Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Gather data in a single file #33

Merged
merged 1 commit into from
Oct 10, 2024
Merged

Gather data in a single file #33

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
92 changes: 5 additions & 87 deletions src/benchmark_lfe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,33 +5,11 @@
)
from transformers import AutoTokenizer

models = [
"NousResearch/Nous-Hermes-llama-2-7b", # 32,000 tokens vocabulary
"gpt2", # 50,257 tokens vocabulary
"NousResearch/Hermes-3-Llama-3.1-8B", # 128,256 tokens vocabulary
"unsloth/gemma-2-2b-it-bnb-4bit", # 256,128 tokens vocabulary
]

regex_case = [
(r"\d{3}-\d{2}-\d{4}", "203-22-1234"),
(
r"(https?:\/\/)?([\da-z\.-]+)\.([a-z\.]{2,6})([\/\w \.-]*)*\/?",
"https://github.com/outlines-dev/outlines",
),
(
r"A: [\w \.\*\-=\+,\?/]{10,50}\. The answer is [1-9][0-9]{0,9}\.",
"A: Some thoughts before answering. The answer is 42.",
),
(
"(0|[1-9][0-9]*)|true|false|([a-zA-Z_][a-zA-Z_0-9]*)",
"AVeryLongStringtoTest1234",
),
(r"\+[1-9]\d{1,14}", "1234567891234"),
]
from .data import json_cases, models, regex_cases


class LMFormatEnforcerRegex:
params = [models, regex_case]
params = [models, regex_cases]
param_names = ["model", "regex"]
timeout = 600

Expand All @@ -48,7 +26,7 @@ def setup(self, model, _):
self.tokenizer_data = build_token_enforcer_tokenizer_data(self.tokenizer)

def time_lfe(self, _, regex):
regex_string, regex_example = regex
regex_string, regex_example = regex["regex"], regex["example"]
regex_example_tokens = self.tokenizer.encode(regex_example)

parser = RegexParser(regex_string)
Expand All @@ -58,68 +36,8 @@ def time_lfe(self, _, regex):
_ = token_enforcer.get_allowed_tokens(regex_example_tokens[: i + 1])


json_case = [
(
{
"$defs": {
"Armor": {
"enum": ["leather", "chainmail", "plate"],
"title": "Armor",
"type": "string",
}
},
"properties": {
"name": {"maxLength": 10, "title": "Name", "type": "string"},
"age": {"title": "Age", "type": "integer"},
"armor": {"$ref": "#/$defs/Armor"},
"strength": {"title": "Strength", "type": "integer"},
},
"required": ["name", "age", "armor", "strength"],
"title": "Character",
"type": "object",
},
"""{'name': 'Super Warrior', 'age': 26, 'armor': 'leather', 'armor': 10}""",
),
(
{
"$schema": "http://json-schema.org/draft-04/schema#",
"title": "Schema for a recording",
"type": "object",
"definitions": {
"artist": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"functions": {"type": "array", "items": {"type": "string"}},
},
"required": ["id", "name", "functions"],
}
},
"properties": {
"id": {"type": "number"},
"work": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"composer": {"$ref": "#/definitions/artist"},
},
},
"recording_artists": {
"type": "array",
"items": {"$ref": "#/definitions/artist"},
},
},
"required": ["id", "work", "recording_artists"],
},
"""{'id': 999, 'work': {'id': 1, 'name': 'Strasbourg Saint-Denis', 'composer': 'Roy Hargrove'}, 'recording_artists': [{'id': 2, 'name': 'Roy Hargrove', 'functions': ['Trumpet', 'Singing']}]}""",
),
]


class LMFormatEnforcerJsonSchema:
params = [models, json_case]
params = [models, json_cases]
param_names = ["model", "json"]
timeout = 600

Expand All @@ -136,7 +54,7 @@ def setup(self, model, _):
self.tokenizer_data = build_token_enforcer_tokenizer_data(self.tokenizer)

def time_lfe(self, _, json):
json_string, json_example = json
json_string, json_example = json["schema"], json["example"]
json_example_tokens = self.tokenizer.encode(json_example)

parser = JsonSchemaParser(json_string)
Expand Down
92 changes: 5 additions & 87 deletions src/benchmark_outlines.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,33 +7,11 @@
from outlines.models.transformers import TransformerTokenizer
from transformers import AutoTokenizer

models = [
"NousResearch/Nous-Hermes-llama-2-7b", # 32,000 tokens vocabulary
"gpt2", # 50,257 tokens vocabulary
"NousResearch/Hermes-3-Llama-3.1-8B", # 128,256 tokens vocabulary
"unsloth/gemma-2-2b-it-bnb-4bit", # 256,128 tokens vocabulary
]

regex_case = [
(r"\d{3}-\d{2}-\d{4}", "203-22-1234"),
(
r"(https?:\/\/)?([\da-z\.-]+)\.([a-z\.]{2,6})([\/\w \.-]*)*\/?",
"https://github.com/outlines-dev/outlines",
),
(
r"A: [\w \.\*\-=\+,\?/]{10,50}\. The answer is [1-9][0-9]{0,9}\.",
"A: Some thoughts before answering. The answer is 42.",
),
(
"(0|[1-9][0-9]*)|true|false|([a-zA-Z_][a-zA-Z_0-9]*)",
"AVeryLongStringtoTest1234",
),
(r"\+[1-9]\d{1,14}", "1234567891234"),
]
from .data import json_cases, models, regex_cases


class OutlinesRegex:
params = [models, regex_case]
params = [models, regex_cases]
param_names = ["model", "regex"]
timeout = 1200

Expand All @@ -59,7 +37,7 @@ def time_outlines(self, _, regex):
"""
caching.clear_cache()

regex_string, regex_example = regex
regex_string, regex_example = regex["regex"], regex["example"]
regex_example_tokens = self.tokenizer.encode(regex_example)[0][0]
guide = RegexGuide(regex_string, self.tokenizer)

Expand All @@ -69,68 +47,8 @@ def time_outlines(self, _, regex):
state = guide.get_next_state(state, token)


json_case = [
(
{
"$defs": {
"Armor": {
"enum": ["leather", "chainmail", "plate"],
"title": "Armor",
"type": "string",
}
},
"properties": {
"name": {"maxLength": 10, "title": "Name", "type": "string"},
"age": {"title": "Age", "type": "integer"},
"armor": {"$ref": "#/$defs/Armor"},
"strength": {"title": "Strength", "type": "integer"},
},
"required": ["name", "age", "armor", "strength"],
"title": "Character",
"type": "object",
},
"""{'name': 'Super Warrior', 'age': 26, 'armor': 'leather', 'armor': 10}""",
),
(
{
"$schema": "http://json-schema.org/draft-04/schema#",
"title": "Schema for a recording",
"type": "object",
"definitions": {
"artist": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"functions": {"type": "array", "items": {"type": "string"}},
},
"required": ["id", "name", "functions"],
}
},
"properties": {
"id": {"type": "number"},
"work": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"composer": {"$ref": "#/definitions/artist"},
},
},
"recording_artists": {
"type": "array",
"items": {"$ref": "#/definitions/artist"},
},
},
"required": ["id", "work", "recording_artists"],
},
"""{'id': 999, 'work': {'id': 1, 'name': 'Strasbourg Saint-Denis', 'composer': 'Roy Hargrove'}, 'recording_artists': [{'id': 2, 'name': 'Roy Hargrove', 'functions': ['Trumpet', 'Singing']}]}""",
),
]


class OutlinesJsonSchema:
params = [models, json_case]
params = [models, json_cases]
param_names = ["model", "json"]
timeout = 1200

Expand All @@ -154,7 +72,7 @@ def time_outlines(self, _, json_case):
regular expression, and walking this index while generating tokens.

"""
json_string, json_example = json_case
json_string, json_example = json_case["schema"], json_case["example"]
json_example_tokens = self.tokenizer.encode(json_example)[0][0]

regex_string = build_regex_from_schema(json.dumps(json_string))
Expand Down
92 changes: 5 additions & 87 deletions src/benchmark_outlines_core.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,33 +5,11 @@
from outlines_core.models.transformers import TransformerTokenizer
from transformers import AutoTokenizer

models = [
"NousResearch/Nous-Hermes-llama-2-7b", # 32,000 tokens vocabulary
"gpt2", # 50,257 tokens vocabulary
"NousResearch/Hermes-3-Llama-3.1-8B", # 128,256 tokens vocabulary
"unsloth/gemma-2-2b-it-bnb-4bit", # 256,128 tokens vocabulary
]

regex_case = [
(r"\d{3}-\d{2}-\d{4}", "203-22-1234"),
(
r"(https?:\/\/)?([\da-z\.-]+)\.([a-z\.]{2,6})([\/\w \.-]*)*\/?",
"https://github.com/outlines-dev/outlines",
),
(
r"A: [\w \.\*\-=\+,\?/]{10,50}\. The answer is [1-9][0-9]{0,9}\.",
"A: Some thoughts before answering. The answer is 42.",
),
(
"(0|[1-9][0-9]*)|true|false|([a-zA-Z_][a-zA-Z_0-9]*)",
"AVeryLongStringtoTest1234",
),
(r"\+[1-9]\d{1,14}", "1234567891234"),
]
from .data import json_cases, models, regex_cases


class OutlinesCoreRegex:
params = [models, regex_case]
params = [models, regex_cases]
param_names = ["model", "regex"]
timeout = 600

Expand All @@ -54,7 +32,7 @@ def time_outlines_core(self, _, regex):
regular expression, and walking this index while generating tokens.

"""
regex_string, regex_example = regex
regex_string, regex_example = regex["regex"], regex["example"]
regex_example_tokens = self.tokenizer.encode(regex_example)[0][0]
guide = RegexGuide(regex_string, self.tokenizer)

Expand All @@ -64,68 +42,8 @@ def time_outlines_core(self, _, regex):
state = guide.get_next_state(state, token)


json_case = [
(
{
"$defs": {
"Armor": {
"enum": ["leather", "chainmail", "plate"],
"title": "Armor",
"type": "string",
}
},
"properties": {
"name": {"maxLength": 10, "title": "Name", "type": "string"},
"age": {"title": "Age", "type": "integer"},
"armor": {"$ref": "#/$defs/Armor"},
"strength": {"title": "Strength", "type": "integer"},
},
"required": ["name", "age", "armor", "strength"],
"title": "Character",
"type": "object",
},
"""{'name': 'Super Warrior', 'age': 26, 'armor': 'leather', 'armor': 10}""",
),
(
{
"$schema": "http://json-schema.org/draft-04/schema#",
"title": "Schema for a recording",
"type": "object",
"definitions": {
"artist": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"functions": {"type": "array", "items": {"type": "string"}},
},
"required": ["id", "name", "functions"],
}
},
"properties": {
"id": {"type": "number"},
"work": {
"type": "object",
"properties": {
"id": {"type": "number"},
"name": {"type": "string"},
"composer": {"$ref": "#/definitions/artist"},
},
},
"recording_artists": {
"type": "array",
"items": {"$ref": "#/definitions/artist"},
},
},
"required": ["id", "work", "recording_artists"],
},
"""{'id': 999, 'work': {'id': 1, 'name': 'Strasbourg Saint-Denis', 'composer': 'Roy Hargrove'}, 'recording_artists': [{'id': 2, 'name': 'Roy Hargrove', 'functions': ['Trumpet', 'Singing']}]}""",
),
]


class OutlinesCoreJsonSchema:
params = [models, json_case]
params = [models, json_cases]
param_names = ["model", "json"]
timeout = 600

Expand All @@ -148,7 +66,7 @@ def time_outlines_core(self, _, json_case):
regular expression, and walking this index while generating tokens.

"""
json_string, json_example = json_case
json_string, json_example = json_case["schema"], json_case["example"]
json_example_tokens = self.tokenizer.encode(json_example)[0][0]

regex_string = build_regex_from_schema(json.dumps(json_string))
Expand Down
Loading