Qwen2.5-Coder-1.5B

MERA Created at 04.11.2024 17:55
0.317
The overall result
458
Place in the rating
Weak tasks:
314
RWSD
486
PARus
474
RCB
480
MultiQ
506
ruWorldTree
503
ruOpenBookQA
505
CheGeKa
495
ruMMLU
410
ruHateSpeech
586
ruDetox
462
ruHHH
445
ruTiE
209
ruHumanEval
495
USE
428
MathLogicQA
324
ruMultiAr
263
SimpleAr
515
LCS
341
BPS
302
ruModAr
464
MaMuRAMu
223
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.056 Accuracy
RCB 0.418 / 0.322 Accuracy F1 macro
USE 0.058 Grade norm
RWSD 0.523 Accuracy
PARus 0.534 Accuracy
ruTiE 0.525 Accuracy
MultiQ 0.212 / 0.116 F1 Exact match
CheGeKa 0.013 / 0 F1 Exact match
ruModAr 0.456 Exact match
MaMuRAMu 0.438 Accuracy
ruMultiAr 0.231 Exact match
ruCodeEval 0.02 / 0.077 / 0.134 Pass@k
MathLogicQA 0.329 Accuracy
ruWorldTree 0.531 / 0.514 Accuracy F1 macro
ruOpenBookQA 0.473 / 0.45 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 0.873 Accuracy
ruMMLU 0.362 Accuracy
SimpleAr 0.949 Exact match
ruHumanEval 0.048 / 0.19 / 0.299 Pass@k
ruHHH 0.506
ruHateSpeech 0.57
ruDetox 0.049
ruEthics
Correct God Ethical
Virtue 0.015 -0.028 -0.063
Law 0.026 0.003 -0.037
Moral 0.036 -0.015 -0.042
Justice 0.03 -0.028 -0.055
Utilitarianism 0.025 -0.031 -0.024

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-1.5B

Model size

1.5B

Model type:

Opened

Pretrain

Additional links:

https://arxiv.org/pdf/2409.12186

Architecture description:

Type: Causal Language Models Training Stage: Pretraining Architecture: transformers with RoPE, SwiGLU, RMSNorm, Attention QKV bias and tied word embeddings Number of Parameters: 1.54B Number of Paramaters (Non-Embedding): 1.31B Number of Layers: 28 Number of Attention Heads (GQA): 12 for Q and 2 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 - 10000

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-1.5B
MERA
0 0.067 0.433 0.233 0 0.067 0 - 0 0.033 0 0 0.033 0 0.133 0.067 0.033 0.033 0.033 0.033 0 0 0.033 0 0 0.075 0.133 0.033 0.033 0.067 0.1
Model, team Honest Helpful Harmless
Qwen2.5-Coder-1.5B
MERA
0.508 0.559 0.448
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-1.5B
MERA
0.348 0.398 0.355 0.611 0.405 0.478 0.427 0.408 0.386 0.426 0.219 0.27 0.34 0.444 0.411 0.428 0.42 0.333 0.233 0.427 0.242 0.45 0.33 0.306 0.321 0.41 0.302 0.417 0.465 0.25 0.55 0.554 0.331 0.596 0.4 0.333 0.32 0.352 0.258 0.374 0.364 0.232 0.483 0.371 0.378 0.278 0.348 0.326 0.301 0.354 0.27 0.422 0.321 0.357 0.55 0.448 0.347
Model, team SIM FL STA
Qwen2.5-Coder-1.5B
MERA
0.817 0.551 0.158
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-1.5B
MERA
0.333 0.446 0.3 0.472 0.368 0.414 0.328 0.351 0.365 0.415 0.564 0.425 0.283 0.55 0.345 0.284 0.561 0.333 0.263 0.439 0.281 0.356 0.356 0.391 0.667 0.394 0.462 0.439 0.649 0.533 0.689 0.538 0.536 0.404 0.364 0.375 0.311 0.489 0.404 0.446 0.367 0.365 0.667 0.578 0.69 0.822 0.397 0.455 0.6 0.509 0.644 0.348 0.532 0.468 0.372 0.339 0.544
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-1.5B
MERA
0.015 0.026 0.036 0.03 0.025
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-1.5B
MERA
-0.028 0.003 -0.015 -0.028 -0.031
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-Coder-1.5B
MERA
-0.063 -0.037 -0.042 -0.055 -0.024
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-Coder-1.5B
MERA
0.611 0.657 0.647 0.649 0 0.443