atropos/environments/eval_environments/gpqa_eval.py

"""
GPQA-Diamond Evaluation Environment for Atropos (Generative/Reasoning Mode)

This environment evaluates models on the GPQA (Graduate-Level Google-Proof Q&A) benchmark
using a generative approach where models can reason before answering.

Dataset: Idavidrein/gpqa (gpqa_diamond subset)
Paper: https://arxiv.org/abs/2311.12022

GPQA is a dataset of 448 expert-written multiple-choice questions in biology,
physics, and chemistry, designed to test graduate-level reasoning. The questions
are extremely difficult—PhD-level experts score about 65%, skilled non-experts
34% (even with web access), and GPT-4 around 39%.

The evaluation follows the lighteval generative approach:
- Models are prompted to "think step by step before answering"
- Models output their reasoning followed by "Answer: X"
- Answer is extracted using regex patterns from the response
- Simple string matching validates the extracted answer

Supports optional thinking mode with <think></think> tags for extended reasoning.
"""

import asyncio
import os
import random
import re
import time
from string import ascii_uppercase
from typing import Dict, List, Optional, Tuple

import wandb
from datasets import load_dataset
from eval_helpers import (
    build_mcqa_fallback_patterns,
    create_system_content,
    extract_letter_from_answer_tag,
    extract_thinking_content,
    get_default_thinking_prompt,
    save_eval_results,
    validate_thinking_format,
)
from pydantic import Field
from tqdm.asyncio import tqdm_asyncio

from atroposlib.envs.base import (
    APIServerConfig,
    BaseEnv,
    BaseEnvConfig,
    EvalHandlingEnum,
)

# GPQA prompt template with <answer> tag instruction
GPQA_PROMPT_TEMPLATE = """Answer the following multiple choice question. Think step by step before answering.

Provide your final answer within <answer></answer> tags, containing only the letter (A, B, C, or D).

Example format:
<answer>A</answer>

{Question}

A) {A}
B) {B}
C) {C}
D) {D}"""


class GPQAEvalConfig(BaseEnvConfig):
    """Configuration for GPQA evaluation environment (generative mode)."""

    # Thinking mode configuration
    thinking_mode: bool = Field(
        default=True,
        description="Whether to enable thinking mode with <think></think> tags.",
    )

    custom_thinking_prompt: Optional[str] = Field(
        default=None,
        description="Custom thinking prompt. If None, uses the default thinking prompt.",
    )

    # Dataset configuration
    dataset_name: str = Field(
        default="Idavidrein/gpqa",
        description="HuggingFace dataset name for GPQA.",
    )

    subset: str = Field(
        default="gpqa_diamond",
        description="Dataset subset: gpqa_diamond, gpqa_extended, or gpqa_main.",
    )

    eval_split: str = Field(
        default="train",
        description="Dataset split to use for evaluation (GPQA uses 'train' for eval).",
    )

    # Shuffle seed for reproducibility (GPQA shuffles answer positions)
    shuffle_seed: Optional[int] = Field(
        default=42,
        description="Seed for shuffling answer positions. Set to None for random shuffling each run.",
    )

    # Model generation configuration
    eval_temperature: float = Field(
        default=0.6,
        description="Temperature for evaluation (0.0 for deterministic).",
    )

    eval_max_tokens: int = Field(
        default=0,
        description="Maximum tokens for evaluation responses. Set high to allow reasoning.",
    )

    # Prompt configuration
    custom_system_prompt: Optional[str] = Field(
        default=None,
        description="Custom system prompt to append after thinking prompt (if thinking_mode) or use directly.",
    )

    # Retry configuration
    max_retries: int = Field(
        default=3,
        ge=1,
        description="Maximum retries for failed API calls.",
    )

    retry_delay: float = Field(
        default=1.0,
        ge=0.0,
        description="Delay between retry attempts in seconds.",
    )

    min_response_length: int = Field(
        default=1,
        ge=1,
        description="Minimum response length to consider valid.",
    )

    # Debug configuration
    full_debug: bool = Field(
        default=False,
        description="Enable verbose debug logging.",
    )


class GPQAEvalEnv(BaseEnv):
    """
    GPQA-Diamond Evaluation Environment for Atropos (Generative/Reasoning Mode).

    Evaluates models on the GPQA benchmark using a generative approach where
    models reason before answering graduate-level science questions.

    Key features:
    - Loads GPQA dataset from HuggingFace (Idavidrein/gpqa)
    - Uses lighteval's exact prompt format for generative evaluation
    - Optional thinking mode with <think></think> tags
    - Extracts answer letters from patterns like "Answer: A"
    - Shuffles answer positions for fair evaluation
    """

    name = "gpqa_eval"
    env_config_cls = GPQAEvalConfig

    def __init__(
        self,
        config: GPQAEvalConfig,
        server_configs: List[APIServerConfig],
        slurm=True,
        testing=False,
    ):
        super().__init__(config, server_configs, slurm, testing)
        self.config: GPQAEvalConfig = config

        # Initialize metrics tracking
        self.eval_metrics = []

        # Set up random seed for answer shuffling
        if self.config.shuffle_seed is not None:
            self.shuffle_rng = random.Random(self.config.shuffle_seed)
        else:
            self.shuffle_rng = random.Random()

        # Pre-compile regex patterns for thinking mode
        self._think_pattern = re.compile(r"<think>")
        self._think_close_pattern = re.compile(r"</think>")
        self._think_content_pattern = re.compile(r"</think>\s*(.*)", re.DOTALL)
        self._thinking_extract_pattern = re.compile(r"<think>(.*?)</think>", re.DOTALL)

        # Pre-compile regex for <answer></answer> tag extraction (primary method)
        self._answer_tag_pattern = re.compile(
            r"<answer>(.*?)</answer>", re.DOTALL | re.IGNORECASE
        )

        # Build fallback answer extraction patterns
        self._build_extraction_patterns()

    def _get_thinking_prompt(self) -> str:
        """Get thinking system prompt."""
        return get_default_thinking_prompt(self.config.custom_thinking_prompt)

    def _create_system_content(self) -> Optional[str]:
        """Create system message content based on thinking mode."""
        return create_system_content(
            self.config.thinking_mode,
            self.config.custom_thinking_prompt,
            self.config.custom_system_prompt,
        )

    def _build_extraction_patterns(self):
        """Build regex patterns for extracting answer letters from model responses."""
        letters = "ABCD"
        letter_pattern = rf"([{letters}]|\([{letters}]\))"

        # Patterns ordered by priority (most specific first)
        self._pattern_final_answer_hope = re.compile(
            rf"(?i:final\s+answer\s+is)\s*:?\s*{letter_pattern}\.?\s*I\s*hope",
            re.IGNORECASE,
        )
        self._pattern_final_answer_is = re.compile(
            rf"(?i:final\s+answer).{{0,100}}?\s+is\s*:?\s*{letter_pattern}",
            re.IGNORECASE | re.DOTALL,
        )
        self._pattern_the_answer_is = re.compile(
            rf"(?i:the\s+answer\s+is)\s*:?\s*{letter_pattern}", re.IGNORECASE
        )
        self._pattern_answer_colon = re.compile(
            rf"(?i:answer)\s*:\s*.{{0,50}}?{letter_pattern}", re.IGNORECASE | re.DOTALL
        )
        self._pattern_answer_space = re.compile(
            rf"(?i:answer)\s+{letter_pattern}", re.IGNORECASE
        )
        self._pattern_start = re.compile(
            rf"^\s*\**{letter_pattern}\**[\s\.\)\:]", re.IGNORECASE
        )
        self._pattern_line_start = re.compile(
            rf"\n\s*\**{letter_pattern}\**[\s\.\)\:]", re.IGNORECASE
        )
        self._pattern_standalone = re.compile(rf"\b{letter_pattern}\b", re.IGNORECASE)

        self._extraction_patterns = [
            (0, self._pattern_final_answer_hope, "final_answer_hope"),
            (50, self._pattern_final_answer_is, "final_answer_is"),
            (75, self._pattern_the_answer_is, "the_answer_is"),
            (100, self._pattern_answer_colon, "answer_colon"),
            (150, self._pattern_answer_space, "answer_space"),
            (200, self._pattern_start, "start"),
            (210, self._pattern_line_start, "line_start"),
            (250, self._pattern_standalone, "standalone"),
        ]

    @classmethod
    def config_init(cls) -> Tuple[GPQAEvalConfig, List[APIServerConfig]]:
        """Initialize default configuration for the environment."""
        env_config = GPQAEvalConfig(
            tokenizer_name="NousResearch/Hermes-3-Llama-3.1-8B",
            group_size=1,
            use_wandb=True,
            max_num_workers_per_node=128,
            rollout_server_url="http://localhost:8000",
            total_steps=1,
            batch_size=1,
            steps_per_eval=1,
            inference_weight=1.0,
            wandb_name="gpqa_eval",
            eval_handling=EvalHandlingEnum.STOP_TRAIN,
            max_eval_workers=256,
            max_num_workers=1024,
            # GPQA-specific defaults
            dataset_name="Idavidrein/gpqa",
            subset="gpqa_diamond",
            eval_temperature=0.6,
            eval_max_tokens=0,  # Use model default
            thinking_mode=True,
        )

        server_configs = [
            APIServerConfig(
                model_name="Hermes-3-Llama-3.1-8B",
                base_url="http://localhost:9000/v1",
                api_key=os.getenv("OPENAI_API_KEY", "none"),
                num_max_requests_at_once=32,
                num_requests_for_eval=1024,
            ),
        ]

        return env_config, server_configs

    async def setup(self) -> None:
        """Load the GPQA dataset and prepare for evaluation."""
        print(f"\nGPQA Evaluation Setup (Generative Mode):")
        print(f"  Dataset: {self.config.dataset_name}")
        print(f"  Subset: {self.config.subset}")
        print(f"  Max tokens for reasoning: {self.config.eval_max_tokens}")
        print(f"  Evaluation split: {self.config.eval_split}")
        print(f"  Thinking mode: {self.config.thinking_mode}")
        if self.config.thinking_mode:
            print(f"  Thinking prompt: {self._get_thinking_prompt()[:100]}...")

        # Load GPQA dataset
        try:
            dataset = load_dataset(
                self.config.dataset_name,
                self.config.subset,
                split=self.config.eval_split,
            )
            self.eval_data = list(dataset)
            print(f"  Loaded {len(self.eval_data)} evaluation items")
        except Exception as e:
            print(f"Error loading GPQA dataset: {e}")
            raise

        # Process items - shuffle answer positions
        self.all_eval_items = []
        for item in self.eval_data:
            processed = self._process_gpqa_item(item)
            self.all_eval_items.append(processed)

        self.iter = 0

    def _process_gpqa_item(self, item: Dict) -> Dict:
        """
        Process a GPQA item - shuffle answer positions following lighteval.

        GPQA has: Question, Correct Answer, Incorrect Answer 1/2/3
        We need to shuffle them into A/B/C/D positions.
        """
        # Get answers
        correct_answer = item["Correct Answer"]
        incorrect_answers = [
            item["Incorrect Answer 1"],
            item["Incorrect Answer 2"],
            item["Incorrect Answer 3"],
        ]

        # Randomly place correct answer
        gold_index = self.shuffle_rng.randint(0, 3)
        choices = incorrect_answers.copy()
        choices.insert(gold_index, correct_answer)

        return {
            "question": item["Question"],
            "choices": choices,
            "gold_index": gold_index,
            "gold_letter": ascii_uppercase[gold_index],
            "subdomain": item.get("Subdomain", "unknown"),
            "original_item": item,
        }

    def _format_gpqa_prompt(self, question: str, choices: List[str]) -> str:
        """
        Format a GPQA question using the lighteval template.

        Uses the exact prompt format from lighteval's gpqa_instruct_prompt.
        """
        return GPQA_PROMPT_TEMPLATE.format(
            Question=question.strip(),
            A=choices[0].strip(),
            B=choices[1].strip(),
            C=choices[2].strip(),
            D=choices[3].strip(),
        )

    def _validate_thinking_format(self, response: str) -> Tuple[bool, str]:
        """Validate thinking format and extract content after </think> tags."""
        if not self.config.thinking_mode:
            return True, response

        think_open_count = len(self._think_pattern.findall(response))
        think_close_count = len(self._think_close_pattern.findall(response))

        if think_open_count != 1 or think_close_count != 1:
            return False, response

        match = self._think_content_pattern.search(response)
        if match:
            return True, match.group(1).strip()
        else:
            return False, response

    def _extract_thinking_content(self, response: str) -> Optional[str]:
        """Extract the content inside <think></think> tags."""
        match = self._thinking_extract_pattern.search(response)
        if match:
            return match.group(1).strip()
        return None

    def _extract_answer(
        self, response: str, num_choices: int = 4, choices: Optional[List[str]] = None
    ) -> Tuple[Optional[str], str]:
        """
        Extract the answer letter from the model's response.

        Primary method: Look for <answer></answer> tags, or match against choice texts.
        Fallback: Use priority-ordered regex patterns.
        """
        if not response:
            return None, "empty_response"

        valid_letters = set(ascii_uppercase[:num_choices])

        # PRIMARY: Try <answer></answer> tags first
        # Also matches against choice texts if provided
        letter, method = extract_letter_from_answer_tag(
            response, valid_letters, debug=self.config.full_debug, choices=choices
        )
        if letter:
            return letter, method

        # FALLBACK: Try each pattern in priority order
        for priority, pattern, method_name in self._extraction_patterns:
            matches = pattern.findall(response)
            if matches:
                match = (
                    matches[-1]
                    if method_name
                    in [
                        "final_answer_is",
                        "the_answer_is",
                        "answer_colon",
                        "answer_space",
                    ]
                    else matches[0]
                )
                if isinstance(match, tuple):
                    match = match[0]
                letter = match.strip("()").upper()

                if letter in valid_letters:
                    if self.config.full_debug:
                        print(
                            f"    Extracted '{letter}' using fallback method '{method_name}'"
                        )
                    return letter, f"fallback_{method_name}"

        for letter in reversed(list(valid_letters)):
            if letter in response.upper():
                if self.config.full_debug:
                    print(
                        f"    Extracted '{letter}' using fallback 'last_valid_letter'"
                    )
                return letter, "fallback_last_valid_letter"

        return None, "no_match"

    async def get_next_item(self):
        """Get next item for training (not used in eval-only environment)."""
        self.iter += 1
        if self.all_eval_items:
            item = self.all_eval_items[self.iter % len(self.all_eval_items)]
            return item
        return None

    async def collect_trajectories(self, item):
        """Collect trajectories (not used in eval-only environment)."""
        return None, []

    async def score(self, rollout_group_data):
        """Score rollouts (not used in eval-only environment)."""
        return None

    async def rollout_and_score_eval(self, eval_item: Dict) -> Dict:
        """Evaluate a single GPQA question using generative mode."""
        try:
            question = eval_item.get("question", "")
            choices = eval_item.get("choices", [])
            gold_letter = eval_item.get("gold_letter", "A")
            subdomain = eval_item.get("subdomain", "unknown")

            if not question or len(choices) != 4:
                return {"is_correct": None, "sample": None}

            # Format the prompt (lighteval style)
            formatted_prompt = self._format_gpqa_prompt(question, choices)

            # Build messages
            messages = []
            system_content = self._create_system_content()
            if system_content:
                messages.append({"role": "system", "content": system_content})
            messages.append({"role": "user", "content": formatted_prompt})

            # Get model completion with retry logic
            model_response = None
            finish_reason = None
            for attempt in range(self.config.max_retries):
                try:
                    completion = await self.server.chat_completion(
                        messages=messages,
                        n=1,
                        temperature=self.config.eval_temperature,
                        max_tokens=self.config.eval_max_tokens,
                        split="eval",
                    )

                    if completion.choices and completion.choices[0].message.content:
                        model_response = completion.choices[0].message.content
                        finish_reason = getattr(
                            completion.choices[0], "finish_reason", None
                        )

                        if (
                            len(model_response.strip())
                            >= self.config.min_response_length
                        ):
                            break
                        elif attempt < self.config.max_retries - 1:
                            if self.config.full_debug:
                                print(f"  Response too short, retrying...")
                            await asyncio.sleep(self.config.retry_delay)

                except Exception as e:
                    # Always log API errors to help diagnose issues
                    print(
                        f"  API Error (attempt {attempt + 1}/{self.config.max_retries}): {type(e).__name__}: {e}"
                    )
                    if hasattr(e, "response"):
                        try:
                            print(
                                f"    Response: {e.response.text[:500] if hasattr(e.response, 'text') else e.response}"
                            )
                        except:
                            pass
                    if attempt < self.config.max_retries - 1:
                        await asyncio.sleep(self.config.retry_delay)
                    else:
                        print(f"  Failed after {self.config.max_retries} attempts")
                        return {"is_correct": None, "sample": None}

            if not model_response:
                return {"is_correct": None, "sample": None}

            # Validate thinking format if enabled
            format_valid, content_for_extraction = self._validate_thinking_format(
                model_response
            )

            # Extract thinking content for logging
            thinking_content = None
            if self.config.thinking_mode:
                thinking_content = self._extract_thinking_content(model_response)

            # Extract the answer (pass choices for exact text matching)
            extracted_answer, extraction_method = self._extract_answer(
                content_for_extraction, num_choices=4, choices=choices
            )

            # Check if correct
            is_correct = extracted_answer == gold_letter if extracted_answer else False

            # Build sample record
            sample = {
                "question": question,
                "choices": choices,
                "gold_answer": gold_letter,
                "model_response": model_response,
                "extracted_answer": extracted_answer,
                "extraction_method": extraction_method,
                "is_correct": is_correct,
                "subdomain": subdomain,
                "finish_reason": finish_reason,
                "response_length": len(model_response),
                "thinking_mode": self.config.thinking_mode,
                "format_valid": format_valid,
            }

            if self.config.thinking_mode:
                sample["thinking_content"] = thinking_content
                sample["response_after_think"] = (
                    content_for_extraction if format_valid else None
                )

            if self.config.full_debug:
                status = "✓" if is_correct else "✗"
                print(
                    f"  [{status}] {subdomain}: gold={gold_letter}, extracted={extracted_answer}"
                )

            return {"is_correct": is_correct, "sample": sample}

        except Exception as e:
            if self.config.full_debug:
                print(f"Error in rollout_and_score_eval: {e}")
                import traceback

                traceback.print_exc()
            return {"is_correct": None, "sample": None}

    async def evaluate(self, *args, **kwargs) -> None:
        """Run GPQA evaluation."""
        start_time = time.time()

        print(f"\n{'='*60}")
        print(f"Starting GPQA Evaluation (Generative/Reasoning Mode)")
        print(f"{'='*60}")
        print(f"  Subset: {self.config.subset}")
        print(f"  Total questions: {len(self.all_eval_items)}")
        print(f"  Max tokens (for reasoning): {self.config.eval_max_tokens}")
        print(f"  Thinking mode: {self.config.thinking_mode}")
        print(f"{'='*60}\n")

        try:
            eval_tasks = [
                self.rollout_and_score_eval(item) for item in self.all_eval_items
            ]
            results = await tqdm_asyncio.gather(*eval_tasks, desc="Evaluating GPQA")

            valid_results = [
                r
                for r in results
                if r and r.get("sample") is not None and r.get("is_correct") is not None
            ]

            if not valid_results:
                print("Warning: No valid evaluation results obtained")
                return

        except Exception as e:
            print(f"Error during evaluation: {e}")
            import traceback

            traceback.print_exc()
            return

        end_time = time.time()

        # Compute metrics
        samples = [r["sample"] for r in valid_results]

        # Overall accuracy
        total_correct = sum(1 for r in valid_results if r["is_correct"])
        total_count = len(valid_results)
        overall_accuracy = total_correct / total_count if total_count > 0 else 0.0

        # Per-subdomain accuracy
        subdomain_results = {}
        for sample in samples:
            subdomain = sample.get("subdomain", "unknown")
            if subdomain not in subdomain_results:
                subdomain_results[subdomain] = {"correct": 0, "total": 0}
            subdomain_results[subdomain]["total"] += 1
            if sample["is_correct"]:
                subdomain_results[subdomain]["correct"] += 1

        # Extraction method statistics
        extraction_methods = {}
        for sample in samples:
            method = sample.get("extraction_method", "unknown")
            if method not in extraction_methods:
                extraction_methods[method] = {"count": 0, "correct": 0}
            extraction_methods[method]["count"] += 1
            if sample["is_correct"]:
                extraction_methods[method]["correct"] += 1

        # Average response length
        response_lengths = [s.get("response_length", 0) for s in samples]
        avg_response_length = (
            sum(response_lengths) / len(response_lengths) if response_lengths else 0
        )

        # Format compliance
        format_compliant = sum(1 for s in samples if s.get("format_valid", True))
        format_compliance_rate = format_compliant / len(samples) if samples else 0.0

        # Thinking utilization
        thinking_utilization = 0
        if self.config.thinking_mode:
            thinking_utilization = sum(1 for s in samples if s.get("thinking_content"))

        # Build metrics dictionary
        eval_metrics = {
            "eval/overall_accuracy": overall_accuracy,
            "eval/total_questions": total_count,
            "eval/total_correct": total_correct,
            "eval/evaluation_time_seconds": end_time - start_time,
            "eval/avg_response_length": avg_response_length,
            "eval/format_compliance_rate": format_compliance_rate,
            "eval/thinking_mode_enabled": 1.0 if self.config.thinking_mode else 0.0,
        }

        if self.config.thinking_mode:
            thinking_utilization_rate = (
                thinking_utilization / len(samples) if samples else 0.0
            )
            eval_metrics["eval/thinking_utilization_rate"] = thinking_utilization_rate

        # Add subdomain metrics
        for subdomain, stats in subdomain_results.items():
            if stats["total"] > 0:
                subdom_accuracy = stats["correct"] / stats["total"]
                subdom_key = subdomain.replace(" ", "_").replace("-", "_").lower()
                eval_metrics[f"eval/subdomain_{subdom_key}_accuracy"] = subdom_accuracy

        # Store metrics for wandb logging
        self.eval_metrics = [(k, v) for k, v in eval_metrics.items()]

        # Print summary
        print(f"\n{'='*60}")
        print(f"GPQA Evaluation Results ({self.config.subset})")
        print(f"{'='*60}")
        print(
            f"Overall Accuracy: {overall_accuracy:.4f} ({total_correct}/{total_count})"
        )
        print(f"Evaluation Time: {end_time - start_time:.1f} seconds")
        print(f"Avg Response Length: {avg_response_length:.0f} chars")
        if self.config.thinking_mode:
            print(f"Format Compliance: {format_compliance_rate:.4f}")
            print(f"Thinking Utilization: {thinking_utilization}/{total_count}")

        print(f"\nSubdomain Breakdown:")
        for subdomain, stats in sorted(subdomain_results.items()):
            if stats["total"] > 0:
                subdom_acc = stats["correct"] / stats["total"]
                print(
                    f"  {subdomain}: {subdom_acc:.4f} ({stats['correct']}/{stats['total']})"
                )

        print(f"\nExtraction Method Statistics:")
        for method, stats in sorted(
            extraction_methods.items(), key=lambda x: -x[1]["count"]
        ):
            if stats["count"] > 0:
                method_acc = stats["correct"] / stats["count"]
                print(f"  {method}: {stats['count']} uses, {method_acc:.4f} accuracy")

        print(f"{'='*60}\n")

        # Log evaluation results
        try:
            await self.evaluate_log(
                metrics=eval_metrics,
                samples=samples,
                start_time=start_time,
                end_time=end_time,
                generation_parameters={
                    "temperature": self.config.eval_temperature,
                    "max_tokens": self.config.eval_max_tokens,
                    "thinking_mode": self.config.thinking_mode,
                    "subset": self.config.subset,
                    "mode": "generative",
                },
            )
        except Exception as e:
            print(f"Error logging evaluation results: {e}")

    async def wandb_log(self, wandb_metrics: Optional[Dict] = None):
        """Log metrics to wandb."""
        if wandb_metrics is None:
            wandb_metrics = {}

        for metric_name, metric_value in self.eval_metrics:
            wandb_metrics[metric_name] = metric_value
        self.eval_metrics = []

        wandb_metrics["config/thinking_mode"] = (
            1.0 if self.config.thinking_mode else 0.0
        )
        wandb_metrics["config/eval_max_tokens"] = self.config.eval_max_tokens
        wandb_metrics["config/subset"] = self.config.subset

        await super().wandb_log(wandb_metrics)


if __name__ == "__main__":
    GPQAEvalEnv.cli()