Merge pull request #612 from EleutherAI/benchmark-scripts

[Refactor] Benchmark scripts

.gitignore

 env
 *.pyc
+output/
 data/
 lm_cache
 .idea

lm_eval/api/registry.py

 import os
 import evaluate
 from lm_eval.api.model import LM
+from lm_eval.logger import eval_logger
 
 MODEL_REGISTRY = {}
 
@@ -130,7 +131,7 @@ searching in HF Evaluate library..."
             metric_object = evaluate.load(name)
             return metric_object.compute
         except Exception:
-            raise Warning(
+            eval_logger.error(
                 "{} not found in the evaluate library!".format(name),
                 "Please check https://huggingface.co/evaluate-metric",
             )
@@ -153,7 +154,7 @@ def get_aggregation(name):
     try:
         return AGGREGATION_REGISTRY[name]
     except KeyError:
-        raise Warning(
+        eval_logger.warning(
             "{} not a registered aggregation metric!".format(name),
         )
 
@@ -162,7 +163,9 @@ def get_default_aggregation(metric_name):
     try:
         return DEFAULT_AGGREGATION_REGISTRY[metric_name]
     except KeyError:
-        raise Warning(f"No default aggregation metric for metric '{metric_name}'!")
+        eval_logger.warning(
+            f"No default aggregation metric for metric '{metric_name}'!"
+        )
 
 
 def is_higher_better(metric_name):
@@ -170,3 +173,6 @@ def is_higher_better(metric_name):
         return HIGHER_IS_BETTER_REGISTRY[metric_name]
     except KeyError:
         raise Warning(f"higher_is_better not specified for metric '{metric_name}'!")
+        eval_logger.warning(
+            f"higher_is_better not specified for metric '{metric_name}'!"
+        )

lm_eval/api/task.py

@@ -70,6 +70,7 @@ class TaskConfig(dict):
     doc_to_target: Union[Callable, str] = None
     doc_to_choice: Union[Callable, str, dict, list] = None
     gold_alias: Union[Callable, str] = None
+    process_results: Union[Callable, str] = None
     use_prompt: str = None
     description: str = ""
     target_delimiter: str = " "
@@ -545,6 +546,16 @@ class ConfigurableTask(Task):
                     for key in metric_config
                     if key not in ["metric", "aggregation", "higher_is_better"]
                 }
+
+                if self._config.process_results is not None:
+                    self._metric_fn_list[metric_name] = None
+                    self._metric_fn_kwargs[metric_name] = {}
+                elif callable(metric_name):
+                    metric_fn = metric_name.__call__
+                    metric_name = metric_name.__name__
+                    self._metric_fn_list[metric_name] = metric_fn
+                    self._metric_fn_kwargs[metric_name] = kwargs
+                else:
                     self._metric_fn_list[metric_name] = get_metric(metric_name)
                     self._metric_fn_kwargs[metric_name] = kwargs
 
@@ -885,8 +896,8 @@ class ConfigurableTask(Task):
 
     def process_results(self, doc, results):
 
-        # if callable(self._config.process_results):
-        #     return self._config.process_results(doc, results)
+        if callable(self._config.process_results):
+            return self._config.process_results(doc, results)
 
         result_dict = {}
         use_metric = list(self._metric_fn_list.keys())
@@ -975,7 +986,11 @@ class ConfigurableTask(Task):
         elif self.OUTPUT_TYPE == "greedy_until":
 
             gold = self.doc_to_target(doc)
+            if type(gold) == int:
+                choices = self.doc_to_choice(doc)
+                gold = choices[gold]
 
+            print(self._metric_fn_list)
             for key, result in zip(self._metric_fn_list.keys(), results):
                 if self.multiple_target:
                     # in the case where we have multiple targets,

lm_eval/evaluator.py

@@ -192,14 +192,35 @@ def evaluate(
 
     # decontaminate = decontamination_ngrams_path is not None
 
+    # stores the final result for each task, for each metric/filter pair.
     results = collections.defaultdict(dict)
+    # Tracks each task's version.
     versions = collections.defaultdict(dict)
+    # Tracks the YAML configs of all chosen tasks.
     configs = collections.defaultdict(dict)
+    # logs info about each document evaluated.
     samples = collections.defaultdict(list)
+    # tracks all Instances/requests a model must generate output on.
     requests = collections.defaultdict(list)
+    # Stores task scores based on task grouping.
+    aggregate = collections.defaultdict(dict)
+    # tracks if a task was chosen via user selecting a group containing it
+    task_groups = collections.defaultdict(dict)
+    # stores the amount to pad out reqs per req. type so that
+    # number of fwd passes per distributed rank is equal
+    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
     for task_name, task in task_dict.items():
+
+        if type(task) == tuple:
+            group, task = task
+            task_groups[task_name] = group
+
         versions[task_name] = task.VERSION
         configs[task_name] = dict(task.dump_config())
 
@@ -243,6 +264,7 @@ def evaluate(
 
             # compute number of pseudobatches to pad with (FSDP/DDP require even batches among ranks)
             numpad = max(gathered_item) - gathered_item[lm.rank]
+            padding_requests[task.OUTPUT_TYPE] += numpad
 
     ### Run LM on inputs, get all outputs ###
     # execute each type of request
@@ -253,8 +275,8 @@ def evaluate(
         for req in reqs:
             cloned_reqs.extend([req] * req.repeats)
 
-        if (lm.world_size > 1) and (numpad > 0):
-            for _ in range(numpad):
+        if (lm.world_size > 1) and (padding_requests[reqtype] > 0):
+            for _ in range(padding_requests[reqtype]):
                 cloned_reqs.extend([req] * req.repeats)
 
         # run requests through model
@@ -270,6 +292,8 @@ def evaluate(
     ### Postprocess outputs ###
     # TODO: del model here, maybe (idea: allow user to specify device of e.g. reward model separately)
     for task_name, task in task_dict.items():
+        if type(task) == tuple:
+            group, task = task
         task.apply_filters()
 
     ### Collect values of metrics on all datapoints ###
@@ -277,6 +301,8 @@ def evaluate(
 
     # unpack results and sort back in order and return control to Task
     for task_name, task in task_dict.items():
+        if type(task) == tuple:
+            group, task = task
         # TODO: make it possible to use a different metric per filter
         # iterate over different filters used
         for key in task.instances[0].filtered_resps.keys():
@@ -362,7 +388,23 @@ def evaluate(
         # aggregate results ; run bootstrap CIs
         for (task_name, key, metric), items in vals.items():
             task = task_dict[task_name]
-            results[task_name][metric + "," + key] = task.aggregation()[metric](items)
+            if type(task) == tuple:
+                group, task = task
+            task_score = task.aggregation()[metric](items)
+            results[task_name][metric + "," + key] = task_score
+
+            # Need to put back in results
+            # pythia | acc
+            #        | perplexity
+            #        | word_perplexity
+            #        | byte_perplexity
+            #        | bits_per_byte
+            if bool(task_groups):
+                group_name = task_groups[task_name]
+                if metric not in aggregate[group_name]:
+                    aggregate[group_name][metric] = [task_score]
+                else:
+                    aggregate[group_name][metric].append(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
@@ -377,10 +419,21 @@ def evaluate(
                 if stderr is not None:
                     results[task_name][metric + "_stderr" + "," + key] = stderr(items)
 
+        if bool(aggregate):
+            for group in aggregate.keys():
+                for metric in aggregate[group].keys():
+                    aggregate[group][metric] = np.average(aggregate[group][metric])
+                    versions[group] = "N/A"
+
         results_dict = {
-            "results": dict(results),
-            "configs": dict(configs),
-            "versions": dict(versions),
+            "results": dict(sorted(results.items())),
+            **(
+                {"aggregate": dict(sorted(aggregate.items()))}
+                if bool(aggregate)
+                else {}
+            ),
+            "configs": dict(sorted(configs.items())),
+            "versions": dict(sorted(versions.items())),
         }
         if log_samples:
             results_dict["samples"] = dict(samples)

lm_eval/prompts/__init__.py

+from lm_eval import utils
 from lm_eval.logger import eval_logger
 
 # Prompt library.
@@ -51,3 +52,17 @@ def get_prompt(prompt_id: str, dataset_name=None, subset_name=None):
                 f"expected only a single `:` as separator between \
                 prompt category and name, but got `{prompt_id}` instead"
             )
+
+
+def load_prompt_list(use_prompt: str, dataset_name=None, subset_name=None, **kwargs):
+
+    from promptsource.templates import DatasetTemplates
+
+    if subset_name is None:
+        prompts = DatasetTemplates(dataset_name=dataset_name)
+    else:
+        prompts = DatasetTemplates(dataset_name=dataset_name, subset_name=subset_name)
+
+    category_name, prompt_name = use_prompt.split(":")
+    prompt_list = utils.pattern_match(prompt_name, prompts.all_template_names)
+    return [":".join([category_name, prompt]) for prompt in prompt_list]

lm_eval/tasks/__init__.py

 import os
+import yaml
 from typing import List, Union
 
 from lm_eval import utils
+from lm_eval import prompts
 from lm_eval.logger import eval_logger
 from lm_eval.api.task import TaskConfig, Task, ConfigurableTask
 from lm_eval.api.registry import (
@@ -13,6 +15,58 @@ from lm_eval.api.registry import (
 )
 
 
+def register_configurable_task(config):
+    SubClass = type(
+        config["task"] + "ConfigurableTask",
+        (ConfigurableTask,),
+        {"CONFIG": TaskConfig(**config)},
+    )
+
+    if "task" in config:
+        task_name = "{}".format(config["task"])
+        register_task(task_name)(SubClass)
+
+    if "group" in config:
+        if type(config["group"]) == str:
+            group_name = [config["group"]]
+        else:
+            group_name = config["group"]
+
+        for group in group_name:
+            register_group(group)(SubClass)
+
+    return 0
+
+
+def check_prompt_config(config):
+    all_configs = []
+    if "use_prompt" in config:
+        prompt_list = prompts.load_prompt_list(
+            use_prompt=config["use_prompt"],
+            dataset_name=config["dataset_path"],
+            subset_name=config["dataset_name"],
+        )
+        for idx, prompt_variation in enumerate(prompt_list):
+            all_configs.append(
+                {
+                    **config,
+                    **{"use_prompt": prompt_variation},
+                    **{
+                        "task": "_".join(
+                            [
+                                get_task_name_from_config(config),
+                                prompt_variation,
+                            ]
+                        )
+                    },
+                    **{"output_type": "greedy_until"},
+                }
+            )
+    else:
+        all_configs.append(config)
+    return all_configs
+
+
 def get_task_name_from_config(task_config):
     if "dataset_name" in task_config:
         return "{dataset_path}_{dataset_name}".format(**task_config)
@@ -31,23 +85,10 @@ def include_task_folder(task_dir):
                     yaml_path = os.path.join(root, f)
                     try:
                         config = utils.load_yaml_config(yaml_path)
+                        all_configs = check_prompt_config(config)
+                        for config in all_configs:
+                            register_configurable_task(config)
 
-                        SubClass = type(
-                            config["task"] + "ConfigurableTask",
-                            (ConfigurableTask,),
-                            {"CONFIG": TaskConfig(**config)},
-                        )
-
-                        if "task" in config:
-                            # task_name = "{}:{}".format(
-                            #     get_task_name_from_config(config), config["task"]
-                            # )
-                            task_name = "{}".format(config["task"])
-                            register_task(task_name)(SubClass)
-
-                        if "group" in config:
-                            for group in config["group"]:
-                                register_group(group)(SubClass)
                     except Exception as error:
                         eval_logger.warning(
                             "Failed to load config in\n"
@@ -57,8 +98,58 @@ def include_task_folder(task_dir):
                         )
 
 
+def include_benchmarks(task_dir, benchmark_dir="benchmarks"):
+
+    for root, subdirs, file_list in os.walk(os.path.join(task_dir, benchmark_dir)):
+        if (subdirs == [] or subdirs == ["__pycache__"]) and (len(file_list) > 0):
+            for f in file_list:
+                if f.endswith(".yaml"):
+                    try:
+                        benchmark_path = os.path.join(root, f)
+
+                        with open(benchmark_path, "rb") as file:
+                            yaml_config = yaml.full_load(file)
+
+                        assert "group" in yaml_config
+                        group = yaml_config["group"]
+                        all_task_list = yaml_config["task"]
+                        config_list = [
+                            task for task in all_task_list if type(task) != str
+                        ]
+                        task_list = [
+                            task for task in all_task_list if type(task) == str
+                        ]
+
+                        for task_config in config_list:
+                            var_configs = check_prompt_config(
+                                {
+                                    **task_config,
+                                    **{"group": group},
+                                }
+                            )
+                            for config in var_configs:
+                                register_configurable_task(config)
+
+                        task_names = utils.pattern_match(task_list, ALL_TASKS)
+                        for task in task_names:
+                            if task in TASK_REGISTRY:
+                                if group in GROUP_REGISTRY:
+                                    GROUP_REGISTRY[group].append(task)
+                                else:
+                                    GROUP_REGISTRY[group] = [task]
+                                    ALL_TASKS.add(group)
+                    except Exception as error:
+                        eval_logger.warning(
+                            "Failed to load benchmark in\n"
+                            f"                                 {benchmark_path}\n"
+                            "                                 Benchmark will not be added to registry\n"
+                            f"                                 Error: {error}"
+                        )
+
+
 task_dir = os.path.dirname(os.path.abspath(__file__)) + "/"
 include_task_folder(task_dir)
+include_benchmarks(task_dir)
 
 
 def get_task(task_name, config):
@@ -97,11 +188,15 @@ def get_task_dict(task_name_list: List[Union[str, dict, Task]], **kwargs):
         if isinstance(task_element, str):
 
             if task_element in GROUP_REGISTRY:
+                group_name = task_element
                 for task_name in GROUP_REGISTRY[task_element]:
                     if task_name not in task_name_from_registry_dict:
                         task_name_from_registry_dict = {
                             **task_name_from_registry_dict,
-                            task_name: get_task(task_name=task_name, config=config),
+                            task_name: (
+                                group_name,
+                                get_task(task_name=task_name, config=config),
+                            ),
                         }
             else:
                 task_name = task_element

lm_eval/tasks/benchmarks/pythia.yaml

+group: pythia
+task:
+  - lambada_openai
+  - wikitext
+  - piqa
+  - sciq
+  - wsc
+  - winogrande
+  - arc_*
+  # - logiqa
+  # - blimp_*
+  # - hendrycksTest*

lm_eval/tasks/benchmarks/t0_eval.yaml

+group: t0_eval
+task:
+  # # Coreference Resolution
+  # - dataset_path: super_glue
+  #   dataset_name: wsc.fixed
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true
+  # # Coreference Resolution
+  # - dataset_path: winogrande
+  #   dataset_name: winogrande_xl
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true
+  # Natural Language Inference
+  - dataset_path: super_glue
+    dataset_name: cb
+    use_prompt: promptsource:*
+    training_split: train
+    validation_split: validation
+    output_type: greedy_until
+    metric_list:
+      - metric: exact_match
+        aggregation: mean
+        higher_is_better: true
+        ignore_case: true
+        ignore_punctuation: true
+  # Natural Language Inference
+  # - dataset_path: super_glue
+  #   dataset_name: rte
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true
+  # # Natural Language Inference
+  # # - dataset_path: anli
+  # #   use_prompt: promptsource:*
+  # #   training_split: train_r1
+  # #   validation_split: dev_r1
+  # # Sentence Completion
+  # - dataset_path: super_glue
+  #   dataset_name: copa
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true
+  # # Natural Language Inference
+  # - dataset_path: hellaswag
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true
+  # # Word Sense Disambiguation
+  # - dataset_path: super_glue
+  #   dataset_name: wic
+  #   use_prompt: promptsource:*
+  #   training_split: train
+  #   validation_split: validation
+  #   metric_list:
+  #     - metric: exact_match
+  #       aggregation: mean
+  #       higher_is_better: true
+  #       ignore_case: true
+  #       ignore_punctuation: true

lm_eval/tasks/super_glue/boolq/default.yaml

 group:
   - super-glue-lm-eval-v1
-task: "boolq"
+task: boolq
 dataset_path: super_glue
 dataset_name: boolq
 output_type: multiple_choice

lm_eval/tasks/super_glue/cb/t5-prompt.yaml

@@ -5,11 +5,15 @@ dataset_path: super_glue
 dataset_name: cb
 training_split: train
 validation_split: validation
+output_type: greedy_until
 doc_to_text: "cb hypothesis: {{hypothesis}} premise {{premise}}"
-doc_to_target: "{% set answer_choices = ['entailment', 'contradiction', 'neutral'] %}{{answer_choices[label]}}"
+doc_to_target: label
+doc_to_choice: ['entailment', 'contradiction', 'neutral']
 metric_list:
   - metric: exact_match
     aggregation: mean
     higher_is_better: true
     ignore_case: true
     ignore_punctuation: true
+  - metric: f1
+    aggregation: !function "aggregate.cb_multi_fi"

lm_eval/tasks/super_glue/copa/t5-prompt.yaml

@@ -5,8 +5,10 @@ dataset_path: super_glue
 dataset_name: copa
 training_split: train
 validation_split: validation
+output_type: greedy_until
 doc_to_text: "copa choice1: {{choice1}} choice2: {{choice2}} question: {{question}}"
-doc_to_target: "{% set answer_choices = ['False', 'True'] %}{{answer_choices[label]}}"
+doc_to_target: label
+doc_to_choice: ['False', 'True']
 metric_list:
   - metric: exact_match
     aggregation: mean

lm_eval/tasks/super_glue/multirc/promptsource-00.yaml

-group:
-  - super-glue-promptsource
-task: "I was going to say…"
-dataset_path: super_glue
-dataset_name: multirc
-training_split: train
-validation_split: validation
-use_prompt: "promptsource:I was going to say…"
-metric_list:
-  - metric: exact_match
-    aggregation: mean
-    higher_is_better: true
-    ignore_case: true
-    ignore_punctuation: true

lm_eval/tasks/super_glue/multirc/promptsource-01.yaml

-include: promptsource-00.yaml
-group:
-  - super-glue-promptsource
-task: "Would it be good to answer…"
-use_prompt: "promptsource:Would it be good to answer…"

lm_eval/tasks/super_glue/multirc/promptsource-02.yaml

-include: promptsource-00.yaml
-group:
-  - super-glue-promptsource
-task: "confirm"
-use_prompt: "promptsource:confirm"

lm_eval/tasks/super_glue/record/default.yaml

-# group:
-#   - super-glue-lm-eval-v1
+group:
+  - super-glue-lm-eval-v1
 task: record
 dataset_path: super_glue
 dataset_name: record
@@ -9,6 +9,10 @@ validation_split: validation
 doc_to_text: !function util.doc_to_text
 doc_to_target: "{{answers}}"
 doc_to_choice: "{{entities}}"
+process_results: !function util.process_results
 metric_list:
   - metric: f1
+    aggregation: mean
   - metric: em
+    higher_is_better: True
+    aggregation: mean

lm_eval/tasks/super_glue/record/promptsource-00.yaml

-group:
-  - super-glue-promptsource
-task: "Add sentence after (continuation choices)"
-dataset_path: super_glue
-dataset_name: record
-training_split: train
-validation_split: validation
-use_prompt: "promptsource:Add sentence after (continuation choices)"
-metric_list:
-  - metric: exact_match
-    aggregation: mean
-    higher_is_better: true
-    ignore_case: true
-    ignore_punctuation: true

lm_eval/tasks/super_glue/record/t5-prompt.yaml

@@ -5,6 +5,7 @@ dataset_path: super_glue
 dataset_name: record
 training_split: train
 validation_split: validation
+output_type: greedy_until
 doc_to_text: "record query: {{query}} entities: {{entities}} passage: {{passage}}"
 doc_to_target: "{{answers}}"
 metric_list:

lm_eval/tasks/super_glue/record/util.py

+import numpy as np
+import transformers.data.metrics.squad_metrics as squad_metrics
+
+from lm_eval.api.metrics import metric_max_over_ground_truths
+
+
 def doc_to_text(doc):
     initial_text, *highlights = doc["passage"].strip().split("\n@highlight\n")
     text = initial_text + "\n\n"
@@ -13,3 +19,25 @@ def format_answer(query, entity):
 def doc_to_target(doc):
     # We only output the first correct entity in a doc
     return format_answer(query=doc["query"], entity=doc["answers"][0])
+
+
+def process_results(doc, results):
+    # ReCoRD's evaluation is actually deceptively simple:
+    # - Pick the maximum likelihood prediction entity
+    # - Evaluate the accuracy and token F1 PER EXAMPLE
+    # - Average over all examples
+    max_idx = np.argmax(np.array([result[0] for result in results]))
+
+    prediction = doc["entities"][max_idx]
+    gold_label_set = doc["answers"]
+    f1 = metric_max_over_ground_truths(
+        squad_metrics.compute_f1, prediction, gold_label_set
+    )
+    em = metric_max_over_ground_truths(
+        squad_metrics.compute_exact, prediction, gold_label_set
+    )
+
+    return {
+        "f1": f1,
+        "em": em,
+    }

lm_eval/tasks/super_glue/rte/promptsource-00.yaml → lm_eval/tasks/super_glue/rte/default.yaml

 group:
-  - super-glue-promptsource
-task: "rte"
+  - super-glue-lm-eval-v1
+task: rte
 dataset_path: super_glue
 dataset_name: rte
+output_type: multiple_choice
 training_split: train
 validation_split: validation
-use_prompt: "promptsource:GPT-3 style"
-generation_kwargs:
-    until:
-    - "\n"
-    - "\n\n"
+doc_to_text: "{{premise}}\nQuestion: {{hypothesis}} True or False?\nAnswer:"
+doc_to_target: label
+doc_to_choice: ['True', 'False']
 metric_list:
-  - metric: exact_match
-    aggregation: mean
-    higher_is_better: true
-    ignore_case: true
-    ignore_punctuation: true
+  - metric: acc

lm_eval/tasks/super_glue/rte/promptsource-01.yaml

-include: promptsource-00.yaml
-group:
-  - super-glue-promptsource
-task: "MNLI crowdsource"
-use_prompt: "promptsource:MNLI crowdsource"