Qwen2.5-Coder-7B-Instruct

MERA Created at 04.11.2024 17:55
0.464
The overall result
207
Place in the rating
Weak tasks:
258
RWSD
331
PARus
187
RCB
160
ruEthics
214
MultiQ
271
ruWorldTree
277
ruOpenBookQA
375
CheGeKa
296
ruMMLU
158
ruHateSpeech
297
ruDetox
306
ruHHH
174
ruTiE
123
ruHumanEval
198
USE
284
MathLogicQA
161
ruMultiAr
182
SimpleAr
195
LCS
57
BPS
237
ruModAr
292
MaMuRAMu
102
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.126 Accuracy
RCB 0.548 / 0.446 Accuracy F1 macro
USE 0.18 Grade norm
RWSD 0.538 Accuracy
PARus 0.716 Accuracy
ruTiE 0.735 Accuracy
MultiQ 0.399 / 0.302 F1 Exact match
CheGeKa 0.041 / 0.022 F1 Exact match
ruModAr 0.483 Exact match
MaMuRAMu 0.617 Accuracy
ruMultiAr 0.308 Exact match
ruCodeEval 0.207 / 0.412 / 0.494 Pass@k
MathLogicQA 0.403 Accuracy
ruWorldTree 0.855 / 0.855 Accuracy F1 macro
ruOpenBookQA 0.745 / 0.744 Accuracy F1 macro

Evaluation on open tasks:

Go to the ratings by subcategory

The table will scroll to the left

Task name Result Metric
BPS 0.991 Accuracy
ruMMLU 0.52 Accuracy
SimpleAr 0.978 Exact match
ruHumanEval 0.173 / 0.357 / 0.427 Pass@k
ruHHH 0.624
ruHateSpeech 0.781
ruDetox 0.161
ruEthics
Correct God Ethical
Virtue 0.366 0.363 0.381
Law 0.355 0.36 0.357
Moral 0.397 0.369 0.395
Justice 0.315 0.3 0.323
Utilitarianism 0.306 0.293 0.354

Information about the submission

Mera version
v.1.2.0
Torch Version
2.4.0
The version of the codebase
9b26db97
CUDA version
12.1
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.44.2
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
vllm

Team:

MERA

Name of the ML model:

Qwen2.5-Coder-7B-Instruct

Model size

7.0B

Model type:

Opened

SFT

Additional links:

https://arxiv.org/pdf/2409.12186

Architecture description:

Type: Causal Language Models Training Stage: Pretraining & Post-training Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias Number of Parameters: 7.61B Number of Paramaters (Non-Embedding): 6.53B Number of Layers: 28 Number of Attention Heads (GQA): 28 for Q and 4 for KV Context Length: Full 131,072 tokens

Description of the training:

-

Pretrain data:

Large-scale, high-quality, and diverse data forms the foundation of pre-trained models. To this end, we constructed a dataset named Qwen2.5-Coder-Data. This dataset comprises five key data types: Source Code Data, Text-Code Grounding Data, Synthetic Data, Math Data, and Text Data, totaling 5.5 trillion tokens.

License:

apache-2.0

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=["\n"]; \nchegeka - do_sample=false;until=["\n"]; \nrudetox - do_sample=false;until=["\n"]; \nrumultiar - do_sample=false;until=["\n"]; \nuse - do_sample=false;until=["\n","."]; \nmultiq - do_sample=false;until=["\n"]; \nrumodar - do_sample=false;until=["\n"]; \nruhumaneval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6; \nrucodeeval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6;

The size of the context:
simplear, bps, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, ruhumaneval, rucodeeval - 32768 \nrutie - 5000

System prompt:
Реши задачу по инструкции ниже. Не давай никаких объяснений и пояснений к своему ответу. Не пиши ничего лишнего. Пиши только то, что указано в инструкции. Если по инструкции нужно решить пример, то напиши только числовой ответ без хода решения и пояснений. Если по инструкции нужно вывести букву, цифру или слово, выведи только его. Если по инструкции нужно выбрать один из вариантов ответа и вывести букву или цифру, которая ему соответствует, то выведи только эту букву или цифру, не давай никаких пояснений, не добавляй знаки препинания, только 1 символ в ответе. Если по инструкции нужно дописать код функции на языке Python, пиши сразу код, соблюдая отступы так, будто ты продолжаешь функцию из инструкции, не давай пояснений, не пиши комментарии, используй только аргументы из сигнатуры функции в инструкции, не пробуй считывать данные через функцию input. Не извиняйся, не строй диалог. Выдавай только ответ и ничего больше.

Ratings by subcategory

Metric: Grade Norm
Model, team 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 8_0 8_1 8_2 8_3 8_4
Qwen2.5-Coder-7B-Instruct
MERA
0.467 0.167 0.733 0.1 0.033 0.067 0 - 0 0 0 0 0 0.033 0.133 0.3 0 0.033 0.033 0 0.033 0.467 0.067 0.1 0.167 0.367 0.167 0.1 0.567 0.233 0.367
Model, team Honest Helpful Harmless
Qwen2.5-Coder-7B-Instruct
MERA
0.541 0.627 0.707
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Management Philosophy Prehistory Human aging Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Biology (college) Physics (college) Human Sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine learning Medical genetics Professional law PR Security studies Chemistry (школьная) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual_physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) European history Government and politics
Qwen2.5-Coder-7B-Instruct
MERA
0.407 0.476 0.625 0.739 0.569 0.687 0.66 0.601 0.506 0.587 0.386 0.397 0.33 0.62 0.605 0.523 0.6 0.507 0.378 0.542 0.249 0.62 0.42 0.514 0.429 0.55 0.377 0.537 0.665 0.35 0.7 0.678 0.552 0.747 0.555 0.568 0.43 0.674 0.43 0.517 0.707 0.445 0.531 0.581 0.685 0.505 0.554 0.404 0.383 0.492 0.53 0.62 0.569 0.597 0.76 0.667 0.575
Model, team SIM FL STA
Qwen2.5-Coder-7B-Instruct
MERA
0.807 0.611 0.373
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Managment Philosophy Pre-History Gerontology Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Bilology (college) Physics (college) Human sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine Learning Genetics Professional law PR Security Chemistry (college) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical Engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional Accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) Europe History Government and politics
Qwen2.5-Coder-7B-Instruct
MERA
0.444 0.584 0.5 0.509 0.618 0.569 0.569 0.667 0.558 0.615 0.603 0.642 0.467 0.651 0.497 0.568 0.682 0.667 0.526 0.684 0.298 0.678 0.756 0.609 0.822 0.652 0.59 0.509 0.825 0.622 0.778 0.654 0.652 0.737 0.485 0.679 0.711 0.778 0.526 0.477 0.633 0.651 0.756 0.733 0.862 0.889 0.724 0.818 0.646 0.772 0.756 0.478 0.709 0.558 0.535 0.462 0.711
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-7B-Instruct
MERA
0.366 0.355 0.397 0.315 0.306
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-7B-Instruct
MERA
0.363 0.36 0.369 0.3 0.293
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-7B-Instruct
MERA
0.381 0.357 0.395 0.323 0.354
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-Coder-7B-Instruct
MERA
0.787 0.686 0.765 0.811 0.571 0.836