Qwen 0.5B Instruct

НГУ Created at 17.08.2024 07:57

0.311

The overall result

The submission does not contain all the required tasks

Ratings for leaderboard tasks

The table will scroll to the left

Task name	Result	Metric
LCS	0.098	Accuracy
RCB	0.336 / 0.219	Accuracy F1 macro
USE	0.008	Grade norm
RWSD	0.515	Accuracy
PARus	0.488	Accuracy
ruTiE	0.5	Accuracy
MultiQ	0.075 / 0.021	F1 Exact match
CheGeKa	0.003 / 0	F1 Exact match
ruModAr	0.243	Exact match
ruMultiAr	0.142	Exact match
MathLogicQA	0.321	Accuracy
ruWorldTree	0.49 / 0.484	Accuracy F1 macro
ruOpenBookQA	0.508 / 0.506	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

0.568

Accuracy

ruMMLU

0.396

Accuracy

SimpleAr

0.679

Exact match

ruHumanEval

0 / 0 / 0

Pass@k

ruHHH

0.466

ruHateSpeech

0.487

ruDetox

0.109

ruEthics

	Correct	God	Ethical
Virtue	0.071	0.046	0.026
Law	0.071	0.058	0.041
Moral	0.078	0.053	0.021
Justice	0.102	0.066	0.052
Utilitarianism	0.069	0.043	0.025

Information about the submission:

Mera version

Torch Version

The version of the codebase

CUDA version

Precision of the model weights

Seed

Butch

Transformers version

The number of GPUs and their type

Architecture

Team:

НГУ

Name of the ML model:

Qwen 0.5B Instruct

Link to the ML model:

https://huggingface.co/Qwen/Qwen2-0.5B-Instruct

Architecture description:

Qwen2 is a language model including decoder of 0.5B size. It is based on the Transformer architecture with SwiGLU activation, attention QKV bias, group query attention, etc. Additionally, tokenizer is improved for adaptation to multiple natural languages and codes.

Description of the training:

The model was pretrained with a large amount of data, after that it was post-trained with both supervised finetuning and direct preference optimization.

Pretrain data:

The model was pretrained with a large amount of data of English, Chinese and 27 additional languages including Russian. In terms of the context length, the model was pretrained on data of the context length of 32K tokens.

Training Details:

The Group Query Attention was applied so that the model can enjoy the benefits of faster speed and less memory usage in model inference. Also, the tying embedding was used as the large sparse embeddings take up a large proportion of the total model parameters.

License:

Apache 2.0

Strategy, generation and parameters:

All the parameters were not changed and are used as prepared by the model's authors. Details: - 1 x NVIDIA A100 80GB - dtype float32- Pytorch 2.3.1 + CUDA 11.7 - Transformers 4.38.2 - Context length 32768

Expand information

Ratings by subcategory

Metric: Accuracy

Model, team	Honest	Helpful	Harmless
Qwen 0.5B Instruct НГУ	0.41	0.458	0.534

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
Qwen 0.5B Instruct НГУ	0.4	0.625	0.4	0.429	0.333	0.4	0.467	0.235	0.2	0.6	0.273	0.3	0.5	0.423	0.273	0.3	0.5	0.407	0.5	0.7	0	0.269	0.3	0.431	0.2	0.364	0.688	0.714	0.5	0.545	0.2	0.222	0.4	0.2	0.636	0.5	0.4	0.381	0.3	0.3	0.291	0.4	0.5	0.4	0.75	0.5	0.3	0.5	0.5	0.5	0.364	0.313	0.471	0.533	0.417	0.333	0.407

Model, team	SIM	FL	STA
Qwen 0.5B Instruct НГУ	0.37	0.568	0.417

Coorect

Good

Ethical

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
Qwen 0.5B Instruct НГУ	0.071	0.071	0.078	0.102	0.069

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
Qwen 0.5B Instruct НГУ	0.046	0.058	0.053	0.066	0.043

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
Qwen 0.5B Instruct НГУ	0.026	0.041	0.021	0.052	0.025

Model, team	Women	Men	LGBT	Nationalities	Migrants	Other
Qwen 0.5B Instruct НГУ	0.454	0.571	0.588	0.432	0.286	0.525

Qwen 0.5B Instruct

Ratings for leaderboard tasks

Evaluation on open tasks:

Information about the submission:

Team:

Name of the ML model:

Link to the ML model:

Architecture description:

Description of the training:

Pretrain data:

Training Details:

License:

Strategy, generation and parameters:

Ratings by subcategory

Confirm the deletion of the sub