in dataset:\n",
    "        tree = entry['tree']\n",
    "        lines = tree.split('\\n')\n",
    "        if len(lines) <= max_lines and len(tree) <= max_chars:\n",
    "            filtered_entries.append(entry)\n",
    "    return filtered_entries\n",
    "\n",
    "# Filter the dataset based on tree length\n",
    "filtered_dapp_trees = filter_by_length(dapp_trees)\n",
    "\n",
    "# If you need only a few examples, select the first N from filtered entries\n",
    "N = 2\n",
    "filtered_dapp_trees_limited = filtered_dapp_trees[:N]\n",
    "\n",
    "# Transform the filtered entries into the desired format\n",
    "transformed_dapp_trees = []\n",
    "for entry in filtered_dapp_trees_limited:\n",
    "    user_prompt = entry['user_prompt']\n",
    "    tree = entry['tree']\n",
    "    transformed_dapp_trees.append({\n",
    "        \"user_prompt\": user_prompt,\n",
    "        \"tree\": tree\n",
    "    })\n",
    "\n",
    "dapp_trees = transformed_dapp_trees\n",
    "\n",
    "# Example of dApps\n",
    "dapp_projects = [\n",
    "    {\n",
    "        \"name\": \"NEAR Calendar App\",\n",
    "        \"description\": \"A simple calendar dApp on NEAR for scheduling events.\",\n",
    "        \"requirements\": [\n",
    "            \"Users should be able to create, edit, and delete events.\",\n",
    "            \"Events should be stored on the NEAR blockchain.\",\n",
    "            \"Provide a user-friendly interface for managing events.\"\n",
    "        ]\n",
    "    },\n",
    "    {\n",
    "        \"name\": \"NEAR Snake Game\",\n",
    "        \"description\": \"A basic snake game dApp on NEAR.\",\n",
    "        \"requirements\": [\n",
    "            \"Implement a classic snake game.\",\n",
    "            \"Store high scores on the NEAR blockchain.\",\n",
    "            \"Provide an interface to view and reset high scores.\"\n",
    "        ]\n",
    "    }\n",
    "]\n",
    "\n",
    "# Define evaluation datasets\n",
    "eval_datasets = {\n",
    "    \"HumanEval\": human_eval_ds,\n",
    "    \"NEAR Protocol Questions\": near_protocol_questions,\n",
    "    \"dApp Trees Generation\": dapp_trees,\n",
    "    \"Basic NEAR dApp Development\": dapp_projects\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "25814b6e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model and tokenizer for StarCoder2-3B from bigcode/starcoder2-3b\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/anaconda3/lib/python3.11/site-packages/huggingface_hub/file_download.py:1132: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model and tokenizer for NEAR-preTrainedStarCoder2 from jcarbonnell/NEAR-preTrainedStarCoder2\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "585002eb62304ea1a2e764346afe0da5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model and tokenizer for NEAR-structTunedStarCoder2 from jcarbonnell/NEAR-structTunedStarCoder2\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a6c9617538ef4326af9bf4b8dd07290a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model and tokenizer for NEARCoder-3b from jcarbonnell/NEARCoder-3b\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "026f3030753e446181b3b7a40fa3b9fc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "All models and tokenizers loaded successfully.\n"
     ]
    }
   ],
   "source": [
    "# List of model names\n",
    "model_names = {\n",
    "    \"StarCoder2-3B\": \"bigcode/starcoder2-3b\",\n",
    "    \"NEAR-preTrainedStarCoder2\": \"jcarbonnell/NEAR-preTrainedStarCoder2\",\n",
    "    \"NEAR-structTunedStarCoder2\": \"jcarbonnell/NEAR-structTunedStarCoder2\",\n",
    "    \"NEARCoder-3b\": \"jcarbonnell/NEARCoder-3b\"\n",
    "}\n",
    "\n",
    "# Dictionary to store tokenizers and models\n",
    "models = {}\n",
    "for model_name, model_path in model_names.items():\n",
    "    print(f\"Loading model and tokenizer for {model_name} from {model_path}\")\n",
    "    # load with memory optimization\n",
    "    tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)\n",
    "    model = AutoModelForCausalLM.from_pretrained(model_path, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16)\n",
    "    models[model_name] = {\n",
    "        \"tokenizer\": tokenizer,\n",
    "        \"model\": model\n",
    "    }\n",
    "\n",
    "print(\"All models and tokenizers loaded successfully.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d366c94",
   "metadata": {},
   "source": [
    "## 2. Run evaluations\n",
    "\n",
    "- Define evaluations for each task\n",
    "- Execute evaluations across all models\n",
    "- Collect and record results"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b66ac9c9",
   "metadata": {},
   "source": [
    "#### General code challenges : Human Eval"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8ec052c8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# normalize code for fair BLEU score\n",
    "def normalize_code(code):\n",
    "    # Remove comments, extra spaces, etc.\n",
    "    code = re.sub(r'\\s+', ' ', code)  # Collapse whitespace\n",
    "    code = re.sub(r'//.*', '', code)  # Remove single line comments\n",
    "    code = re.sub(r'/\\*.*?\\*/', '', code, flags=re.DOTALL)  # Remove block comments\n",
    "    code = code.strip()\n",
    "    return code.replace('\\n', ' ').strip()\n",
    "\n",
    "# Define evaluation functions for HumanEval\n",
    "def evaluate_human_eval(model, tokenizer, dataset):\n",
    "    # Initialize the evaluation pipeline and BLEU scorer\n",
    "    text_generation_pipeline = pipeline(\"text-generation\", model=model, tokenizer=tokenizer)\n",
    "    bleu = evaluate.load('bleu')\n",
    "    # List to store evaluation metrics\n",
    "    evaluation_metrics = []\n",
    "    \n",
    "    # Iterate over each row in the DataFrame\n",
    "    for index, challenge in dataset.iterrows():\n",
    "        # Generate code from the prompt\n",
    "        generated_code = text_generation_pipeline(\n",
    "            challenge[\"prompt\"],\n",
    "            max_length=1024,\n",
    "            num_return_sequences=1,\n",
    "            truncation=False,\n",
    "            pad_token_id=tokenizer.eos_token_id\n",
    "        )[0][\"generated_text\"]\n",
    "        \n",
    "        # Normalize generated and reference code\n",
    "        generated_code = normalize_code(generated_code)\n",
    "        reference = normalize_code(challenge[\"canonical_solution\"])\n",
    "        \n",
    "        # Compute the BLEU score\n",
    "        bleu_score = bleu.compute(predictions=[generated_code], references=[[reference]])['bleu']\n",
    "        # Append the BLEU score to the evaluation metrics list\n",
    "        evaluation_metrics.append(bleu_score)\n",
    "        \n",
    "        # Print information about the challenge and BLEU score\n",
    "        print(f\"Challenge Prompt: {challenge['prompt']}\")\n",
    "        print(f\"Generated Code: {generated_code}\")\n",
    "        print(f\"Canonical Solution: {reference}\")\n",
    "        print(f\"BLEU Score: {bleu_score:.2f}\\n\")\n",
    "\n",
    "    return evaluation_metrics\n",
    "\n",
    "# Define a function to save evaluation results\n",
    "def save_evaluation_results(results, file_path):\n",
    "    results.to_csv(file_path, index=False)\n",
    "\n",
    "# Define the path to save/load evaluation results\n",
    "human_eval_results_path = \"human_eval_results.csv\"\n",
    "\n",
    "# Define a dictionary to store evaluation results\n",
    "evaluation_results = {}\n",
    "\n",
    "# Iterate over each model\n",
    "for model_name, model_data in models.items():\n",
    "    print(f\"Evaluating model: {model_name}\")\n",
    "    model = model_data[\"model\"]\n",
    "    tokenizer = model_data[\"tokenizer\"]\n",
    "\n",
    "    # Evaluate the model on the HumanEval task\n",
    "    evaluation_metrics = evaluate_human_eval(model, tokenizer, eval_datasets[\"HumanEval\"])\n",
    "\n",
    "    # Store the evaluation results in the dictionary\n",
    "    evaluation_results[model_name] = evaluation_metrics\n",
    "\n",
    "# Convert evaluation results to a DataFrame\n",
    "all_results = []\n",
    "for model_name, metrics in evaluation_results.items():\n",
    "    for metric in metrics:\n",
    "        #metric['model'] = model_name\n",
    "        all_results.append(metric)\n",
    "\n",
    "evaluation_df = pd.DataFrame(all_results)\n",
    "\n",
    "# Save the evaluation results\n",
    "save_evaluation_results(evaluation_df, human_eval_results_path)\n",
    "\n",
    "# Print or export evaluation results\n",
    "print(evaluation_df)\n",
    "\n",
    "# Clear cache after saving results\n",
    "del model, tokenizer, evaluation_df, evaluation_results, all_results\n",
    "gc.collect()\n",
    "\n",
    "# If using psutil to monitor memory usage\n",
    "print(f\"Memory usage after clearing: {psutil.virtual_memory().percent}%\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f50f6dfe",
   "metadata": {},
   "source": [
    "#### NEAR Protocol general knowledge:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "19b6b494",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded 1136 reference texts.\n"
     ]
    }
   ],
   "source": [
    "reference_files_path = '/Users/juliencarbonnell/opencampus/nearDataPretraining'\n",
    "\n",
    "# Load reference corpus function\n",
    "def load_reference_corpus(file_path_pattern):\n",
    "    reference_texts = []\n",
    "    for file_path in glob.glob(file_path_pattern):\n",
    "        with open(file_path, 'r', encoding='utf-8') as file:\n",
    "            reference_texts.append(file.read())\n",
    "    return reference_texts\n",
    "\n",
    "# Load reference texts\n",
    "reference_texts = load_reference_corpus(reference_files_path + \"/*.txt\")\n",
    "\n",
    "# Print the number of references loaded\n",
    "print(f\"Loaded {len(reference_texts)} reference texts.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "0001138c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'StarCoder2-3B': ['What is the Nightshade Sharding in NEAR Protocol?\\n\\nThe Nightshade Sharding is a new sharding protocol that is designed to be compatible with the NEAR Protocol.\\n\\n## What is the Nightshade Sharding?\\n\\nThe Nightshade Sharding is a new sharding protocol that is designed to be compatible with the NEAR Protocol.\\n\\n## What is the Nightshade Sharding?\\n\\nThe Nightshade Sharding is a new sharding protocol that is designed to be compatible with the NEAR Protocol.\\n\\n## What is the Nightshade Sharding?\\n\\nThe Nightshade Sharding is a new sharding protocol that is designed to be compatible with the NEAR Protocol.\\n\\n## What is', 'Explain the role of validators in the NEAR Protocol.\\n\\nValidators are the nodes that participate in the consensus process. Validators are responsible for validating transactions and blocks. Validators are also responsible for maintaining the state of the blockchain. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for maintaining the list of validators. Validators are also responsible for', 'How does NEAR handle transactions and smart contract execution?\\n\\nThe NEAR protocol is a blockchain-based platform that allows developers to build and deploy smart contracts on the NEAR platform. The NEAR platform is a decentralized network that allows developers to build and deploy smart contracts on the NEAR platform.\\n\\nThe NEAR platform is a blockchain-based platform that allows developers to build and deploy smart contracts on the NEAR platform. The NEAR platform is a decentralized network that allows developers to build and deploy smart contracts on the NEAR platform.\\n\\nThe NEAR platform is a blockchain-based platform that allows developers to build and deploy smart contracts on the NEAR platform. The NEAR platform is a decentral', \"What are the advantages of NEAR's Proof of Stake consensus mechanism?\\n\\n## What is the NEAR protocol?\\n\\nNEAR is a blockchain protocol that allows users to store and transfer data, and to interact with smart contracts.\\n\\n## What is the NEAR protocol?\\n\\nNEAR is a blockchain protocol that allows users to store and transfer data, and to interact with smart contracts.\\n\\n## What is the NEAR protocol?\\n\\nNEAR is a blockchain protocol that allows users to store and transfer data, and to interact with smart contracts.\\n\\n## What is the NEAR protocol?\\n\\nNEAR is a blockchain protocol that allows users to store and transfer data, and to interact with smart contracts.\", \"Describe how NEAR's Rainbow Bridge works for cross-chain communication.\\n\\n## Rainbow Bridge\\n\\nThe Rainbow Bridge is a bridge that allows NEAR to communicate with other blockchains.\\n\\n### Rainbow Bridge Protocol\\n\\nThe Rainbow Bridge protocol is a set of rules that allow NEAR to communicate with other blockchains.\\n\\nThe Rainbow Bridge protocol is based on the [Ethereum 2.0 specification](https://github.com/ethereum/consensus-specs/blob/dev/specs/phase0/beacon-chain.md).\\n\\nThe Rainbow Bridge protocol is based on the [Ethereum 2.0 specification](https://github.com\", 'What is the Aurora Engine on NEAR Protocol?\\n\\nThe Aurora Engine is a smart contract that allows users to create and manage their own Aurora tokens.\\n\\n## What is Aurora?\\n\\nAurora is a new cryptocurrency that is being developed by the Aurora Foundation. Aurora is a decentralized, open-source, and permissionless cryptocurrency that is built on the NEAR Protocol.\\n\\n## What is NEAR Protocol?\\n\\nNEAR Protocol is a decentralized, permissionless, and open-source blockchain that is built on the Ethereum Virtual Machine (EVM).\\n\\n## What is the Aurora Foundation?\\n\\nThe A'], 'NEAR-preTrainedStarCoder2': ['What is the Nightshade Sharding in NEAR Protocol?\\n---\\n\\nNightshade is a sharding solution for NEAR Protocol. It allows developers to shard their blockchain without having to change their code.\\n\\n## What is Nightshade?\\n\\nNightshade is a sharding solution for NEAR Protocol. It allows developers to shard their blockchain without having to change their code.\\n\\n## Why Nightshade?\\n\\nNightshade is a sharding solution for NEAR Protocol. It allows developers to shard their blockchain without having to change their code.\\n\\n## How Nightshade Works\\n\\nNightshade is a sharding solution for NEAR Protocol. It allows developers to shard their blockchain without having to change their code.', 'Explain the role of validators in the NEAR Protocol.\\n\\n## Validators\\n\\nValidators are the entities that maintain the state of the blockchain. They are responsible for ensuring that the state of the blockchain is consistent and secure. They are also responsible for ensuring that the network is functioning properly.\\n\\nValidators are responsible for maintaining the following:\\n\\n- Providing the necessary security and reliability for the network\\n- Maintaining the integrity of the state of the blockchain\\n- Ensuring that the network is functioning properly\\n\\nValidators are also responsible for ensuring that the network is secure and reliable. They are responsible for:\\n\\n- Providing the necessary security and reliability for the network\\n- Maintaining the integrity of the', 'How does NEAR handle transactions and smart contract execution?\\n---\\n\\nNEAR is a decentralized network that uses a [transaction-based model](https://docs.near.org/concepts/transactions) to process transactions and execute smart contracts. This means that transactions are processed in a sequential manner, and smart contracts are executed in a parallel manner.\\n\\nTransactions are processed by a [transaction pool](https://docs.near.org/concepts/transactions/transaction-pool), which is a queue of transactions that are waiting to be processed. The transaction pool is responsible for validating transactions and ensuring that they are valid before they are processed.\\n\\nWhen a transaction is received by the network, it is added to the transaction pool.', \"What are the advantages of NEAR's Proof of Stake consensus mechanism?\\n\\nThe Proof of Stake consensus mechanism is a consensus mechanism that uses the stake of the validators to determine the order of transactions. This mechanism is designed to prevent malicious actors from creating invalid transactions and causing the network to be unstable.\\n\\nThe Proof of Stake consensus mechanism has several advantages over other consensus mechanisms:\\n\\n- It is more secure than other consensus mechanisms because it uses the stake of the validators to determine the order of transactions. This means that validators who have more stake will have more influence over the network and will be more likely to be selected as validators.\\n- It is more efficient than other consensus mechanisms because\", \"Describe how NEAR's Rainbow Bridge works for cross-chain communication.\\n\\n## Rainbow Bridge\\n\\nThe Rainbow Bridge is a bridge between NEAR and Ethereum. It allows users to send and receive NEAR tokens on Ethereum.\\n\\nThe Rainbow Bridge is a decentralized application that allows users to send and receive NEAR tokens on Ethereum. It is built on NEAR Protocol and uses the Ethereum Virtual Machine (EVM) as its backend.\\n\\nThe Rainbow Bridge is designed to be compatible with existing Ethereum applications and protocols. It allows users to send and receive NEAR tokens on Ethereum without changing their existing Ethereum applications or protocols.\\n\\n## Rainbow Bridge Features\\n\\nThe\", 'What is the Aurora Engine on NEAR Protocol?\\n---\\n\\nThe Aurora Engine is a highly scalable, low-latency, and high-throughput blockchain. It is designed to be compatible with the NEAR Protocol, allowing developers to build decentralized applications and applications that are compatible with the NEAR Protocol.\\n\\n## What is the Aurora Engine?\\n\\nThe Aurora Engine is a highly scalable, low-latency, and high-throughput blockchain. It is designed to be compatible with the NEAR Protocol, allowing developers to build decentralized applications and applications that are compatible with the NEAR Protocol.\\n\\n## How does the Aurora Engine work?\\n\\nThe'], 'NEAR-structTunedStarCoder2': ['What is the Nightshade Sharding in NEAR Protocol?\\n---\\n\\nNightshade is a sharding protocol that is designed to be compatible with the [near-core](https://github.com/near/nearcore) runtime.\\n\\n## Nightshade Sharding\\n\\nNightshade sharding is a sharding protocol that is designed to be compatible with the [near-core](https://github.com/near/nearcore) runtime.\\n\\nNightshade sharding is a sharding protocol that is designed to be compatible with the [near-core](https://github.com/near/nearcore) runtime.\\n\\nNightshade sharding is a sharding protocol that is designed to be compatible with the [', 'Explain the role of validators in the NEAR Protocol.\\n\\n## Validator\\n\\nA validator is a node that runs a full node and participates in consensus.\\n\\n## Validator Set\\n\\nThe validator set is the set of validators that are currently active.\\n\\n## Validator Set Size\\n\\nThe validator set size is the number of validators that are currently active.\\n\\n## Validator Set Shard\\n\\nThe validator set shard is the shard that the validator set belongs to.\\n\\n## Validator Set Shard Size\\n\\nThe validator set shard size is the number of validators in the validator set shard.\\n\\n## Validator Set Shard Index\\n\\nThe validator set shard index is the index of the validator set shard.', 'How does NEAR handle transactions and smart contract execution?\\n---\\n\\nNEAR uses a [transaction-based model](https://nomicon.io/Standards/Tokens/FungibleToken/Transaction) for fungible token transfers.\\n\\nTransactions are signed by the sender, and the receiver can verify that the transaction was signed by the sender.\\n\\nThe transaction is then sent to the network, and the network will verify that the transaction is valid.\\n\\nIf the transaction is valid, the network will then execute the transaction.\\n\\nThe transaction is executed by the [runtime](https://nomicon.io/Runtime) which is responsible for:\\n\\n- Validating the transaction\\n- Executing the transaction', \"What are the advantages of NEAR's Proof of Stake consensus mechanism?\\n---\\n\\n## Advantages of Proof of Stake\\n\\n- **Security**: The security of the network is based on the fact that the network is run by a group of validators who are selected by the community. The validators are selected by the community based on their stake. The more stake they have, the more likely they are to be selected. This means that the more stake a validator has, the more likely they are to be elected. This is a very important property of the network.\\n- **Efficiency**: The more stake a validator has, the more time they have to wait to be selected. This means that the more stake\", \"Describe how NEAR's Rainbow Bridge works for cross-chain communication.\\n\\n## Rainbow Bridge\\n\\nThe Rainbow Bridge is a bridge between Ethereum and NEAR. It allows Ethereum contracts to call NEAR functions and vice versa.\\n\\n### Rainbow Bridge Contract\\n\\nThe Rainbow Bridge contract is deployed to `rainbow.near`.\\n\\n### Rainbow Bridge Interface\\n\\n```rust\\n// RainbowBridge.near\\n\\n// RainbowBridge.near/get_account_id\\n// Returns the account_id of the given public key.\\n//\\n// Args:\\n//   public_key: The public key to lookup.\\n//\\n// Returns\", 'What is the Aurora Engine on NEAR Protocol?\\n---\\n\\nThe Aurora Engine is a [runtime](https://docs.aurora.dev/docs/runtime) that runs on the [Aurora Network](https://aurora.dev/aurora-network).\\n\\nThe Aurora Engine is a [Rust](https://www.rust-lang.org/) implementation of the [Runtime Interface](https://github.com/near-examples/runtime-interface) that is used by [near-sdk-rs](https://github.com/near/near-sdk-rs).\\n\\n## Why is it called Aurora?\\n\\nThe Aurora Network is a ['], 'NEARCoder-3b': ['What is the Nightshade Sharding in NEAR Protocol?\\n\\nThe Nightshade Sharding is a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding will be a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding will be a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding will be a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding will be a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding will be a new sharding protocol that will be deployed in the NEAR Protocol. The Nightshade Sharding', 'Explain the role of validators in the NEAR Protocol.\\n\\nValidators are the people who are responsible for keeping the network secure. They are responsible for ensuring that the network is always running smoothly', 'How does NEAR handle transactions and smart contract execution?\\n\\nNEAR uses a Tendermint consensus engine to ensure that transactions are processed in a deterministic order. This means that the order of transactions is guaranteed to be the same for every node. This is important because it allows for a deterministic execution of smart contracts.\\n\\nNEAR uses a state machine to store and update the state of the blockchain. This means that the state of the blockchain is always consistent and up-to-date. This is important because it allows for a fast and efficient execution of smart contracts.\\n\\nNEAR uses a permissioning system to control access to the blockchain. This means that only authorized users can interact with the blockchain. This is important because it prevents unauthorized users', \"What are the advantages of NEAR's Proof of Stake consensus mechanism?\\n\\n## What is the difference between the NEAR Proof of Stake consensus mechanism and the NEAR Proof of Work consensus mechanism?\\n\\n## What is the difference between the NEAR Proof of Stake consensus mechanism and the NEAR Proof of Authority consensus mechanism?\\n\\n## What is the difference between the NEAR Proof of Stake consensus mechanism and the NEAR Proof of Authority consensus mechanism?\\n\\n## What is the difference between the NEAR Proof of Stake consensus mechanism and the NEAR Proof of Authority consensus mechanism?\\n\\n## What is the difference between the NEAR Proof of\", \"Describe how NEAR's Rainbow Bridge works for cross-chain communication.\\n\\n## Rainbow Bridge\\n\\nRainbow Bridge is a bridge between Ethereum and NEAR. It allows users to send and receive NEAR tokens on Ethereum and vice versa.\\n\\nRainbow Bridge is a decentralized application (DApp) that allows users to send and receive NEAR tokens on Ethereum and vice versa. It is built on top of the NEAR blockchain and uses the NEAR Protocol.\\n\\nRainbow Bridge is a bridge between Ethereum and NEAR. It allows users to send and receive NEAR tokens on Ethereum and vice versa.\\n\\nRainbow Bridge is a decentralized application (DApp) that\", 'What is the Aurora Engine on NEAR Protocol?\\n\\nThe Aurora Engine is a smart contract that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage their own NEAR accounts. It is a decentralized application that allows users to create and manage']}\n"
     ]
    }
   ],
   "source": [
    "# Initialize dictionaries to store predictions for each model\n",
    "predictions_per_model = {model_name: [] for model_name in model_names}\n",
    "\n",
    "# Generate predictions for each question for each model\n",
    "for model_name, model_data in models.items():\n",
    "    print(f\"Generating predictions for model: {model_name}\")\n",
    "    tokenizer = model_data[\"tokenizer\"]\n",
    "    model = model_data[\"model\"]\n",
    "    \n",
    "    # Initialize text generation pipeline for the current model\n",
    "    text_generation_pipeline = pipeline(\n",
    "        \"text-generation\",\n",
    "        model=model,\n",
    "        tokenizer=tokenizer,\n",
    "        device=device\n",
    "    )\n",
    "    \n",
    "    # Generate predictions for each question\n",
    "    for question in near_protocol_questions:\n",
    "        generated_answer = text_generation_pipeline(\n",
    "            question,\n",
    "            max_length=150,  # Set appropriate max length\n",
    "            num_return_sequences=1,\n",
    "            truncation=True,\n",
    "            pad_token_id=tokenizer.eos_token_id\n",
    "        )[0][\"generated_text\"]\n",
    "        \n",
    "        predictions_per_model[model_name].append(generated_answer.strip())\n",
    "        \n",
    "print(predictions_per_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "0c2ce1d8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating metrics for model: StarCoder2-3B\n",
      "Prediction 1 BLEU Score: 0.51\n",
      "Prediction 1 ROUGE-1 Score: 0.24\n",
      "\n",
      "Prediction 2 BLEU Score: 0.33\n",
      "Prediction 2 ROUGE-1 Score: 0.19\n",
      "\n",
      "Prediction 3 BLEU Score: 0.55\n",
      "Prediction 3 ROUGE-1 Score: 0.34\n",
      "\n",
      "Prediction 4 BLEU Score: 0.55\n",
      "Prediction 4 ROUGE-1 Score: 0.33\n",
      "\n",
      "Prediction 5 BLEU Score: 0.54\n",
      "Prediction 5 ROUGE-1 Score: 0.25\n",
      "\n",
      "Prediction 6 BLEU Score: 0.70\n",
      "Prediction 6 ROUGE-1 Score: 0.30\n",
      "\n",
      "Average BLEU Score for StarCoder2-3B: 0.53\n",
      "Average ROUGE-1 Score for StarCoder2-3B: 0.28\n",
      "Evaluating metrics for model: NEAR-preTrainedStarCoder2\n",
      "Prediction 1 BLEU Score: 0.37\n",
      "Prediction 1 ROUGE-1 Score: 0.22\n",
      "\n",
      "Prediction 2 BLEU Score: 0.54\n",
      "Prediction 2 ROUGE-1 Score: 0.27\n",
      "\n",
      "Prediction 3 BLEU Score: 0.60\n",
      "Prediction 3 ROUGE-1 Score: 0.38\n",
      "\n",
      "Prediction 4 BLEU Score: 0.44\n",
      "Prediction 4 ROUGE-1 Score: 0.29\n",
      "\n",
      "Prediction 5 BLEU Score: 0.59\n",
      "Prediction 5 ROUGE-1 Score: 0.32\n",
      "\n",
      "Prediction 6 BLEU Score: 0.57\n",
      "Prediction 6 ROUGE-1 Score: 0.28\n",
      "\n",
      "Average BLEU Score for NEAR-preTrainedStarCoder2: 0.52\n",
      "Average ROUGE-1 Score for NEAR-preTrainedStarCoder2: 0.29\n",
      "Evaluating metrics for model: NEAR-structTunedStarCoder2\n",
      "Prediction 1 BLEU Score: 0.47\n",
      "Prediction 1 ROUGE-1 Score: 0.26\n",
      "\n",
      "Prediction 2 BLEU Score: 0.48\n",
      "Prediction 2 ROUGE-1 Score: 0.24\n",
      "\n",
      "Prediction 3 BLEU Score: 0.63\n",
      "Prediction 3 ROUGE-1 Score: 0.31\n",
      "\n",
      "Prediction 4 BLEU Score: 0.61\n",
      "Prediction 4 ROUGE-1 Score: 0.29\n",
      "\n",
      "Prediction 5 BLEU Score: 0.52\n",
      "Prediction 5 ROUGE-1 Score: 0.22\n",
      "\n",
      "Prediction 6 BLEU Score: 0.56\n",
      "Prediction 6 ROUGE-1 Score: 0.25\n",
      "\n",
      "Average BLEU Score for NEAR-structTunedStarCoder2: 0.54\n",
      "Average ROUGE-1 Score for NEAR-structTunedStarCoder2: 0.26\n",
      "Evaluating metrics for model: NEARCoder-3b\n",
      "Prediction 1 BLEU Score: 0.32\n",
      "Prediction 1 ROUGE-1 Score: 0.25\n",
      "\n",
      "Prediction 2 BLEU Score: 0.64\n",
      "Prediction 2 ROUGE-1 Score: 0.25\n",
      "\n",
      "Prediction 3 BLEU Score: 0.58\n",
      "Prediction 3 ROUGE-1 Score: 0.33\n",
      "\n",
      "Prediction 4 BLEU Score: 0.41\n",
      "Prediction 4 ROUGE-1 Score: 0.26\n",
      "\n",
      "Prediction 5 BLEU Score: 0.57\n",
      "Prediction 5 ROUGE-1 Score: 0.28\n",
      "\n",
      "Prediction 6 BLEU Score: 0.41\n",
      "Prediction 6 ROUGE-1 Score: 0.27\n",
      "\n",
      "Average BLEU Score for NEARCoder-3b: 0.49\n",
      "Average ROUGE-1 Score for NEARCoder-3b: 0.27\n"
     ]
    }
   ],
   "source": [
    "# Load evaluation metrics\n",
    "bleu = load(\"bleu\", trust_remote_code=True)\n",
    "rouge = load(\"rouge\", trust_remote_code=True)\n",
    "\n",
    "# Iterate over each model to compute BLEU and ROUGE scores\n",
    "for model_name in predictions_per_model:\n",
    "    print(f\"Evaluating metrics for model: {model_name}\")\n",
    "\n",
    "    # Get predictions for the current model\n",
    "    predictions = predictions_per_model[model_name]\n",
    "    \n",
    "    # Ensure each prediction has a corresponding reference(s)\n",
    "    # For this example, we'll pair each prediction with all references for simplicity\n",
    "    # You might need to change this depending on your evaluation strategy\n",
    "    references_for_predictions = [reference_texts] * len(predictions)\n",
    "\n",
    "    # Compute BLEU and ROUGE scores\n",
    "    bleu_scores = []\n",
    "    rouge_scores = []\n",
    "\n",
    "    for i, prediction in enumerate(predictions):\n",
    "        # Prepare single prediction for BLEU and ROUGE\n",
    "        single_prediction = [prediction]\n",
    "\n",
    "        # Prepare references for the single prediction\n",
    "        # Here we assume each prediction is evaluated against all reference texts\n",
    "        # Adjust this according to your specific needs\n",
    "        single_references = [reference_texts]\n",
    "\n",
    "        # Compute BLEU score using the internal tokenizer\n",
    "        bleu_score = bleu.compute(predictions=single_prediction, references=single_references, tokenizer=word_tokenize)\n",
    "        bleu_scores.append(bleu_score['bleu'])\n",
    "\n",
    "        # Compute ROUGE score\n",
    "        rouge_score = rouge.compute(predictions=single_prediction, references=single_references)\n",
    "\n",
    "        # Append ROUGE-1 F1 score\n",
    "        rouge_1_f1 = rouge_score['rouge1'] if 'rouge1' in rouge_score else 0\n",
    "        rouge_scores.append(rouge_1_f1)\n",
    "        \n",
    "        # Print individual prediction scores\n",
    "        print(f\"Prediction {i + 1} BLEU Score: {bleu_score['bleu']:.2f}\")\n",
    "        print(f\"Prediction {i + 1} ROUGE-1 Score: {rouge_1_f1:.2f}\\n\")\n",
    "\n",
    "    # Print average BLEU and ROUGE scores for the current model\n",
    "    avg_bleu_score = sum(bleu_scores) / len(bleu_scores) if bleu_scores else 0\n",
    "    avg_rouge_score = sum(rouge_scores) / len(rouge_scores) if rouge_scores else 0\n",
    "\n",
    "    print(f\"Average BLEU Score for {model_name}: {avg_bleu_score:.2f}\")\n",
    "    print(f\"Average ROUGE-1 Score for {model_name}: {avg_rouge_score:.2f}\")\n",
    "\n",
    "# Optionally, save updated predictions_per_model back to the JSON file\n",
    "with open('predictions_per_model.json', 'w') as f:\n",
    "    json.dump(predictions_per_model, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "945ffaaf",
   "metadata": {},
   "source": [
    "#### NEAR dApp structure generation:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "7eab157e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating model: StarCoder2-3B\n",
      "Structural Similarity: 0.03\n",
      "Comment Quality: 0.00\n",
      "\n",
      "Structural Similarity: 0.02\n",
      "Comment Quality: 0.00\n",
      "\n",
      "Evaluation Metrics for StarCoder2-3B:\n",
      " {'user_prompt': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.'], 'generated_tree': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.\\n\\n## Solution\\n\\nThe following steps are followed to create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. Create a new directory for the project.\\n2. Create a `package.json` file in the project directory.\\n3. Create a `src` directory in the project directory.\\n4. Create a `main.js` file in the `src` directory.\\n5. Create a `index.html` file in the `src` directory.\\n6. Create a `App.vue` file in the `src` directory.\\n7. Create a `App.css` file in the `src` directory.\\n8. Create a `App.js` file in the `src` directory.\\n9. Create a `App.spec.js` file in the `src` directory.\\n10. Create a `App.vue` file in the `src` directory.\\n11. Create a `App.css` file in the `src` directory.\\n12. Create a `App.js` file in the `src` directory.\\n13. Create a `App.spec.js` file in the `src` directory.\\n14. Create a `App.vue` file in the `src` directory.\\n15. Create a `App.css` file in the `src` directory.\\n16. Create a `App.js` file in the `src` directory.\\n17. Create a `App.spec.js` file in the `src` directory.\\n18. Create a `App.vue` file in the `src` directory.\\n19. Create a `App.css` file in the `src` directory.\\n20. Create a `App.js` file in the `src` directory.\\n21. Create a `App.spec.js` file in the `src` directory.\\n22. Create a `App.vue` file in the `src` directory.\\n23. Create a `App.css` file in the `src` directory.\\n24. Create a `App.js` file in the `src` directory.\\n25. Create a `App.spec.js` file in the `src` directory.\\n26. Create a `App.vue` file in the `src` directory.\\n27. Create a `App.css` file in the `src` directory.\\n28. Create a `App.js` file in the `src` directory.\\n29. Create a `App.spec.js` file in the `src` directory.\\n30. Create a `App.vue` file in the `src` directory.\\n31. Create a `App.css` file in the `src` directory.\\n32. Create a `App.js` file in the `src` directory.\\n33. Create a `App.spec.js` file in the `src` directory.\\n34. Create a `App.vue` file in the `src` directory.\\n35. Create a `App.css` file in the `src` directory.\\n36. Create a `App.js` file in the `src` directory.\\n37. Create a `App.spec.js` file in the `src` directory.\\n38. Create a `App.vue` file in the `src` directory.\\n39. Create a `App.css` file in the `src` directory.\\n40. Create a `App.js` file in the `src` directory.\\n41. Create a `App.spec.js` file in the `src` directory.\\n42. Create a `App.vue` file in the `src` directory.\\n43. Create a `App.css` file in the `src` directory.\\n44. Create a `App.js` file in the `src` directory.\\n45. Create a `App.spec.js` file in the `src` directory.\\n46.', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\n## Create a project structure for a NEAR DApp based on the following requirements:\\n\\n### 1. The project should be based on a Rust Smart Contract template.\\n\\nThe project should be based on a Rust Smart Contract template.\\n\\n### 2. Include a `Cargo.toml` file with the project name and author names updated.\\n\\nThe `Cargo.toml` file should include the project name and author names.\\n\\n### 3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n\\nThe `README.md` file should include guidelines similar to starting with a Rust Smart Contract.\\n\\n### 4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n\\nThe `build.sh` script should be used for building the project on Unix systems and the `build.bat` script should be used for building on Windows systems.\\n\\n### 5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n\\nThe `src` directory should contain a `lib.rs` file where the smart contract will be written.\\n\\n### 6. Include a `test.sh` script for testing the contract.\\n\\nThe `test.sh` script should be used for testing the contract.\\n\\n### 7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n\\nAll necessary dependencies for a Rust NEAR DApp should be included and the project configurations should be compatible with NEAR smart contract development.\\n\\n### 8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\nFollow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\n## Create a project structure for a NEAR DApp based on the following requirements:\\n\\n### 1. The project should be based on a Rust Smart Contract template.\\n\\nThe project should be based on a Rust Smart Contract template.\\n\\n### 2. Include a `Cargo.toml` file with the project name and author names updated.\\n\\nThe `Cargo.toml` file should include the project name and author names.\\n\\n### 3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n\\nThe `README.md` file should include guidelines similar to starting with a Rust Smart Contract.\\n\\n### 4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n\\nThe `build.sh` script should be used for building the project on Unix systems and the `build.bat` script should be used for building on Windows systems.\\n\\n### 5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n\\nThe `src` directory should contain a `lib.rs` file where the smart contract will be written.\\n\\n### 6. Include a `test.sh` script for testing the contract.\\n\\nThe `test.sh` script should be used for testing the contract.\\n\\n### 7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n\\nAll necessary dependencies for a Rust NEAR DApp should be included and the project configurations should be compatible with NEAR smart contract development.\\n\\n### 8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\nFollow common conventions and best practices for structuring a NEAR DApp based on the'], 'structural_similarity': [0.02857142857142857, 0.020618556701030927], 'comment_quality': [0.0, 0.0]}\n",
      "Evaluating model: NEAR-preTrainedStarCoder2\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Structural Similarity: 0.03\n",
      "Comment Quality: 0.00\n",
      "\n",
      "Structural Similarity: 0.03\n",
      "Comment Quality: 0.00\n",
      "\n",
      "Evaluation Metrics for NEAR-preTrainedStarCoder2:\n",
      " {'user_prompt': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.'], 'generated_tree': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements. This will help developers understand the structure of the project and easily navigate through it.\\n\\n## Example\\n\\n```\\n├── package.json\\n├── package-lock.json\\n├── README.md\\n├── src\\n│   ├── App.vue\\n│   ├── assets\\n│   │   └── logo.png\\n│   ├── components\\n│   │   └── HelloWorld.vue\\n│   ├── main.js\\n│   └── styles\\n│       └── style.css\\n└── vue.config.js\\n```\\n\\n## Key Takeaways\\n\\n- Create a new Vue.js project using the `vue create` command.\\n- Configure the project to use the necessary dependencies and packages.\\n- Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n- Adhere to common Vue.js project structure conventions while setting up the project.\\n\\n## Further Reading\\n\\n- [Vue.js Documentation](https://vuejs.org/): Learn about Vue.js and its features.\\n- [Vue CLI Documentation](https://cli.vuejs.org/): Learn about Vue CLI and its features.\\n- [Vue CLI Project Structure](https://cli.vuejs.org/guide/project-structure.html): Learn about the structure of a Vue CLI project.\\n- [Vue CLI Setup](https://cli.vuejs.org/guide/installation.html): Learn how to set up Vue CLI.\\n- [Vue CLI Configuration](https://cli.vuejs.org/config/): Learn about the configuration options for Vue CLI.\\n- [Vue CLI Plugins](https://cli.vuejs.org/guide/plugins.html): Learn about the plugins available for Vue CLI.\\n- [Vue CLI Commands](https://cli.vuejs.org/guide/commands.html): Learn about the commands available for Vue CLI.\\n- [Vue CLI Project Structure](https://cli.vuejs.org/guide/project-structure.html): Learn about the structure of a Vue CLI project.\\n- [Vue CLI Setup](https://cli.vuejs.org/guide/installation.html): Learn how to set up Vue CLI.\\n- [Vue CLI Configuration](https://cli.vuejs.org/config/): Learn about the configuration options for Vue CLI.\\n- [Vue CLI Plugins](https://cli.vuejs.org/guide/plugins.html): Learn about the plugins available for Vue CLI.\\n- [Vue CLI Commands](https://cli.vuejs.org/guide/commands.html): Learn about the commands available for Vue CLI.\\n- [Vue CLI Project Structure](https://cli.vuejs.org/guide/project-structure.html): Learn about the structure of a Vue CLI project.\\n- [Vue CLI Setup](https://cli.vuejs.org/guide/installation.html): Learn how to set up Vue CLI.\\n- [Vue CLI Configuration](https://cli.vuejs.org/config/): Learn about the configuration options for Vue CLI.\\n- [Vue CLI Plugins](https://cli.vuejs.org/guide/plugins.html): Learn about the plugins available for Vue CLI.\\n- [Vue CLI Commands](https://cli.vuejs.org/guide/commands.html): Learn about the commands available for Vue CLI.\\n- [Vue CLI Project Structure](https://cli.vuejs.org/guide/project-structure.html): Learn about the structure of a Vue CLI project.\\n- [Vue CLI Setup](https://cli.vuejs.org/guide/installation.html): Learn how to set up Vue CLI.\\n- [Vue CLI Configuration](https://cli.vuejs.org/config/): Learn about the configuration options for Vue CLI.\\n- [Vue CLI Plugins](https://cli.vuejs.org/guide/plugins.html): Learn about the plugins available for Vue CLI.\\n- [Vue CLI Commands](https://cli.vuejs.org/guide/commands.html): Learn about the commands available for Vue CLI.\\n- [Vue CLI Project Structure](https://cli.vuejs.org/guide/project', \"Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\n## Example\\n\\nHere's an example of a project structure for a NEAR DApp based on the requirements listed above:\\n\\n```bash\\n.\\n├── Cargo.toml\\n├── README.md\\n├── build.bat\\n├── build.sh\\n├── src\\n│   └── lib.rs\\n└── test.sh\\n```\\n\\n## Conclusion\\n\\nBy following the provided guidelines, you can create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n\\n## Further Reading\\n\\n- [Rust Smart Contract Template](https://github.com/near-examples/rust-smart-contract-template)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Development](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-development.html)\\n- [Rust Smart Contract Testing](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-testing.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/concepts/rust-smart-contract-best-practices.html)\\n- [Rust Smart Contract Best Practices](https://docs.rs/near-sdk/latest/near_sdk/\"], 'structural_similarity': [0.02857142857142857, 0.02702702702702703], 'comment_quality': [0.0, 0.0]}\n",
      "Evaluating model: NEAR-structTunedStarCoder2\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Structural Similarity: 0.03\n",
      "Comment Quality: 1.00\n",
      "\n",
      "Structural Similarity: 0.03\n",
      "Comment Quality: 1.00\n",
      "\n",
      "Evaluation Metrics for NEAR-structTunedStarCoder2:\n",
      " {'user_prompt': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.'], 'generated_tree': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.\\n\\n```\\n.\\n├── README.md\\n├── build\\n│\\xa0\\xa0 ├── build.js\\n│\\xa0\\xa0 ├── check-version.js\\n│\\xa0\\xa0 ├── dev-server.js\\n│\\xa0\\xa0 ├── dev-server.conf.js\\n│\\xa0\\xa0 ├── dev-server.js\\n│\\xa0\\xa0 ├── dev-server.js.map\\n│\\xa0\\xa0 ├── dev-server.js.orig\\n│\\xa0\\xa0 ├── dev-server.js.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp\\n│\\xa0\\xa0 ├── dev-server.js.swp.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.orig.map\\n│\\xa0\\xa0 ├── dev-server.js.swp.orig.orig.orig.orig.orig.orig.orig.orig.orig', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n9. `Cargo.toml` should have the following sections:\\n    - `[package]`\\n        - `name`\\n        - `version`\\n        - `description`\\n        - `license`\\n        - `repository`\\n        - `homepage`\\n        - `documentation`\\n        - `readme`\\n        - `keywords`\\n        - `categories`\\n        - `authors`\\n        - `edition`\\n    - `[dependencies]`\\n        - `near-sdk`\\n        - `near-sdk-derive`\\n        - `near-sdk-test-env`\\n        - `near-sdk-test-utils`\\n        - `serde`\\n        - `serde_json`\\n        - `serde_derive`\\n        - `serde_with`\\n        - `serde_with_macros`\\n        - `near-bindgen`\\n        - `near-bindgen-derive`\\n        - `near-bindgen-derive-macro`\\n        - `near-bindgen-macro-support`\\n        - `near-bindgen-macro-support-derive`\\n        - `near-bindgen-macro-support-derive-macro`\\n        - `near-bindgen-macro-support-proc-macro`\\n        - `near-bindgen-macro-support-proc-macro-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-macro-support-proc-macro-derive-macro-support-derive-macro-support-derive-macro-support-derive`\\n        - `near-bindgen-'], 'structural_similarity': [0.03389830508474576, 0.025], 'comment_quality': [1.0, 1.0]}\n",
      "Evaluating model: NEARCoder-3b\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Structural Similarity: 0.02\n",
      "Comment Quality: 0.00\n",
      "\n",
      "Structural Similarity: 0.03\n",
      "Comment Quality: 1.00\n",
      "\n",
      "Evaluation Metrics for NEARCoder-3b:\n",
      " {'user_prompt': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.'], 'generated_tree': ['Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be built using Vue.js.\\n2. Include necessary files typically required for a Vue.js project such as configuration files, build scripts, and main application files.\\n3. Incorporate all relevant dependencies and packages needed for a Vue.js project.\\n4. Ensure the project includes npm setup for installing dependencies and scripts for serving, building, and analyzing the production bundle.\\n5. Adhere to common Vue.js project structure conventions while setting up the project.\\n\\nProvide a well-organized directory structure and file list based on these requirements.\\n\\n## Project Structure\\n\\n```\\n├── README.md\\n├── package.json\\n├── public\\n│   ├── favicon.ico\\n│   └── index.html\\n├── src\\n│   ├── App.vue\\n│   ├── assets\\n│   │   └── logo.png\\n│   ├── components\\n│   │   ├── AppHeader.vue\\n│   │   ├── AppFooter.vue\\n│   │   ├── AppMain.vue\\n│   │   ├── AppSidebar.vue\\n│   │   ├── AppSidebarItem.vue\\n│   │   ├── AppSidebarDivider.vue\\n│   │   ├── AppSidebarMinimizer.vue\\n│   │   ├── AppSidebarNav.vue\\n│   │   ├── AppSidebarNavItem.vue\\n│   │   ├── AppSidebarNavLink.vue\\n│   │   ├── AppSidebarSearch.vue\\n│   │   ├── AppSidebarSubmenu.vue\\n│   │   ├── AppSidebarToggler.vue\\n│   │   ├── AppSidebarUser.vue\\n│   │   ├── AppSidebarUserDropdown.vue\\n│   │   ├── AppSidebarUserItem.vue\\n│   │   ├── AppSidebarUserLink.vue\\n│   │   ├── AppSidebarUserMenu.vue\\n│   │   ├── AppSidebarUserSubmenu.vue\\n│   │   ├── AppSidebarUserToggle.vue\\n│   │   ├── AppSidebarUserTooltip.vue\\n│   │   ├── AppSidebarWrapper.vue\\n│   │   ├── AppSidebarWrapperItem.vue\\n│   │   ├── AppSidebarWrapperLink.vue\\n│   │   ├── AppSidebarWrapperSubmenu.vue\\n│   │   ├── AppSidebarWrapperToggle.vue\\n│   │   ├── AppSidebarWrapperTooltip.vue\\n│   │   ├── AppSidebarWrapperUser.vue\\n│   │   ├── AppSidebarWrapperUserDropdown.vue\\n│   │   ├── AppSidebarWrapperUserItem.vue\\n│   │   ├── AppSidebarWrapperUserLink.vue\\n│   │   ├── AppSidebarWrapperUserMenu.vue\\n│   │   ├── AppSidebarWrapperUserSubmenu.vue\\n│   │   ├── AppSidebarWrapperUserToggle.vue\\n│   │   ├── AppSidebarWrapperUserTooltip.vue\\n│   │   ├── AppSidebarWrapperWrapper.vue\\n│   │   ├── AppSidebarWrapperWrapperItem.vue\\n│   │   ├── AppSidebarWrapperWrapperLink.vue\\n│   │   ├── AppSidebarWrapperWrapperSubmenu.vue\\n│   │   ├── AppSidebarWrapperWrapperToggle.vue\\n│   │   ├── AppSidebarWrapperWrapperTooltip.vue\\n│   │   ├── AppSidebarWrapperWrapperUser.vue\\n│   │   ├── AppSidebarWrapperWrapperUserDropdown.vue\\n│   │   ├── AppSidebarWrapperWrapperUserItem.vue\\n│   │   ├── AppSidebarWrapperWrapperUserLink.vue\\n│   │   ├── AppSidebarWrapperWrapperUserMenu.vue\\n│   │   ├── AppSidebarWrapperWrapperUserSubmenu.vue\\n│   │   ├── AppSidebarWrapperWrapperUserToggle.vue\\n│   │   ├── AppSidebarWrapperWrapperUserTooltip.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapper.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperItem.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperLink.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperSubmenu.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperToggle.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperTooltip.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUser.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserDropdown.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserItem.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserLink.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserMenu.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserSubmenu.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserToggle.vue\\n│   │   ├── AppSidebarWrapperWrapperWrapperUserTooltip.vue\\n│   │   ├── AppSidebarWrapperWrapper', 'Create a project structure for a NEAR DApp based on the following requirements:\\n\\n1. The project should be based on a Rust Smart Contract template.\\n2. Include a `Cargo.toml` file with the project name and author names updated.\\n3. The project must have a `README.md` file with guidelines similar to starting with a Rust Smart Contract.\\n4. Utilize a `build.sh` script for building the project on Unix systems and a `build.bat` script for building on Windows systems.\\n5. Use a `src` directory containing a `lib.rs` file where the smart contract will be written.\\n6. Include a `test.sh` script for testing the contract.\\n7. Ensure all necessary dependencies for a Rust NEAR DApp are included and the project configurations are compatible with NEAR smart contract development.\\n8. Follow common conventions and best practices for structuring a NEAR DApp based on the provided README instructions.\\n9. Ensure the project is properly formatted and linted using `cargo fmt` and `cargo clippy`.\\n10. Ensure the project is tested using `cargo test`.\\n11. Ensure the project is compiled using `cargo build --release`.\\n12. Ensure the project is packaged using `cargo package`.\\n13. Ensure the project is published to the NEAR Registry using `cargo publish`.\\n14. Ensure the project is documented using `cargo doc`.\\n15. Ensure the project is formatted and linted using `rustfmt` and `clippy` before committing.\\n16. Ensure the project is tested using `cargo test` before committing.\\n17. Ensure the project is compiled using `cargo build --release` before committing.\\n18. Ensure the project is packaged using `cargo package` before committing.\\n19. Ensure the project is published to the NEAR Registry using `cargo publish` before committing.\\n20. Ensure the project is documented using `cargo doc` before committing.\\n21. Ensure the project is formatted and linted using `rustfmt` and `clippy` before pushing to the remote repository.\\n22. Ensure the project is tested using `cargo test` before pushing to the remote repository.\\n23. Ensure the project is compiled using `cargo build --release` before pushing to the remote repository.\\n24. Ensure the project is packaged using `cargo package` before pushing to the remote repository.\\n25. Ensure the project is published to the NEAR Registry using `cargo publish` before pushing to the remote repository.\\n26. Ensure the project is documented using `cargo doc` before pushing to the remote repository.\\n27. Ensure the project is formatted and linted using `rustfmt` and `clippy` before merging to the main branch.\\n28. Ensure the project is tested using `cargo test` before merging to the main branch.\\n29. Ensure the project is compiled using `cargo build --release` before merging to the main branch.\\n30. Ensure the project is packaged using `cargo package` before merging to the main branch.\\n31. Ensure the project is published to the NEAR Registry using `cargo publish` before merging to the main branch.\\n32. Ensure the project is documented using `cargo doc` before merging to the main branch.\\n33. Ensure the project is formatted and linted using `rustfmt` and `clippy` before creating a pull request.\\n34. Ensure the project is tested using `cargo test` before creating a pull request.\\n35. Ensure the project is compiled using `cargo build --release` before creating a pull request.\\n36. Ensure the project is packaged using `cargo package` before creating a pull request.\\n37. Ensure the project is published to the NEAR Registry using `cargo publish` before creating a pull request.\\n38. Ensure the project is documented using `cargo doc` before creating a pull request.\\n39. Ensure the project is formatted and linted using `rustfmt` and `clippy` before merging the pull request.\\n40. Ensure the project is tested using `cargo test` before merging the pull request.\\n41. Ensure the project is compiled using `cargo build --release` before merging the pull request.\\n42. Ensure the project is packaged using `cargo package` before merging the pull request.\\n43. Ensure the project is published to the NEAR Registry using `cargo publish` before merging the pull request.\\n44. Ensure the project is documented using `cargo doc` before merging the pull request.\\n45. Ensure the project is formatted and linted using `rustfmt` and `clippy` before closing the pull request.\\n46. Ensure the project is tested using `cargo test` before closing the'], 'structural_similarity': [0.020618556701030927, 0.029850746268656716], 'comment_quality': [0.0, 1.0]}\n"
     ]
    }
   ],
   "source": [
    "# Define evaluation functions for dApp tree generation\n",
    "def calculate_structural_similarity(generated_tree, reference_tree):\n",
    "    generated_lines = generated_tree.split('\\n')\n",
    "    reference_lines = reference_tree.split('\\n')\n",
    "    similarity = difflib.SequenceMatcher(None, generated_lines, reference_lines).ratio()\n",
    "    return similarity\n",
    "\n",
    "def evaluate_comment_quality(generated_tree, reference_tree):\n",
    "    generated_comments = [line for line in generated_tree.split('\\n') if line.strip().startswith('#')]\n",
    "    reference_comments = [line for line in reference_tree.split('\\n') if line.strip().startswith('#')]\n",
    "    similarity = difflib.SequenceMatcher(None, generated_comments, reference_comments).ratio()\n",
    "    return similarity\n",
    "\n",
    "def evaluate_dapp_trees_generation(model, tokenizer, dataset):\n",
    "    text_generation_pipeline = pipeline(\"text-generation\", model=model, tokenizer=tokenizer)\n",
    "    evaluation_metrics = {\n",
    "        \"user_prompt\": [],\n",
    "        \"generated_tree\": [],\n",
    "        \"structural_similarity\": [],\n",
    "        \"comment_quality\": []\n",
    "    }\n",
    "\n",
    "    for tree_example in dataset:\n",
    "        generated_tree = text_generation_pipeline(\n",
    "            tree_example[\"user_prompt\"], \n",
    "            max_length=1024, \n",
    "            num_return_sequences=1,\n",
    "            truncation=True,\n",
    "            pad_token_id=tokenizer.eos_token_id\n",
    "        )[0][\"generated_text\"]\n",
    "\n",
    "        reference_tree = tree_example[\"tree\"]\n",
    "        structural_similarity = calculate_structural_similarity(generated_tree, reference_tree)\n",
    "        comment_quality = evaluate_comment_quality(generated_tree, reference_tree)\n",
    "\n",
    "        # Collect evaluation metrics\n",
    "        evaluation_metrics[\"user_prompt\"].append(tree_example[\"user_prompt\"])\n",
    "        evaluation_metrics[\"generated_tree\"].append(generated_tree)\n",
    "        evaluation_metrics[\"structural_similarity\"].append(structural_similarity)\n",
    "        evaluation_metrics[\"comment_quality\"].append(comment_quality)\n",
    "\n",
    "        # Print evaluation metrics for the current example\n",
    "        #print(f\"User Prompt: {tree_example['user_prompt']}\")\n",
    "        #print(f\"Generated Tree:\\n{generated_tree}\")\n",
    "        #print(f\"Reference Tree:\\n{reference_tree}\")\n",
    "        print(f\"Structural Similarity: {structural_similarity:.2f}\")\n",
    "        print(f\"Comment Quality: {comment_quality:.2f}\\n\")\n",
    "\n",
    "    return evaluation_metrics\n",
    "\n",
    "# Evaluate each preloaded model on dApp trees generation\n",
    "for model_name in model_names.keys():\n",
    "    print(f\"Evaluating model: {model_name}\")\n",
    "    model = models[model_name][\"model\"]\n",
    "    tokenizer = models[model_name][\"tokenizer\"]\n",
    "    evaluation_metrics = evaluate_dapp_trees_generation(model, tokenizer, dapp_trees)\n",
    "    print(f\"Evaluation Metrics for {model_name}:\\n\", evaluation_metrics)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "74baa0bf",
   "metadata": {},
   "source": [
    "#### Basic NEAR dApp development:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "07e499aa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluating model: StarCoder2-3B\n",
      "Project Name: NEAR Calendar App\n",
      "Project Description: A simple calendar dApp on NEAR for scheduling events.\n",
      "Generated Code:\n",
      "A simple calendar dApp on NEAR for scheduling events.\n",
      "\n",
      "## How to run\n",
      "\n",
      "### Install NEAR CLI\n",
      "\n",
      "```\n",
      "npm install -g near-cli\n",
      "```\n",
      "\n",
      "### Create a new account\n",
      "\n",
      "```\n",
      "near create-account <account-id>\n",
      "```\n",
      "\n",
      "### Deploy the contract\n",
      "\n",
      "```\n",
      "near deploy --wasm <path-to-wasm-file> --accountId <account-id>\n",
      "```\n",
      "\n",
      "### Run the contract\n",
      "\n",
      "```\n",
      "near call <account-id> <contract-id> <method-name> --accountId <account-id> --amount 0.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Project Name: NEAR Snake Game\n",
      "Project Description: A basic snake game dApp on NEAR.\n",
      "Generated Code:\n",
      "A basic snake game dApp on NEAR.\n",
      "\n",
      "## How to run\n",
      "\n",
      "```\n",
      "npm install\n",
      "npm run dev\n",
      "```\n",
      "\n",
      "## How to deploy\n",
      "\n",
      "```\n",
      "npm run deploy\n",
      "```\n",
      "\n",
      "## How to test\n",
      "\n",
      "```\n",
      "npm run test\n",
      "```\n",
      "\n",
      "## How to play\n",
      "\n",
      "- Open [https://near-snake.vercel.app/](https://near-snake.vercel.app/) in your browser.\n",
      "- Click on the \"Play\" button.\n",
      "- Click on the \"Start\" button.\n",
      "- Use the arrow keys to move the snake.\n",
      "- Click on the \"Pause\" button to pause the game.\n",
      "- Click on the \"Restart\" button to restart the game.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"Leaderboard\" button to see the top 10 scores.\n",
      "- Click on the \"\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.17\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Evaluation Metrics for StarCoder2-3B:\n",
      " {'name': ['NEAR Calendar App', 'NEAR Snake Game'], 'description': ['A simple calendar dApp on NEAR for scheduling events.', 'A basic snake game dApp on NEAR.'], 'generated_code': ['A simple calendar dApp on NEAR for scheduling events.\\n\\n## How to run\\n\\n### Install NEAR CLI\\n\\n```\\nnpm install -g near-cli\\n```\\n\\n### Create a new account\\n\\n```\\nnear create-account <account-id>\\n```\\n\\n### Deploy the contract\\n\\n```\\nnear deploy --wasm <path-to-wasm-file> --accountId <account-id>\\n```\\n\\n### Run the contract\\n\\n```\\nnear call <account-id> <contract-id> <method-name> --accountId <account-id> --amount 0.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000', 'A basic snake game dApp on NEAR.\\n\\n## How to run\\n\\n```\\nnpm install\\nnpm run dev\\n```\\n\\n## How to deploy\\n\\n```\\nnpm run deploy\\n```\\n\\n## How to test\\n\\n```\\nnpm run test\\n```\\n\\n## How to play\\n\\n- Open [https://near-snake.vercel.app/](https://near-snake.vercel.app/) in your browser.\\n- Click on the \"Play\" button.\\n- Click on the \"Start\" button.\\n- Use the arrow keys to move the snake.\\n- Click on the \"Pause\" button to pause the game.\\n- Click on the \"Restart\" button to restart the game.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"Leaderboard\" button to see the top 10 scores.\\n- Click on the \"'], 'functionality': [0.0, 0.0], 'usability': [0.0, 0.16666666666666666], 'code_quality': [0.08333333333333333, 0.08333333333333333]}\n",
      "Evaluating model: NEAR-preTrainedStarCoder2\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Project Name: NEAR Calendar App\n",
      "Project Description: A simple calendar dApp on NEAR for scheduling events.\n",
      "Generated Code:\n",
      "A simple calendar dApp on NEAR for scheduling events.\n",
      "\n",
      "## Getting started\n",
      "\n",
      "### Prerequisites\n",
      "\n",
      "- [Node.js](https://nodejs.org/en/)\n",
      "- [Yarn](https://yarnpkg.com/)\n",
      "- [Git](https://git-scm.com/)\n",
      "- [NEAR CLI](https://docs.near.org/tools/near-cli)\n",
      "\n",
      "### Installation\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/calendar-dapp.git\n",
      "cd calendar-dapp\n",
      "yarn install\n",
      "```\n",
      "\n",
      "### Configuration\n",
      "\n",
      "Create a `.env` file in the root directory of the project.\n",
      "\n",
      "```bash\n",
      "touch.env\n",
      "```\n",
      "\n",
      "Add the following environment variables to the `.env` file.\n",
      "\n",
      "```bash\n",
      "# NEAR account ID\n",
      "ACCOUNT_ID=your-account-id.testnet\n",
      "\n",
      "# NEAR account private key\n",
      "PRIVATE_KEY=your-account-private-key\n",
      "\n",
      "# NEAR network ID\n",
      "NETWORK_ID=testnet\n",
      "```\n",
      "\n",
      "### Running the app\n",
      "\n",
      "```bash\n",
      "yarn start\n",
      "```\n",
      "\n",
      "## Contributing\n",
      "\n",
      "We welcome community contributions to this project. Please see our [Contributing Guidelines](https://github.com/near-examples/calendar-dapp/blob/master/CONTRIBUTING.md) for more information.\n",
      "\n",
      "## License\n",
      "\n",
      "[MIT](https://github.com/near-examples/calendar-dapp/blob/master/LICENSE)\n",
      "\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.17\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Project Name: NEAR Snake Game\n",
      "Project Description: A basic snake game dApp on NEAR.\n",
      "Generated Code:\n",
      "A basic snake game dApp on NEAR.\n",
      "\n",
      "## Getting Started\n",
      "\n",
      "### Prerequisites\n",
      "\n",
      "- [NEAR CLI](https://docs.near.org/tools/near-cli)\n",
      "- [Node.js](https://nodejs.org/en/)\n",
      "\n",
      "### Installation\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/snake-game-dapp.git\n",
      "cd snake-game-dapp\n",
      "npm install\n",
      "```\n",
      "\n",
      "### Running the App\n",
      "\n",
      "```bash\n",
      "npm run dev\n",
      "```\n",
      "\n",
      "### Deploying the App\n",
      "\n",
      "```bash\n",
      "npm run deploy\n",
      "```\n",
      "\n",
      "## How It Works\n",
      "\n",
      "The game is played by a snake that moves around the board. The snake can move in four directions: up, down, left, and right. The snake can also eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\n",
      "\n",
      "The snake game is played on\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Evaluation Metrics for NEAR-preTrainedStarCoder2:\n",
      " {'name': ['NEAR Calendar App', 'NEAR Snake Game'], 'description': ['A simple calendar dApp on NEAR for scheduling events.', 'A basic snake game dApp on NEAR.'], 'generated_code': ['A simple calendar dApp on NEAR for scheduling events.\\n\\n## Getting started\\n\\n### Prerequisites\\n\\n- [Node.js](https://nodejs.org/en/)\\n- [Yarn](https://yarnpkg.com/)\\n- [Git](https://git-scm.com/)\\n- [NEAR CLI](https://docs.near.org/tools/near-cli)\\n\\n### Installation\\n\\n```bash\\ngit clone https://github.com/near-examples/calendar-dapp.git\\ncd calendar-dapp\\nyarn install\\n```\\n\\n### Configuration\\n\\nCreate a `.env` file in the root directory of the project.\\n\\n```bash\\ntouch.env\\n```\\n\\nAdd the following environment variables to the `.env` file.\\n\\n```bash\\n# NEAR account ID\\nACCOUNT_ID=your-account-id.testnet\\n\\n# NEAR account private key\\nPRIVATE_KEY=your-account-private-key\\n\\n# NEAR network ID\\nNETWORK_ID=testnet\\n```\\n\\n### Running the app\\n\\n```bash\\nyarn start\\n```\\n\\n## Contributing\\n\\nWe welcome community contributions to this project. Please see our [Contributing Guidelines](https://github.com/near-examples/calendar-dapp/blob/master/CONTRIBUTING.md) for more information.\\n\\n## License\\n\\n[MIT](https://github.com/near-examples/calendar-dapp/blob/master/LICENSE)\\n', 'A basic snake game dApp on NEAR.\\n\\n## Getting Started\\n\\n### Prerequisites\\n\\n- [NEAR CLI](https://docs.near.org/tools/near-cli)\\n- [Node.js](https://nodejs.org/en/)\\n\\n### Installation\\n\\n```bash\\ngit clone https://github.com/near-examples/snake-game-dapp.git\\ncd snake-game-dapp\\nnpm install\\n```\\n\\n### Running the App\\n\\n```bash\\nnpm run dev\\n```\\n\\n### Deploying the App\\n\\n```bash\\nnpm run deploy\\n```\\n\\n## How It Works\\n\\nThe game is played by a snake that moves around the board. The snake can move in four directions: up, down, left, and right. The snake can also eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on a 20x20 board. The snake starts at the center of the board and can move in any direction. The snake can eat food that appears on the board. The snake dies when it hits a wall or itself.\\n\\nThe snake game is played on'], 'functionality': [0.0, 0.0], 'usability': [0.16666666666666666, 0.0], 'code_quality': [0.08333333333333333, 0.08333333333333333]}\n",
      "Evaluating model: NEAR-structTunedStarCoder2\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Project Name: NEAR Calendar App\n",
      "Project Description: A simple calendar dApp on NEAR for scheduling events.\n",
      "Generated Code:\n",
      "A simple calendar dApp on NEAR for scheduling events.\n",
      "\n",
      "## Getting started\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/calendar-frontend.git\n",
      "cd calendar-frontend\n",
      "yarn install\n",
      "yarn start\n",
      "```\n",
      "\n",
      "## Deploying\n",
      "\n",
      "```bash\n",
      "yarn deploy\n",
      "```\n",
      "\n",
      "## License\n",
      "\n",
      "[MIT](LICENSE)\n",
      "\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Project Name: NEAR Snake Game\n",
      "Project Description: A basic snake game dApp on NEAR.\n",
      "Generated Code:\n",
      "A basic snake game dApp on NEAR.\n",
      "\n",
      "## Getting Started\n",
      "\n",
      "### Prerequisites\n",
      "\n",
      "- [Node.js](https://nodejs.org/en/)\n",
      "- [yarn](https://yarnpkg.com/)\n",
      "- [near-cli](https://github.com/near/near-cli)\n",
      "\n",
      "### Installation\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/snake-game-js\n",
      "cd snake-game-js\n",
      "yarn\n",
      "```\n",
      "\n",
      "### Compile\n",
      "\n",
      "```bash\n",
      "yarn build\n",
      "```\n",
      "\n",
      "### Deploy\n",
      "\n",
      "```bash\n",
      "near deploy\n",
      "```\n",
      "\n",
      "### Run\n",
      "\n",
      "```bash\n",
      "yarn start\n",
      "```\n",
      "\n",
      "### Test\n",
      "\n",
      "```bash\n",
      "yarn test\n",
      "```\n",
      "\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Evaluation Metrics for NEAR-structTunedStarCoder2:\n",
      " {'name': ['NEAR Calendar App', 'NEAR Snake Game'], 'description': ['A simple calendar dApp on NEAR for scheduling events.', 'A basic snake game dApp on NEAR.'], 'generated_code': ['A simple calendar dApp on NEAR for scheduling events.\\n\\n## Getting started\\n\\n```bash\\ngit clone https://github.com/near-examples/calendar-frontend.git\\ncd calendar-frontend\\nyarn install\\nyarn start\\n```\\n\\n## Deploying\\n\\n```bash\\nyarn deploy\\n```\\n\\n## License\\n\\n[MIT](LICENSE)\\n', 'A basic snake game dApp on NEAR.\\n\\n## Getting Started\\n\\n### Prerequisites\\n\\n- [Node.js](https://nodejs.org/en/)\\n- [yarn](https://yarnpkg.com/)\\n- [near-cli](https://github.com/near/near-cli)\\n\\n### Installation\\n\\n```bash\\ngit clone https://github.com/near-examples/snake-game-js\\ncd snake-game-js\\nyarn\\n```\\n\\n### Compile\\n\\n```bash\\nyarn build\\n```\\n\\n### Deploy\\n\\n```bash\\nnear deploy\\n```\\n\\n### Run\\n\\n```bash\\nyarn start\\n```\\n\\n### Test\\n\\n```bash\\nyarn test\\n```\\n'], 'functionality': [0.0, 0.0], 'usability': [0.0, 0.0], 'code_quality': [0.08333333333333333, 0.08333333333333333]}\n",
      "Evaluating model: NEARCoder-3b\n",
      "Project Name: NEAR Calendar App\n",
      "Project Description: A simple calendar dApp on NEAR for scheduling events.\n",
      "Generated Code:\n",
      "A simple calendar dApp on NEAR for scheduling events.\n",
      "\n",
      "## Getting started\n",
      "\n",
      "### Prerequisites\n",
      "\n",
      "- [Node.js](https://nodejs.org/)\n",
      "- [Yarn](https://yarnpkg.com/)\n",
      "- [NEAR CLI](https://github.com/near/near-cli)\n",
      "- [NEAR testnet](https://testnet.near.org/)\n",
      "\n",
      "### Installation\n",
      "\n",
      "Clone the repository:\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/calendar-contract.git\n",
      "cd calendar-contract\n",
      "```\n",
      "\n",
      "Install dependencies:\n",
      "\n",
      "```bash\n",
      "yarn\n",
      "```\n",
      "\n",
      "### Compile and deploy\n",
      "\n",
      "Compile the contract:\n",
      "\n",
      "```bash\n",
      "yarn build\n",
      "```\n",
      "\n",
      "Deploy the contract to the testnet:\n",
      "\n",
      "```bash\n",
      "yarn deploy\n",
      "```\n",
      "\n",
      "### Run tests\n",
      "\n",
      "Run the tests:\n",
      "\n",
      "```bash\n",
      "yarn test\n",
      "```\n",
      "\n",
      "## License\n",
      "\n",
      "[MIT](LICENSE)\n",
      "\n",
      "## Contributing\n",
      "\n",
      "See [CONTRIBUTING.md](CONTRIBUTING.md)\n",
      "\n",
      "## Code of Conduct\n",
      "\n",
      "See [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md)\n",
      "\n",
      "## Acknowledgements\n",
      "\n",
      "- [NEAR](https://near.org/)\n",
      "- [NEAR Protocol](https://protocol.near.org/)\n",
      "- [NEAR Developer Resources](https://docs.near.org/)\n",
      "- [NEAR Developer Guides](https://docs.near.org/docs/develop)\n",
      "- [NEAR Community](https://community.near.org/)\n",
      "- [NEAR Governance](https://near.org/governance)\n",
      "- [NEAR Wallet](https://wallet.near.org/)\n",
      "- [NEAR Explorer](https://explorer.near.org/)\n",
      "- [NEAR Wallet Explorer](https://wallet.explorer.near.org/)\n",
      "- [NEAR API Reference](https://docs.near.org/docs/develop/front-end/js-api-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Developer Tools](https://docs.near.org/docs/develop/tools)\n",
      "- [NEAR Developer Tools Reference](https://docs.near.org/docs/develop/tools/tools-reference)\n",
      "- [NEAR Developer Guides](https://docs.near.org/docs/develop)\n",
      "- [NEAR Developer Guides Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts](https://github.com/near/near-contract-standards)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\n",
      "- [\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.17\n",
      "\n",
      "Project Name: NEAR Snake Game\n",
      "Project Description: A basic snake game dApp on NEAR.\n",
      "Generated Code:\n",
      "A basic snake game dApp on NEAR.\n",
      "\n",
      "## Getting started\n",
      "\n",
      "### Prerequisites\n",
      "\n",
      "- [Node.js](https://nodejs.org/)\n",
      "- [Yarn](https://yarnpkg.com/)\n",
      "- [NEAR CLI](https://github.com/near/near-cli)\n",
      "\n",
      "### Installation\n",
      "\n",
      "Clone the repository:\n",
      "\n",
      "```bash\n",
      "git clone https://github.com/near-examples/snake-game.git\n",
      "cd snake-game\n",
      "```\n",
      "\n",
      "Install dependencies:\n",
      "\n",
      "```bash\n",
      "yarn install\n",
      "```\n",
      "\n",
      "### Compile and deploy\n",
      "\n",
      "Compile and deploy the contract:\n",
      "\n",
      "```bash\n",
      "yarn deploy\n",
      "```\n",
      "\n",
      "### Run tests\n",
      "\n",
      "Run unit tests:\n",
      "\n",
      "```bash\n",
      "yarn test\n",
      "```\n",
      "\n",
      "Run integration tests:\n",
      "\n",
      "```bash\n",
      "yarn test:integration\n",
      "```\n",
      "\n",
      "### Run the app\n",
      "\n",
      "Start the app:\n",
      "\n",
      "```bash\n",
      "yarn start\n",
      "```\n",
      "\n",
      "Open [http://localhost:3000](http://localhost:3000) to view it in the browser.\n",
      "\n",
      "## License\n",
      "\n",
      "[MIT](LICENSE)\n",
      "\n",
      "## Acknowledgements\n",
      "\n",
      "- [NEAR](https://near.org/)\n",
      "- [NEAR Developer Hub](https://docs.near.org/docs/develop/front-end/react)\n",
      "- [NEAR React Starter Kit](https://github.com/near-examples/react-starter-kit)\n",
      "- [NEAR React Snake Game](https://github.com/near-examples/react-snake-game)\n",
      "- [NEAR React Snake Game Tutorial](https://docs.near.org/docs/tutorials/react-snake-game)\n",
      "- [NEAR React Snake Game Tutorial Part 2](https://docs.near.org/docs/tutorials/react-snake-game-part-2)\n",
      "- [NEAR React Snake Game Tutorial Part 3](https://docs.near.org/docs/tutorials/react-snake-game-part-3)\n",
      "- [NEAR React Snake Game Tutorial Part 4](https://docs.near.org/docs/tutorials/react-snake-game-part-4)\n",
      "- [NEAR React Snake Game Tutorial Part 5](https://docs.near.org/docs/tutorials/react-snake-game-part-5)\n",
      "- [NEAR React Snake Game Tutorial Part 6](https://docs.near.org/docs/tutorials/react-snake-game-part-6)\n",
      "- [NEAR React Snake Game Tutorial Part 7](https://docs.near.org/docs/tutorials/react-snake-game-part-7)\n",
      "- [NEAR React Snake Game Tutorial Part 8](https://docs.near.org/docs/tutorials/react-snake-game-part-8)\n",
      "- [NEAR React Snake Game Tutorial Part 9](https://docs.near.org/docs/tutorials/react-snake-game-part-9)\n",
      "- [NEAR React Snake Game Tutorial Part 10](https://docs.near.org/docs/tutorials/react-snake-game-part-10)\n",
      "- [NEAR React Snake Game Tutorial Part 11](https://docs.near.org/docs/tutorials/react-snake-game-part-11)\n",
      "- [NEAR React Snake Game Tutorial Part 12](https://docs.near.org/docs/tutorials/react-snake-game-part-12)\n",
      "- [NEAR React Snake Game Tutorial Part 13](https://docs.near.org/docs/tutorials/react-snake-game-part-13)\n",
      "- [NEAR React Snake Game Tutorial Part 14](https://docs.near.org/docs/tutorials/react-snake-game-part-14)\n",
      "- [NEAR React Snake Game Tutorial Part 15](https://docs.near.org/docs/tutorials/react-snake-game-part-15)\n",
      "- [NEAR React Snake Game Tutorial Part 16](https://docs.near.org/docs/tutorials/react-snake-game-part-16)\n",
      "- [NEAR React Snake Game Tutorial Part 17](https://docs.near.org/docs/tutorials/react-snake-game-part-17)\n",
      "- [NEAR React Snake Game Tutorial Part 18](https://docs.near.org/docs/tutorials/react-snake-game-part-18)\n",
      "- [NEAR React Snake Game Tutorial Part 19](https://docs.near.org/docs/tutorials/react-snake-game-part-19)\n",
      "- [NEAR React Snake Game Tutorial Part 20](\n",
      "Functionality Score: 0.00\n",
      "Usability Score: 0.00\n",
      "Code Quality Score: 0.08\n",
      "\n",
      "Evaluation Metrics for NEARCoder-3b:\n",
      " {'name': ['NEAR Calendar App', 'NEAR Snake Game'], 'description': ['A simple calendar dApp on NEAR for scheduling events.', 'A basic snake game dApp on NEAR.'], 'generated_code': ['A simple calendar dApp on NEAR for scheduling events.\\n\\n## Getting started\\n\\n### Prerequisites\\n\\n- [Node.js](https://nodejs.org/)\\n- [Yarn](https://yarnpkg.com/)\\n- [NEAR CLI](https://github.com/near/near-cli)\\n- [NEAR testnet](https://testnet.near.org/)\\n\\n### Installation\\n\\nClone the repository:\\n\\n```bash\\ngit clone https://github.com/near-examples/calendar-contract.git\\ncd calendar-contract\\n```\\n\\nInstall dependencies:\\n\\n```bash\\nyarn\\n```\\n\\n### Compile and deploy\\n\\nCompile the contract:\\n\\n```bash\\nyarn build\\n```\\n\\nDeploy the contract to the testnet:\\n\\n```bash\\nyarn deploy\\n```\\n\\n### Run tests\\n\\nRun the tests:\\n\\n```bash\\nyarn test\\n```\\n\\n## License\\n\\n[MIT](LICENSE)\\n\\n## Contributing\\n\\nSee [CONTRIBUTING.md](CONTRIBUTING.md)\\n\\n## Code of Conduct\\n\\nSee [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md)\\n\\n## Acknowledgements\\n\\n- [NEAR](https://near.org/)\\n- [NEAR Protocol](https://protocol.near.org/)\\n- [NEAR Developer Resources](https://docs.near.org/)\\n- [NEAR Developer Guides](https://docs.near.org/docs/develop)\\n- [NEAR Community](https://community.near.org/)\\n- [NEAR Governance](https://near.org/governance)\\n- [NEAR Wallet](https://wallet.near.org/)\\n- [NEAR Explorer](https://explorer.near.org/)\\n- [NEAR Wallet Explorer](https://wallet.explorer.near.org/)\\n- [NEAR API Reference](https://docs.near.org/docs/develop/front-end/js-api-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Developer Tools](https://docs.near.org/docs/develop/tools)\\n- [NEAR Developer Tools Reference](https://docs.near.org/docs/develop/tools/tools-reference)\\n- [NEAR Developer Guides](https://docs.near.org/docs/develop)\\n- [NEAR Developer Guides Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts](https://github.com/near/near-contract-standards)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [NEAR Contracts Reference](https://docs.near.org/docs/develop/front-end/js-contract-reference)\\n- [', 'A basic snake game dApp on NEAR.\\n\\n## Getting started\\n\\n### Prerequisites\\n\\n- [Node.js](https://nodejs.org/)\\n- [Yarn](https://yarnpkg.com/)\\n- [NEAR CLI](https://github.com/near/near-cli)\\n\\n### Installation\\n\\nClone the repository:\\n\\n```bash\\ngit clone https://github.com/near-examples/snake-game.git\\ncd snake-game\\n```\\n\\nInstall dependencies:\\n\\n```bash\\nyarn install\\n```\\n\\n### Compile and deploy\\n\\nCompile and deploy the contract:\\n\\n```bash\\nyarn deploy\\n```\\n\\n### Run tests\\n\\nRun unit tests:\\n\\n```bash\\nyarn test\\n```\\n\\nRun integration tests:\\n\\n```bash\\nyarn test:integration\\n```\\n\\n### Run the app\\n\\nStart the app:\\n\\n```bash\\nyarn start\\n```\\n\\nOpen [http://localhost:3000](http://localhost:3000) to view it in the browser.\\n\\n## License\\n\\n[MIT](LICENSE)\\n\\n## Acknowledgements\\n\\n- [NEAR](https://near.org/)\\n- [NEAR Developer Hub](https://docs.near.org/docs/develop/front-end/react)\\n- [NEAR React Starter Kit](https://github.com/near-examples/react-starter-kit)\\n- [NEAR React Snake Game](https://github.com/near-examples/react-snake-game)\\n- [NEAR React Snake Game Tutorial](https://docs.near.org/docs/tutorials/react-snake-game)\\n- [NEAR React Snake Game Tutorial Part 2](https://docs.near.org/docs/tutorials/react-snake-game-part-2)\\n- [NEAR React Snake Game Tutorial Part 3](https://docs.near.org/docs/tutorials/react-snake-game-part-3)\\n- [NEAR React Snake Game Tutorial Part 4](https://docs.near.org/docs/tutorials/react-snake-game-part-4)\\n- [NEAR React Snake Game Tutorial Part 5](https://docs.near.org/docs/tutorials/react-snake-game-part-5)\\n- [NEAR React Snake Game Tutorial Part 6](https://docs.near.org/docs/tutorials/react-snake-game-part-6)\\n- [NEAR React Snake Game Tutorial Part 7](https://docs.near.org/docs/tutorials/react-snake-game-part-7)\\n- [NEAR React Snake Game Tutorial Part 8](https://docs.near.org/docs/tutorials/react-snake-game-part-8)\\n- [NEAR React Snake Game Tutorial Part 9](https://docs.near.org/docs/tutorials/react-snake-game-part-9)\\n- [NEAR React Snake Game Tutorial Part 10](https://docs.near.org/docs/tutorials/react-snake-game-part-10)\\n- [NEAR React Snake Game Tutorial Part 11](https://docs.near.org/docs/tutorials/react-snake-game-part-11)\\n- [NEAR React Snake Game Tutorial Part 12](https://docs.near.org/docs/tutorials/react-snake-game-part-12)\\n- [NEAR React Snake Game Tutorial Part 13](https://docs.near.org/docs/tutorials/react-snake-game-part-13)\\n- [NEAR React Snake Game Tutorial Part 14](https://docs.near.org/docs/tutorials/react-snake-game-part-14)\\n- [NEAR React Snake Game Tutorial Part 15](https://docs.near.org/docs/tutorials/react-snake-game-part-15)\\n- [NEAR React Snake Game Tutorial Part 16](https://docs.near.org/docs/tutorials/react-snake-game-part-16)\\n- [NEAR React Snake Game Tutorial Part 17](https://docs.near.org/docs/tutorials/react-snake-game-part-17)\\n- [NEAR React Snake Game Tutorial Part 18](https://docs.near.org/docs/tutorials/react-snake-game-part-18)\\n- [NEAR React Snake Game Tutorial Part 19](https://docs.near.org/docs/tutorials/react-snake-game-part-19)\\n- [NEAR React Snake Game Tutorial Part 20]('], 'functionality': [0.0, 0.0], 'usability': [0.0, 0.0], 'code_quality': [0.16666666666666666, 0.08333333333333333]}\n"
     ]
    }
   ],
   "source": [
    "# Evaluation functions for NEAR dApp generation\n",
    "def evaluate_functionality(generated_code, requirements):\n",
    "    functionality_score = sum(1 for requirement in requirements if all(keyword.lower() in generated_code.lower() for keyword in requirement.split()))\n",
    "    return functionality_score / len(requirements) if requirements else 0\n",
    "\n",
    "def evaluate_usability(generated_code):\n",
    "    ui_elements = ['button', 'input', 'form', 'interface', 'user', 'display']\n",
    "    usability_score = sum(1 for element in ui_elements if element.lower() in generated_code.lower())\n",
    "    return usability_score / len(ui_elements)\n",
    "\n",
    "def evaluate_code_quality(generated_code):\n",
    "    quality_indicators = ['def ', 'function ', 'class ', '#', 'import ', 'contract ']\n",
    "    code_quality_score = sum(1 for indicator in quality_indicators if indicator.lower() in generated_code.lower())\n",
    "    total_lines = len(generated_code.split('\\n'))\n",
    "    indented_lines = len([line for line in generated_code.split('\\n') if re.match(r'^\\s+', line)])\n",
    "    indentation_score = indented_lines / total_lines if total_lines > 0 else 0\n",
    "    return (code_quality_score / len(quality_indicators) + indentation_score) / 2\n",
    "\n",
    "def evaluate_basic_near_dapp_development(model, tokenizer, dataset):\n",
    "    text_generation_pipeline = pipeline(\"text-generation\", model=model, tokenizer=tokenizer)\n",
    "    evaluation_metrics = {\n",
    "        \"name\": [],\n",
    "        \"description\": [],\n",
    "        \"generated_code\": [],\n",
    "        \"functionality\": [],\n",
    "        \"usability\": [],\n",
    "        \"code_quality\": []\n",
    "    }\n",
    "\n",
    "    for project in dataset:\n",
    "        generated_code = text_generation_pipeline(\n",
    "            project[\"description\"], \n",
    "            max_length=1024, \n",
    "            num_return_sequences=1,\n",
    "            truncation=True,\n",
    "            pad_token_id=tokenizer.eos_token_id\n",
    "        )[0][\"generated_text\"]\n",
    "        \n",
    "        functionality_score = evaluate_functionality(generated_code, project[\"requirements\"])\n",
    "        usability_score = evaluate_usability(generated_code)\n",
    "        code_quality_score = evaluate_code_quality(generated_code)\n",
    "\n",
    "        # Collect evaluation metrics\n",
    "        evaluation_metrics[\"name\"].append(project[\"name\"])\n",
    "        evaluation_metrics[\"description\"].append(project[\"description\"])\n",
    "        evaluation_metrics[\"generated_code\"].append(generated_code)\n",
    "        evaluation_metrics[\"functionality\"].append(functionality_score)\n",
    "        evaluation_metrics[\"usability\"].append(usability_score)\n",
    "        evaluation_metrics[\"code_quality\"].append(code_quality_score)\n",
    "\n",
    "        # Print evaluation metrics for the current project\n",
    "        print(f\"Project Name: {project['name']}\")\n",
    "        print(f\"Project Description: {project['description']}\")\n",
    "        print(f\"Generated Code:\\n{generated_code}\")\n",
    "        print(f\"Functionality Score: {functionality_score:.2f}\")\n",
    "        print(f\"Usability Score: {usability_score:.2f}\")\n",
    "        print(f\"Code Quality Score: {code_quality_score:.2f}\\n\")\n",
    "\n",
    "    return evaluation_metrics\n",
    "\n",
    "# Evaluate each preloaded model on dApp generation\n",
    "for model_name in model_names.keys():\n",
    "    print(f\"Evaluating model: {model_name}\")\n",
    "    model = models[model_name][\"model\"]\n",
    "    tokenizer = models[model_name][\"tokenizer\"]\n",
    "    evaluation_metrics = evaluate_basic_near_dapp_development(model, tokenizer, dapp_projects)\n",
    "    print(f\"Evaluation Metrics for {model_name}:\\n\", evaluation_metrics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "32c82bd7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                        model  humanEval_bleu nearQuestions_bleu  \\\n",
      "0               StarCoder2-3b           0.076               0.53   \n",
      "1   NEAR-preTrainedStarCoder2           0.080               0.52   \n",
      "2  NEAR-structTunedStarCoder2           0.069               0.54   \n",
      "3                NEARCoder-3b           0.075               0.49   \n",
      "\n",
      "  nearQuestions_rouge1  dAppTrees_structure  dAppTrees_comment  \\\n",
      "0                 0.28               0.0240                0.0   \n",
      "1                 0.29               0.0275                0.0   \n",
      "2                 0.26               0.0290                1.0   \n",
      "3                 0.27               0.0245                0.5   \n",
      "\n",
      "   dAppGen_functionality  dAppGen_usability  dAppGen_quality  \n",
      "0                    0.0               0.08           0.0830  \n",
      "1                    0.0               0.08           0.0830  \n",
      "2                    0.0               0.00           0.0830  \n",
      "3                    0.0               0.00           0.1215  \n"
     ]
    }
   ],
   "source": [
    "data = [\n",
    "    ['StarCoder2-3b', \n",
    "     np.mean([0.07, 0.04, 0.00, 0.43, 0.05, 0.03, 0.03, 0.00, 0.05, 0.06]), \n",
    "     '0.53',\n",
    "     '0.28',\n",
    "     np.mean([0.028, 0.020]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.0, 0.16]),\n",
    "     np.mean([0.083, 0.083])\n",
    "    ], \n",
    "    ['NEAR-preTrainedStarCoder2', \n",
    "     np.mean([0.00, 0.04, 0.00, 0.30, 0.13, 0.07, 0.14, 0.00, 0.04, 0.08]), \n",
    "     '0.52',\n",
    "     '0.29',\n",
    "     np.mean([0.028, 0.027]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.16, 0.0]),\n",
    "     np.mean([0.083, 0.083])\n",
    "    ], \n",
    "    ['NEAR-structTunedStarCoder2', \n",
    "     np.mean([0.05, 0.09, 0.00, 0.29, 0.04, 0.04, 0.07, 0.00, 0.02, 0.09]), \n",
    "     '0.54',\n",
    "     '0.26',\n",
    "     np.mean([0.033, 0.025]),\n",
    "     np.mean([1.0, 1.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.083, 0.083])\n",
    "    ], \n",
    "    ['NEARCoder-3b', \n",
    "     np.mean([0.00, 0.00, 0.00, 0.13, 0.11, 0.07, 0.00, 0.00, 0.13, 0.31]), \n",
    "     '0.49',\n",
    "     '0.27',\n",
    "     np.mean([0.020, 0.029]),\n",
    "     np.mean([0.0, 1.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.0, 0.0]),\n",
    "     np.mean([0.16, 0.083])\n",
    "    ]\n",
    "]\n",
    "\n",
    "df = pd.DataFrame(data, columns=['model',\n",
    "                                 'humanEval_bleu',\n",
    "                                 'nearQuestions_bleu',\n",
    "                                 'nearQuestions_rouge1',\n",
    "                                 'dAppTrees_structure',\n",
    "                                 'dAppTrees_comment',\n",
    "                                 'dAppGen_functionality',\n",
    "                                 'dAppGen_usability',\n",
    "                                 'dAppGen_quality',\n",
    "                                ])\n",
    "print(df)\n",
    "\n",
    "# Save DataFrame to CSV file\n",
    "csv_filename = \"NEARCoderEval_metrics.csv\"\n",
    "df.to_csv(csv_filename, index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57f84bb9",
   "metadata": {},
   "source": [
    "### 3. Compare evaluation results\n",
    "\n",
    "- Look where NEARCoder performs better or worst than StarCoder, or where the finetuning doesn't change performance.\n",
    "- Use graphs to visualize the results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "7551f0a8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                        model  humanEval_bleu  nearQuestions_bleu  \\\n",
      "0               StarCoder2-3b           0.076                0.53   \n",
      "1   NEAR-preTrainedStarCoder2           0.080                0.52   \n",
      "2  NEAR-structTunedStarCoder2           0.069                0.54   \n",
      "3                NEARCoder-3b           0.075                0.49   \n",
      "\n",
      "   nearQuestions_rouge1  dAppTrees_structure  dAppTrees_comment  \\\n",
      "0                  0.28               0.0240                0.0   \n",
      "1                  0.29               0.0275                0.0   \n",
      "2                  0.26               0.0290                1.0   \n",
      "3                  0.27               0.0245                0.5   \n",
      "\n",
      "   dAppGen_functionality  dAppGen_usability  dAppGen_quality  \n",
      "0                    0.0               0.08           0.0830  \n",
      "1                    0.0               0.08           0.0830  \n",
      "2                    0.0               0.00           0.0830  \n",
      "3                    0.0               0.00           0.1215  \n"
     ]
    }
   ],
   "source": [
    "NEARCoderEval = pd.read_csv(\"NEARCoderEval_metrics.csv\")\n",
    "print(NEARCoderEval)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "b839b987",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/anaconda3/lib/python3.11/site-packages/evaluate/visualization.py:47: UserWarning: Attempting to set identical low and high ylims makes transformation singular; automatically expanding.\n",
      "  ax.set_ylim(*ranges[j])\n",
      "/opt/anaconda3/lib/python3.11/site-packages/evaluate/visualization.py:56: UserWarning: Attempting to set identical low and high ylims makes transformation singular; automatically expanding.\n",
      "  ax.set_ylim(*ranges[j])\n",
      "/opt/anaconda3/lib/python3.11/site-packages/evaluate/visualization.py:124: RuntimeWarning: invalid value encountered in scalar divide\n",
      "  sdata.append((d - y1) / (y2 - y1) * (x2 - x1) + x1)\n",
      "/var/folders/92/7zcym99s7h54kvkc6wxlmn5w0000gn/T/ipykernel_75559/1229957070.py:21: UserWarning: FigureCanvasAgg is non-interactive, and thus cannot be shown\n",
      "  plot.show()\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# visualize models' performance on a graph\n",
    "metrics = [\n",
    "    'humanEval_bleu', \n",
    "    'nearQuestions_bleu', \n",
    "    'nearQuestions_rouge1', \n",
    "    'dAppTrees_structure', \n",
    "    'dAppTrees_comment', \n",
    "    'dAppGen_functionality', \n",
    "    'dAppGen_usability', \n",
    "    'dAppGen_quality'\n",
    "]\n",
    "\n",
    "# Convert DataFrame to a list of dictionaries\n",
    "data_for_plot = []\n",
    "for index, row in NEARCoderEval.iterrows():\n",
    "    entry = {metric: row[metric] for metric in metrics}\n",
    "    data_for_plot.append(entry)\n",
    "\n",
    "# Create radar plot using evaluate\n",
    "plot = radar_plot(data=data_for_plot, model_names=NEARCoderEval['model'])\n",
    "plot.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7b440509",
   "metadata": {},
   "outputs": [],
   "source": [
    "# which model performs best on which task?\n"
   ]
  }
 ],
 "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": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}