File size: 8,962 Bytes
1613f96 cc7bfac 1613f96 a8bf9ff e9469f3 a8bf9ff 1613f96 16e1e6b 1613f96 16e1e6b 1613f96 9b88843 1613f96 22ec7d4 811fa97 1613f96 197af1c c430a49 811fa97 c430a49 197af1c 811fa97 197af1c c430a49 811fa97 c430a49 197af1c c430a49 811fa97 69139f5 811fa97 197af1c 811fa97 22ec7d4 1613f96 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 |
from dataclasses import dataclass
from enum import Enum
@dataclass
class Task:
benchmark: str
metric: str
col_name: str
# Select your tasks here
# ---------------------------------------------------
class Tasks(Enum):
# task_key in the json file, metric_key in the json file, name to display in the leaderboard
task0 = Task("hf", "score", "LLM JUDGE")
task1 = Task("rouge1", "score", "ROUGE-unigrams")
task2 = Task("rouge2", "score", "ROUGE-bigrams")
task3 = Task("rougeL", "score", "ROUGE-Longest Common Subsequence")
task4 = Task("rougeLsum", "score", "ROUGE-Lsum")
task5 = Task("bleu", "score", "Bleu")
task6 = Task("brevity_penalty", "score", "Brevity Penalty")
task7 = Task("hf", "score", "LLM JUDGE")
# task = Task("precisions", "score", "Precision")
NUM_FEWSHOT = 0 # Change with your few shot
# ---------------------------------------------------
# Your leaderboard name
TITLE = """<img src="https://iq.wiki/branding/downloadassets/logooriginalbrain.svg" height="40" style="display: block; margin-left: auto; margin-right: auto;">
<h1 align="center" id="space-title"></h1>"""
# What does your leaderboard evaluate?
INTRODUCTION_TEXT = """
<center>Solbench: Evaluating Solidity Code Generation</center>
"""
# Which evaluations are you running? how can people reproduce what you have?
LLM_BENCHMARKS_TEXT = """
## How it works
## Reproducibility
To reproduce our results, here is the commands you can run:
"""
EVALUATION_QUEUE_TEXT = """
## Some good practices before submitting a model
### 1) Make sure you can load your model and tokenizer using AutoClasses:
```python
from transformers import AutoConfig, AutoModel, AutoTokenizer
config = AutoConfig.from_pretrained("your model name", revision=revision)
model = AutoModel.from_pretrained("your model name", revision=revision)
tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
```
If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
Note: make sure your model is public!
Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!
### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
### 3) Make sure your model has an open license!
This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
### 4) Fill up your model card
When we add extra information about models to the leaderboard, it will be automatically taken from the model card
## In case of model failure
If your model is displayed in the `FAILED` category, its execution stopped.
Make sure you have followed the above steps first.
If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task).
"""
EVALUATION_SCRIPT = '''
To evaluate the model you can access the colab notebook at [this link](https://colab.research.google.com/drive/145KAGvgdAb8BrkObUrxAVWBd9EGDqy8N?usp=sharing).
## First install the necessary libraries
```
pip install accelerate openai anthropic datasets
```
## Setup your :
* OPENAI_API_KEY
* ANTHROPIC_API_KEY
* HF_TOKEN
## Select a model
```python
MODEL_ID = # model_id_here
```
## Then run the following script
````python
from transformers import pipeline
import torch
import os
import json
from openai import OpenAI
import anthropic
from huggingface_hub.utils._token import get_token
from huggingface_hub import InferenceClient
HF_TOKEN = get_token()
from datasets import load_dataset
ds = load_dataset("braindao/solbench-naive-judge-random-v1",split="test")
pipe = pipeline("text-generation", model= MODEL_ID , torch_dtype=torch.bfloat16, device_map="auto")
def generate(message):
messages = [
{"role": "user", "content": message},
]
return pipe(messages,max_new_tokens=1024)[0]["generated_text"][1]["content"]
def convert_to_int(text):
value = 0
try :
value = int(text)
except :
pass
return value
def anthropic_judge(code,baseline):
prompt = f"""Analyze the provided Solidity code and assign a score from 0 to 10 based on these criteria:
1. Functionality (0-2 points)
2. Security (0-2 points)
3. Efficiency (0-2 points)
4. Readability and Style (0-2 points)
5. Similarity with the Expert Code (0-2 points)
We
Evaluate the code thoroughly, sum up the points, and return ONLY an integer value representing the final score. Your entire response should consist of a single integer between 0 and 10, inclusive.
Solidity code to evaluate:
```solidity
{code}
```
Expert Code:
```solidity
{baseline}
```
OUTPUT FORMAT: [integer]"""
sys = """You are a solidity code judge,
You will only reply with an integer value between 0-10"""
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-3-5-sonnet-20240620",
max_tokens=1000,
temperature=0,
system=sys,
messages=[
{
"role": "user",
"content": [
{
"type": "text",
"text": prompt
}
]
}
]
)
return convert_to_int(message.content[0].text)
def openai_judge(code,baseline):
prompt = f"""evaluate the following solidity code and return a score between 0 and 10 based how far the code achieves the following criteria:
1. Functionality (0-2 points)
2. Security (0-2 points)
3. Efficiency (0-2 points)
4. Readability and Style (0-2 points)
5. Similarity with the Expert Code (0-2 points)
code to evaluate:
{code}
expert code:
{baseline}
return only an integer value and no additional comment, score should be either 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10.
"""
client = OpenAI()
completion = client.chat.completions.create(
model="gpt-4o",
messages=[
{"role": "user", "content": prompt}
]
)
return convert_to_int(completion.choices[0].message.content)
def hf_judge(code,baseline):
prompt = f"""evaluate the following solidity code and return a score between 0 and 10 based how far the code achieves the following criteria:
1. Functionality (0-2 points)
2. Security (0-2 points)
3. Efficiency (0-2 points)
4. Readability and Style (0-2 points)
5. Similarity with the Expert Code (0-2 points)
code to evaluate:
{code}
expert code:
{baseline}
return only an integer value and no additional comment, score should be either 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10.
"""
client = InferenceClient(
"meta-llama/Meta-Llama-3.1-405B-Instruct",
token=HF_TOKEN,
)
out = ""
try :
for message in client.chat_completion(
messages=[{"role":"system","content" : "you are a solidity code judge, you will only reply with an integer value between 0-10"},
{"role": "user", "content": prompt}],
max_tokens=500,
stream=True,
):
out += message.choices[0].delta.content
except :
pass
return convert_to_int(out)
def LLM_JUDGE(code,baseline,judges=["openai","anthropic","hf"]) :
out = {}
if "openai" in judges :
out["openai"] = openai_judge(code,baseline)
if "anthropic" in judges :
out["anthropic"] = anthropic_judge(code,baseline)
if "hf" in judges :
out["hf"] = hf_judge(code,baseline)
return out
# Judge model against data
from tqdm import tqdm
scores = {"openai":[],"anthropic":[],"hf":[]}
for sample in tqdm(ds) :
score = evaluate_sample(sample)
for key in score.keys():
scores[key].append(score[key])
# normalize scores
for key in scores.keys():
scores[key] = sum(scores[key])/(10*len(scores[key]))
d = {
"config": {
"model_dtype": "torch.bfloat16",
"model_name": MODEL_ID,
"model_sha": "main"
},
"results": {
"openai": {
"score": 0
},
"anthropic": {
"score": 0
},
"hf": {
"score": 0
}
}
}
for key in scores.keys() :
d["results"][key]["score"] = scores[key]
# Serializing json
json_object = json.dumps(d, indent=4)
# Writing to sample.json
file_name = MODEL_ID.split("/")[1] + ".json"
with open(file_name, "w") as outfile:
outfile.write(json_object)
````
## if you are not part of braindao set `create_pr` to **True**
```python
from huggingface_hub import upload_file
upload_file(path_or_fileobj = file_name,
path_in_repo=f"{MODEL_ID}.json",
repo_id="braindao/results",
repo_type="dataset",
create_pr=False)
```
'''
CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
CITATION_BUTTON_TEXT = r"""
"""
|