Merge remote-tracking branch 'origin/big-refactor' into calibration

lm_eval/decontamination/janitor.py

 import re
 import string
-import timeit
 import pickle
 import traceback
 from pprint import pprint
+from typing import Iterator, Sequence, TypeVar
 
 # This is a cpp module. Compile janitor_util.cpp with:
 # c++ -O3 -Wall -shared -std=c++11 -fPIC $(python3 -m pybind11 --includes) janitor_util.cpp -o janitor_util$(python3-config --extension-suffix) -undefined dynamic_lookup
@@ -16,10 +16,12 @@ except Exception:
     traceback.print_exc()
     JANITOR_CPP = False
 
+T = TypeVar("T")
+
 
 # Implementation from nltk source
 # https://www.nltk.org/_modules/nltk/util.html
-def form_ngrams(sequence, n):
+def form_ngrams(sequence: Iterator[T], n: int) -> Iterator[tuple[T, ...]]:
     history = []
     while n > 1:
         # PEP 479, prevent RuntimeError from being raised when StopIteration bubbles out of generator
@@ -36,7 +38,7 @@ def form_ngrams(sequence, n):
         del history[0]
 
 
-def word_ngrams(s, n):
+def word_ngrams(s: str, n: int) -> Iterator[str]:
     """Splits a string into ngram words"""
     tokens = s.split()  # not a generator :(
     ngram_seqs = form_ngrams(iter(tokens), n)
@@ -68,14 +70,14 @@ def word_ngrams(s, n):
 
 
 # https://stackoverflow.com/questions/13734451/string-split-with-indices-in-python
-def split_indices(s):
+def split_indices(s: str) -> Iterator[tuple[str, tuple[int, int]]]:
     """Splits a string on whitespaces and records the indices of each in the original string.
     @:return generator((word, (start_idx, end_idx)), ...)
     """
     return ((m.group(0), (m.start(), m.end() - 1)) for m in re.finditer(r"\S+", s))
 
 
-def word_ngrams_indices(s, n):
+def word_ngrams_indices(s: str, n: int) -> Iterator[tuple[str, tuple[int, int]]]:
     """Splits a string into pairs of (ngram words, their start/end indices)"""
     tokens_with_indices = split_indices(s)
 
@@ -104,16 +106,15 @@ def word_ngrams_indices(s, n):
 
 
 class Janitor:
-
     # FIXME delete_chars: Should anything else go here? Special chars?
     def __init__(
         self,
-        ngram_n=13,
-        window_to_remove=200,
-        too_dirty_cutoff=10,
-        minimum_slice_length=200,
-        delete_chars=string.punctuation,
-    ):
+        ngram_n: int = 13,
+        window_to_remove: int = 200,
+        too_dirty_cutoff: int = 10,
+        minimum_slice_length: int = 200,
+        delete_chars: str = string.punctuation,
+    ) -> None:
         self.ngram_n = ngram_n
         self.window_to_remove = window_to_remove
         self.too_dirty_cutoff = too_dirty_cutoff
@@ -135,11 +136,11 @@ class Janitor:
     # I/O for saving contamination ngrams
     ##############
 
-    def save_contamination_ngrams(self, filename):
+    def save_contamination_ngrams(self, filename: str) -> None:
         with open(filename, "wb") as fp:
             pickle.dump(filename, fp)
 
-    def load_contamination_ngrams(self, filename):
+    def load_contamination_ngrams(self, filename: str) -> None:
         with open(filename, "rb") as fp:
             self.dirt_ngrams = pickle.load(fp)
 
@@ -147,7 +148,7 @@ class Janitor:
     # Call these :)
     ##############
 
-    def register_contaminant(self, dirt_string):
+    def register_contaminant(self, dirt_string: str) -> None:
         """Register a string as contamination to be removed, e.g. a test set
         This breaks the dirt_string into ngrams to store for future cleaning"""
         if JANITOR_CPP:
@@ -156,7 +157,7 @@ class Janitor:
             print("WARNING: Janitor running in python mode")
             return self.register_contaminant_python(dirt_string)
 
-    def clean(self, dirty_string):
+    def clean(self, dirty_string: str) -> list[str]:
         """Clean a string (e.g. a training set) by removing all ngrams previously
         registered as contaminants. Returns a list of clean chunks, or empty if
         the string was too dirty"""
@@ -166,7 +167,9 @@ class Janitor:
             print("WARNING: Janitor running in python mode")
             return self.clean_python(dirty_string)
 
-    def _split_chunks(self, dirty_string, dirty_parts):
+    def _split_chunks(
+        self, dirty_string: str, dirty_parts: Sequence[tuple]
+    ) -> list[str]:
         clean_chunks = []
         splice_idx = 0
         end = -1
@@ -189,12 +192,12 @@ class Janitor:
     # Fast C++
     ##############
 
-    def register_contaminant_cpp(self, dirt_string):
+    def register_contaminant_cpp(self, dirt_string) -> None:
         self.dirt_ngrams.update(
             janitor_util.clean_ngram(dirt_string, self.delete_chars, self.ngram_n)
         )
 
-    def clean_cpp(self, dirty_string):
+    def clean_cpp(self, dirty_string: str) -> list[str]:
         contamination_indices = janitor_util.clean_ngram_with_indices(
             dirty_string, self.delete_chars, self.ngram_n
         )
@@ -204,15 +207,15 @@ class Janitor:
     # Slow python
     ##############
 
-    def normalize_string(self, s):
+    def normalize_string(self, s: str) -> str:
         return s.translate(self.translation_table)
 
-    def register_contaminant_python(self, dirt_string):
+    def register_contaminant_python(self, dirt_string: str) -> None:
         self.dirt_ngrams.update(
             word_ngrams(self.normalize_string(dirt_string), self.ngram_n)
         )
 
-    def clean_python(self, dirty_string):
+    def clean_python(self, dirty_string: str) -> list[str]:
         contamination_indices = (
             (None, *idx_pair)
             for dirty_ngram, idx_pair in word_ngrams_indices(dirty_string, self.ngram_n)

lm_eval/evaluator.py

@@ -40,11 +40,11 @@ def simple_evaluate(
     device=None,
     use_cache=None,
     limit=None,
-    bootstrap_iters=100000,
-    check_integrity=False,
+    bootstrap_iters: int = 100000,
+    check_integrity: bool = False,
     decontamination_ngrams_path=None,
-    write_out=False,
-    log_samples=True,
+    write_out: bool = False,
+    log_samples: bool = True,
 ):
     """Instantiate and evaluate a model on a list of tasks.
 
@@ -172,17 +172,17 @@ def evaluate(
     lm,
     task_dict,
     limit=None,
-    bootstrap_iters=100000,
+    bootstrap_iters: int = 100000,
     decontamination_ngrams_path=None,
-    write_out=False,
-    log_samples=True,
+    write_out: bool = False,
+    log_samples: bool = True,
 ):
     """Instantiate and evaluate a model on a list of tasks.
 
     :param lm: obj
         Language Model
     :param task_dict: dict[str, Task]
-        Dictionary of tasks. Tasks will be taken to have name task.EVAL_HARNESS_NAME if defined and type(task).__name__ otherwise.
+        Dictionary of tasks. Tasks will be taken to have name type(task).config.task .
     :param limit: int, optional
         Limit the number of examples per task (only use this for testing)
     :param bootstrap_iters:
@@ -216,7 +216,6 @@ def evaluate(
     padding_requests = collections.defaultdict(int)
 
     # Stores group related keys and values for group-aggregation
-    aggregate = collections.defaultdict(dict)
     task_groups = collections.defaultdict(dict)
 
     # get lists of each type of request
@@ -224,6 +223,7 @@ def evaluate(
         if type(task) == tuple:
             group, task = task
             task_groups[task_name] = group
+            aggregate[task_name] = {}
 
         versions[task_name] = task.VERSION
         configs[task_name] = dict(task.dump_config())
@@ -248,7 +248,8 @@ def evaluate(
                 # print the prompt for the first few documents
                 if inst.doc_id < 1:
                     eval_logger.info(
-                        f"Task: {task_name}; document {inst.doc_id}; context prompt (starting on next line):\n{inst.args[0]}\n(end of prompt on previous line)"
+                        f"Task: {task_name}; document {inst.doc_id}; context prompt (starting on next line):\
+\n{inst.args[0]}\n(end of prompt on previous line)\ntarget string or answer choice index (starting on next line):\n{task.doc_to_target(inst.doc)}\n(end of target on previous line)"
                     )
                     eval_logger.info(f"Request: {str(inst)}")
 
@@ -357,28 +358,35 @@ def evaluate(
             if type(items[0]) == tuple:
                 numitem = len(items[0])
 
-            # distributed gather requires all ranks to have same dimensions
-            # so we pad out with float32 min value
-            pad_value = torch.finfo(torch.float32).min
-            metrics_tensor = torch.tensor(items, device=lm.device)
-
-            original_dtype = metrics_tensor.dtype  # store original dtype
-            torch_device_tensor = lm.accelerator.pad_across_processes(
-                metrics_tensor.to(torch.float32), pad_index=pad_value
-            )
-            gathered_item = lm.accelerator.gather(torch_device_tensor)
+            if isinstance(items[0], (str, list)):
+                # handle the string case
+                gathered_items = [None] * lm.accelerator.num_processes
+                torch.distributed.all_gather_object(gathered_items, items)
 
-            if numitem > 0:
-                gathered_filtered = gathered_item[gathered_item[:, 0] != pad_value]
+                gathered_item = list(itertools.chain.from_iterable(gathered_items))
             else:
-                gathered_filtered = gathered_item[gathered_item != pad_value]
+                # distributed gather requires all ranks to have same dimensions
+                # so we pad out with float32 min value
+                pad_value = torch.finfo(torch.float32).min
+                metrics_tensor = torch.tensor(items, device=lm.device)
+
+                original_dtype = metrics_tensor.dtype  # store original dtype
+                torch_device_tensor = lm.accelerator.pad_across_processes(
+                    metrics_tensor.to(torch.float32), pad_index=pad_value
+                )
+                gathered_item = lm.accelerator.gather(torch_device_tensor)
 
-            gathered_item = (
-                gathered_filtered.to(original_dtype).cpu().detach().numpy().tolist()
-            )
-            # reconvert if we were passed a tuple of values
-            if numitem > 0:
-                gathered_item = [tuple(g) for g in gathered_item]
+                if numitem > 0:
+                    gathered_filtered = gathered_item[gathered_item[:, 0] != pad_value]
+                else:
+                    gathered_filtered = gathered_item[gathered_item != pad_value]
+
+                gathered_item = (
+                    gathered_filtered.to(original_dtype).cpu().detach().numpy().tolist()
+                )
+                # reconvert if we were passed a tuple of values
+                if numitem > 0:
+                    gathered_item = [tuple(g) for g in gathered_item]
 
             if lm.rank == 0:
                 vals_torch[(task_name, key, metric)] = gathered_item
@@ -401,16 +409,16 @@ def evaluate(
             #        | word_perplexity
             #        | byte_perplexity
             #        | bits_per_byte
-            if bool(task_groups):
+            if task_name in task_groups:
                 group_name = task_groups[task_name]
-                if metric not in aggregate[group_name]:
-                    aggregate[group_name][metric] = [task_score]
-                else:
+                if metric in list(aggregate[group_name].keys()):
                     aggregate[group_name][metric].append(task_score)
+                else:
+                    aggregate[group_name][metric] = [task_score]
 
             # hotfix: bleu, chrf, ter seem to be really expensive to bootstrap
             # so we run them less iterations. still looking for a cleaner way to do this
-            if bootstrap_iters > 0:
+            if False:  # bootstrap_iters > 0:
                 stderr = lm_eval.api.metrics.stderr_for_metric(
                     metric=task.aggregation()[metric],
                     bootstrap_iters=min(bootstrap_iters, 1000)

lm_eval/filters/__init__.py

@@ -17,14 +17,16 @@ FILTER_REGISTRY = {
 
 
 def get_filter(filter_name):
-    return FILTER_REGISTRY[filter_name]
+    if filter_name in FILTER_REGISTRY:
+        return FILTER_REGISTRY[filter_name]
+    else:
+        return filter_name
 
 
 def build_filter_ensemble(filter_name, components):
     """
     Create a filtering pipeline.
     """
-
     filters = []
     for (function, kwargs) in components:
         if kwargs is None:

lm_eval/filters/decontamination.py

@@ -9,7 +9,7 @@ class DecontaminationFilter(Filter):
 
     name = "track_decontamination"
 
-    def __init__(self, path):
+    def __init__(self, path) -> None:
         """
 
         TODO: make sure only ever run one time on the train set (should this be cached as a class var? keyed by value for "path").
@@ -17,7 +17,7 @@ class DecontaminationFilter(Filter):
         """
         self._decontam_results = None
 
-    def apply(self, reps):
+    def apply(self, resps, docs) -> None:
         """
         Return {"no_contamination", "only_contamination"} keys for the 2 different subsets
         """

lm_eval/filters/extraction.py

@@ -6,7 +6,9 @@ from lm_eval.api.filter import Filter
 class RegexFilter(Filter):
     """ """
 
-    def __init__(self, regex_pattern=r"#### (\-?[0-9\.\,]+)", fallback="[invalid]"):
+    def __init__(
+        self, regex_pattern: str = r"#### (\-?[0-9\.\,]+)", fallback: str = "[invalid]"
+    ) -> None:
         """
         pass a string `regex` to run `re.compile(r"regex")` on.
         `fallback` defines the output returned if no matches for the regex are located.
@@ -15,7 +17,7 @@ class RegexFilter(Filter):
         self.regex = re.compile(regex_pattern)
         self.fallback = fallback
 
-    def apply(self, resps):
+    def apply(self, resps, docs):
         # here, we assume we have a list, in which each element is
         # a list of model responses for some particular input/target pair.
         # so we process each of these (same input/target response sets)
@@ -41,12 +43,11 @@ class RegexFilter(Filter):
 class WhitespaceFilter(Filter):
     """ """
 
-    def __init__(self):
+    def __init__(self) -> None:
         pass
 
-    def apply(self, resps):
+    def apply(self, resps, docs):
         def filter_set(inst):
-
             filtered_resp = []
             for resp in inst:
                 if resp.startswith(" "):

lm_eval/filters/selection.py

@@ -4,12 +4,12 @@ from lm_eval.api.filter import Filter
 
 
 class TakeFirstFilter(Filter):
-    def __init__(self):
+    def __init__(self) -> None:
         """
         Can define custom behavior here, if an individual instantiation of a Filter class should have state.
         """
 
-    def apply(self, resps):
+    def apply(self, resps, docs):
         """
         Assuming each entry of `resps` is a list of model responses, we discard all but the first response.
         """
@@ -17,13 +17,12 @@ class TakeFirstFilter(Filter):
 
 
 class TakeKFilter(Filter):
-    def __init__(self, *args, **kwargs):
-
+    def __init__(self, *args, **kwargs) -> None:
         self.k = kwargs.pop("k")
 
         super().__init__(*args, **kwargs)
 
-    def apply(self, resps):
+    def apply(self, resps, docs):
         # check we have at least k responses per doc, else we can't take the first k
         assert (
             len(resps[0]) >= self.k
@@ -32,12 +31,12 @@ class TakeKFilter(Filter):
 
 
 class MajorityVoteFilter(Filter):
-    def __init__(self):
+    def __init__(self) -> None:
         """
         Can define custom behavior here, if an individual instantiation of a Filter class should have state.
         """
 
-    def apply(self, resps):
+    def apply(self, resps, docs):
         """
         Each entry of `resps` is a list of model responses.
         We select the response that occurs most frequently in each entry of `resps`.

lm_eval/logger.py

@@ -6,3 +6,5 @@ logging.basicConfig(
     level=logging.INFO,
 )
 eval_logger = logging.getLogger("lm-eval")
+
+SPACING = " " * 47

lm_eval/models/anthropic_llms.py

@@ -76,7 +76,7 @@ class AnthropicLM(LM):
         max_tokens_to_sample: int = 256,
         temperature: float = 0,  # defaults to 1
         **kwargs,  # top_p, top_k, etc.
-    ):
+    ) -> None:
         """Anthropic API wrapper.
 
         :param model: str
@@ -135,11 +135,10 @@ please install anthropic via `pip install lm-eval[anthropic]` or `pip install -e
     def tok_decode(self, tokens: List[int]) -> str:
         return self.tokenizer.decode(tokens)
 
-    def _loglikelihood_tokens(self, requests, disable_tqdm=False):
+    def _loglikelihood_tokens(self, requests, disable_tqdm: bool = False):
         raise NotImplementedError("No support for logits.")
 
     def greedy_until(self, requests) -> List[str]:
-
         if not requests:
             return []
 

lm_eval/models/dummy.py

@@ -5,7 +5,7 @@ from lm_eval.api.registry import register_model
 
 @register_model("dummy")
 class DummyLM(LM):
-    def __init__(self):
+    def __init__(self) -> None:
         super().__init__()
 
     @classmethod

lm_eval/models/huggingface.py

+import os
+
 import torch
 import transformers
 from transformers.models.auto.modeling_auto import (
@@ -20,7 +22,7 @@ from lm_eval.api.registry import register_model
 
 from lm_eval.utils import MultiTokenEOSCriteria, stop_sequences_criteria
 
-from accelerate import Accelerator, find_executable_batch_size
+from accelerate import Accelerator, find_executable_batch_size, DistributedType
 from typing import List, Optional, Union
 
 
@@ -67,6 +69,7 @@ class HFLM(LM):
         revision: Optional[str] = "main",
         subfolder: Optional[str] = None,
         tokenizer: Optional[str] = None,
+        truncation: Optional[bool] = False,
         max_length: Optional[int] = None,
         device: Optional[str] = "cuda",
         dtype: Optional[Union[str, torch.dtype]] = "auto",
@@ -75,6 +78,7 @@ class HFLM(LM):
         low_cpu_mem_usage: Optional[bool] = True,
         trust_remote_code: Optional[bool] = False,
         use_fast_tokenizer: Optional[bool] = True,
+        cache_dir: Optional[Union[str, os.PathLike]] = None,
         # arguments used for splitting a model across GPUs naively.
         # only used if `parallelize=True`.
         parallelize: Optional[bool] = False,
@@ -90,7 +94,7 @@ class HFLM(LM):
         bnb_4bit_compute_dtype: Optional[Union[str, torch.dtype]] = None,
         gptq: Optional[Union[bool, str]] = False,
         gptq_use_triton: Optional[bool] = False,
-    ):
+    ) -> None:
         super().__init__()
 
         assert isinstance(device, str)
@@ -240,6 +244,8 @@ class HFLM(LM):
             use_fast=use_fast_tokenizer,
         )
 
+        self.truncation = truncation
+
         self.vocab_size = self.tokenizer.vocab_size
         self.tokenizer.pad_token_id = self.tokenizer.eos_token_id
 
@@ -289,9 +295,16 @@ class HFLM(LM):
                         "Failed to place model onto specified device. This may be because the model is quantized via `bitsandbytes`. If the desired GPU is being used, this message is safe to ignore."
                     )
             else:
-                self._model = accelerator.prepare_model(
-                    self.model, evaluation_mode=True
-                )
+                assert accelerator.distributed_type in [
+                    DistributedType.FSDP,
+                    DistributedType.MULTI_GPU,
+                ], "Unsupported distributed type provided. Only DDP and FSDP are supported."
+                if accelerator.distributed_type == DistributedType.FSDP:
+                    self._model = accelerator.prepare(self.model)
+                else:
+                    self._model = accelerator.prepare_model(
+                        self.model, evaluation_mode=True
+                    )
                 self._device = torch.device(f"cuda:{accelerator.local_process_index}")
                 self.accelerator = accelerator
 
@@ -334,7 +347,7 @@ class HFLM(LM):
         return self._DEFAULT_MAX_LENGTH
 
     @property
-    def max_gen_toks(self):
+    def max_gen_toks(self) -> int:
         return 256
 
     @property
@@ -353,7 +366,7 @@ class HFLM(LM):
     def world_size(self):
         return self._world_size
 
-    def _detect_batch_size(self, requests=None, pos=0):
+    def _detect_batch_size(self, requests=None, pos: int = 0):
         if requests:
             _, context_enc, continuation_enc = requests[pos]
             max_length = len(
@@ -419,7 +432,11 @@ class HFLM(LM):
         return encoding
 
     def tok_batch_encode(
-        self, strings: List[str], padding_side="left", left_truncate_len=None
+        self,
+        strings: List[str],
+        padding_side: str = "left",
+        left_truncate_len: int = None,
+        truncation: bool = False,
     ):
         # encode a batch of strings. converts to tensors and pads automatically, unlike tok_encode.
         old_padding_side = self.tokenizer.padding_side
@@ -432,6 +449,7 @@ class HFLM(LM):
 
         encoding = self.tokenizer(
             strings,
+            truncation=truncation,
             padding="longest",
             return_tensors="pt",
             add_special_tokens=add_special_tokens,
@@ -595,7 +613,9 @@ class HFLM(LM):
 
         return loglikelihoods
 
-    def _loglikelihood_tokens(self, requests, disable_tqdm=False, override_bs=None):
+    def _loglikelihood_tokens(
+        self, requests, disable_tqdm: bool = False, override_bs=None
+    ):
         # TODO: implement some kind of efficient-request-middleware that lumps together requests with the same context
         res = []
 
@@ -856,7 +876,9 @@ class HFLM(LM):
 
                 # encode, pad, and truncate contexts for this batch
                 context_enc, attn_masks = self.tok_batch_encode(
-                    contexts, left_truncate_len=max_ctx_len
+                    contexts,
+                    left_truncate_len=max_ctx_len,
+                    truncation=self.truncation,
                 )
                 context_enc = context_enc.to(self.device)
                 attn_masks = attn_masks.to(self.device)

lm_eval/models/openai_completions.py

@@ -69,7 +69,7 @@ class OpenaiCompletionsLM(LM):
         engine: str = "text-davinci-003",
         truncate: bool = False,
         batch_size: int = 1,
-    ):
+    ) -> None:
         """
 
         :param engine: str
@@ -99,12 +99,12 @@ class OpenaiCompletionsLM(LM):
         return self.end_of_text_token_id
 
     @property
-    def max_length(self):
+    def max_length(self) -> int:
         # Note: the OpenAI API supports up to 2049 tokens, with the first token being the first input token
         return 2048
 
     @property
-    def max_gen_toks(self):
+    def max_gen_toks(self) -> int:
         return 256
 
     @property
@@ -152,7 +152,7 @@ class OpenaiCompletionsLM(LM):
         return self._loglikelihood_tokens(new_reqs)
 
     def _loglikelihood_tokens(
-        self, requests, disable_tqdm=False
+        self, requests, disable_tqdm: bool = False
     ) -> List[Tuple[float, bool]]:
         res = []
 

lm_eval/models/textsynth.py

@@ -41,7 +41,7 @@ def textsynth_completion(**kwargs):
 
 @register_model("textsynth")
 class TextSynthLM(LM):
-    def __init__(self, engine, truncate=False):
+    def __init__(self, engine, truncate: bool = False) -> None:
         """
         :param engine: str
             TextSynth API engine (e.g. `gptj_6B`)
@@ -62,12 +62,12 @@ class TextSynthLM(LM):
         raise NotImplementedError()
 
     @property
-    def max_length(self):
+    def max_length(self) -> int:
         # NOTE: Turn on truncation to avoid errors on long inputs.
         return 2048
 
     @property
-    def max_gen_toks(self):
+    def max_gen_toks(self) -> int:
         return 256
 
     @property

lm_eval/prompts/__init__.py

@@ -5,7 +5,7 @@ from lm_eval.logger import eval_logger
 # Stores prompts in a dictionary indexed by 2 levels:
 # prompt category name, and prompt name.
 # This allows us to access prompts
-PROMPT_REGISTRY = {
+PROMPT_REGISTRY: dict[str, dict[str, str]] = {
     "qa-basic": {
         "question-newline-answer": "Question: {{question}}\nAnswer:",
         "q-newline-a": "Q: {{question}}\nA:",
@@ -13,7 +13,7 @@ PROMPT_REGISTRY = {
 }
 
 
-def get_prompt(prompt_id: str, dataset_name=None, subset_name=None):
+def get_prompt(prompt_id: str, dataset_name: str = None, subset_name: str = None):
     # unpack prompt name
     category_name, prompt_name = prompt_id.split(":")
     if subset_name is None:

lm_eval/tasks/README.md

@@ -5,8 +5,8 @@ Boxes should be checked iff tasks are implemented in the refactor and tested for
 
 - [x] Glue
 - [x] SuperGlue
-- [ ] CoQA (Lintang)
-- [ ] DROP (Lintang)
+- [x] CoQA
+- [x] DROP
 - [x] ~~Lambada~~
 - [x] Lambada (Cloze variants)
 - [x] ~~Lambada (Multilingual)~~
@@ -29,7 +29,7 @@ Boxes should be checked iff tasks are implemented in the refactor and tested for
 - [x] HeadQA
 - [x] MathQA
 - [x] WebQs
-- [ ] WSC273 (Lintang)
+- [x] WSC273
 - [x] Winogrande
 - [x] ANLI
 - [x] Hendrycks Ethics (missing some tasks/metrics, see PR 660: <https://github.com/EleutherAI/lm-evaluation-harness/pull/660> for more info)
@@ -38,11 +38,11 @@ Boxes should be checked iff tasks are implemented in the refactor and tested for
 - [x] TruthfulQA (gen)
 - [ ] MuTual
 - [ ] Hendrycks Math (Hailey)
-- [ ] Asdiv
+- [x] Asdiv
 - [ ] GSM8k
 - [x] Arithmetic
 - [ ] MMMLU (Hailey)
-- [ ] Translation (WMT) suite (Hailey)
+- [x] Translation (WMT) suite
 - [x] Unscramble
 - [x] ~~Pile (perplexity)~~
 - [x] BLiMP
@@ -56,7 +56,7 @@ Boxes should be checked iff tasks are implemented in the refactor and tested for
 - [x] XWinograd
 - [x] PAWS-X
 - [x] XNLI
-- [ ] MGSM (Lintang)
+- [x] MGSM
 - [ ] SCROLLS
 - [x] Babi
 

lm_eval/tasks/__init__.py

@@ -15,7 +15,7 @@ from lm_eval.api.registry import (
 )
 
 
-def register_configurable_task(config):
+def register_configurable_task(config: dict[str, str]) -> int:
     SubClass = type(
         config["task"] + "ConfigurableTask",
         (ConfigurableTask,),
@@ -38,7 +38,7 @@ def register_configurable_task(config):
     return 0
 
 
-def check_prompt_config(config):
+def check_prompt_config(config: dict[str, str]) -> List[dict[str, str]]:
     all_configs = []
     if "use_prompt" in config:
         prompt_list = prompts.load_prompt_list(
@@ -69,14 +69,14 @@ def check_prompt_config(config):
     return all_configs
 
 
-def get_task_name_from_config(task_config):
+def get_task_name_from_config(task_config: dict[str, str]) -> str:
     if "dataset_name" in task_config:
         return "{dataset_path}_{dataset_name}".format(**task_config)
     else:
         return "{dataset_path}".format(**task_config)
 
 
-def include_task_folder(task_dir):
+def include_task_folder(task_dir: str) -> None:
     """
     Calling this function
     """

lm_eval/tasks/asdiv/default.yaml

+task: asdiv
+dataset_path: EleutherAI/asdiv
+output_type: loglikelihood
+validation_split: validation
+doc_to_text: "{{body}}\nQuestion:{{question}}\nAnswer:"
+doc_to_target: "{{answer.split(' (')[0]}}"
+should_decontaminate: true
+doc_to_decontamination_query: "{{body}} {{question}}"
+metric_list:
+  - metric: acc
+    aggregation: mean
+    higher_is_better: true

lm_eval/tasks/coqa/README.md

+# CoQA
+
+### Paper
+
+Title: `CoQA: A Conversational Question Answering Challenge`
+
+Abstract: https://arxiv.org/pdf/1808.07042.pdf
+
+CoQA is a large-scale dataset for building Conversational Question Answering
+systems. The goal of the CoQA challenge is to measure the ability of machines to
+understand a text passage and answer a series of interconnected questions that
+appear in a conversation.
+
+Homepage: https://stanfordnlp.github.io/coqa/
+
+### Citation
+
+```
+BibTeX-formatted citation goes here
+```
+
+### Groups and Tasks
+
+#### Groups
+
+* Not part of a group yet
+
+#### Tasks
+
+* `coqa`
+
+### Checklist
+
+For adding novel benchmarks/datasets to the library:
+* [ ] Is the task an existing benchmark in the literature?
+  * [ ] Have you referenced the original paper that introduced the task?
+  * [ ] If yes, does the original paper provide a reference implementation? If so, have you checked against the reference implementation and documented how to run such a test?
+
+
+If other tasks on this dataset are already supported:
+* [ ] Is the "Main" variant of this task clearly denoted?
+* [ ] Have you provided a short sentence in a README on what each new variant adds / evaluates?
+* [ ] Have you noted which, if any, published evaluation setups are matched by this variant?

lm_eval/tasks/coqa/default.yaml

+task: coqa
+dataset_path: EleutherAI/coqa
+output_type: greedy_until
+training_split: train
+validation_split: validation
+doc_to_text: !function utils.doc_to_text
+doc_to_target: !function utils.doc_to_target
+process_results: !function utils.process_results
+should_decontaminate: true
+doc_to_decontamination_query: "{{story}} {{question.input_text|join('\n')}}"
+generation_kwargs:
+  until:
+    - "\nQ:"
+metric_list:
+  - metric: em
+    aggregation: mean
+    higher_is_better: true
+  - metric: f1
+    aggregation: mean
+    higher_is_better: true

lm_eval/tasks/coqa/utils.py

+from itertools import zip_longest
+
+import transformers.data.metrics.squad_metrics as squad_metrics
+
+
+def doc_to_text(doc):
+    # Given a passage p, the conversation history {q1, a1, . . . qi−1, ai−1}
+    # and a question qi, the task is to predict the answer ai
+    doc_text = doc["story"] + "\n\n"
+    for (q, a) in zip_longest(
+        doc["questions"]["input_text"], doc["answers"]["input_text"][:-1]
+    ):  # omit target answer ai
+        question = f"Q: {q}\n\n"
+        answer = f"A: {a}\n\n" if a is not None else "A:"
+        doc_text += question + answer
+    return doc_text
+
+
+def doc_to_target(doc):
+
+    turn_id = len(doc["questions"]["input_text"])
+    # Returns unique answers and valid alternatives (Some questions in CoQA have multiple valid answers).
+    answers = []
+    answer_forturn = doc["answers"]["input_text"][turn_id - 1]
+    answers.append(answer_forturn)
+
+    additional_answers = doc.get("additional_answers")
+    if additional_answers:
+        for key in additional_answers:
+            additional_answer_for_turn = additional_answers[key]["input_text"][
+                turn_id - 1
+            ]
+            if additional_answer_for_turn.lower() not in map(str.lower, answers):
+                answers.append(additional_answer_for_turn)
+    return answers
+
+
+def em(gold_list, pred):
+    # tests for exact match and on the normalised answer (compute_exact)
+    em_sum = 0.0
+    if len(gold_list) > 1:
+        for i in range(len(gold_list)):
+            gold_answers = gold_list[0:i] + gold_list[i + 1 :]
+            # predictions compared against (n) golds and take maximum
+            em_sum += max(squad_metrics.compute_exact(a, pred) for a in gold_answers)
+    else:
+        em_sum += max(squad_metrics.compute_exact(a, pred) for a in gold_list)
+
+    return em_sum / max(1, len(gold_list))
+
+
+def compute_scores(gold_list, pred):
+    # tests for exact match and on the normalised answer (compute_exact)
+    # test for overlap (compute_f1)
+    f1_sum = 0.0
+    em_sum = 0.0
+    if len(gold_list) > 1:
+        for i in range(len(gold_list)):
+            gold_answers = gold_list[0:i] + gold_list[i + 1 :]
+            # predictions compared against (n) golds and take maximum
+            em_sum += max(squad_metrics.compute_exact(a, pred) for a in gold_answers)
+            f1_sum += max(squad_metrics.compute_f1(a, pred) for a in gold_answers)
+    else:
+        em_sum += max(squad_metrics.compute_exact(a, pred) for a in gold_list)
+        f1_sum += max(squad_metrics.compute_f1(a, pred) for a in gold_list)
+
+    return {
+        "em": em_sum / max(1, len(gold_list)),
+        "f1": f1_sum / max(1, len(gold_list)),
+    }
+
+
+def process_results(doc, results):
+
+    gold_list = doc_to_target(doc)
+    pred = results[0].strip().split("\n")[0]
+
+    scores = compute_scores(gold_list, pred)
+    return scores

lm_eval/tasks/drop/README.md

+# DROP
+
+### Paper
+
+Title: `DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs`
+
+Abstract: https://aclanthology.org/attachments/N19-1246.Supplementary.pdf
+
+DROP is a QA dataset which tests comprehensive understanding of paragraphs. In
+this crowdsourced, adversarially-created, 96k question-answering benchmark, a
+system must resolve multiple references in a question, map them onto a paragraph,
+and perform discrete operations over them (such as addition, counting, or sorting).
+
+Homepage: https://allenai.org/data/drop
+
+Acknowledgement: This implementation is based on the official evaluation for `DROP`:
+https://github.com/allenai/allennlp-reading-comprehension/blob/master/allennlp_rc/eval/drop_eval.py
+
+### Citation
+
+```
+@misc{dua2019drop,
+    title={DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs},
+    author={Dheeru Dua and Yizhong Wang and Pradeep Dasigi and Gabriel Stanovsky and Sameer Singh and Matt Gardner},
+    year={2019},
+    eprint={1903.00161},
+    archivePrefix={arXiv},
+    primaryClass={cs.CL}
+}
+```
+
+### Groups and Tasks
+
+#### Groups
+
+* Not part of a group yet.
+
+#### Tasks
+
+* `drop`
+
+### Checklist
+
+For adding novel benchmarks/datasets to the library:
+* [ ] Is the task an existing benchmark in the literature?
+  * [ ] Have you referenced the original paper that introduced the task?
+  * [ ] If yes, does the original paper provide a reference implementation? If so, have you checked against the reference implementation and documented how to run such a test?
+
+
+If other tasks on this dataset are already supported:
+* [ ] Is the "Main" variant of this task clearly denoted?
+* [ ] Have you provided a short sentence in a README on what each new variant adds / evaluates?
+* [ ] Have you noted which, if any, published evaluation setups are matched by this variant?