M4-ai/tau-0.5B-GGUF

Quantized GGUF model files for tau-0.5B from M4-ai

Name Quant method Size
tau-0.5b.fp16.gguf fp16 1.25 GB
tau-0.5b.q2_k.gguf q2_k 298.41 MB
tau-0.5b.q3_k_m.gguf q3_k_m 349.88 MB
tau-0.5b.q4_k_m.gguf q4_k_m 407.16 MB
tau-0.5b.q5_k_m.gguf q5_k_m 459.24 MB
tau-0.5b.q6_k.gguf q6_k 514.58 MB
tau-0.5b.q8_0.gguf q8_0 664.60 MB

Original Model Card:

tau-0.5B

Model Details

  • Model Name: tau-0.5B
  • Base Model: Qwen1.5-0.5B
  • Dataset: UltraTextbooks-2.0
  • Model Size: 0.5B parameters
  • Model Type: Language Model
  • Training Procedure: Further pre-training of Qwen1.5-0.5B on UltraTextbooks-2.0.

Model Use

tau-0.5B is designed to be a general-purpose language model with enhanced capabilities in the domains of machine learning, mathematics, and coding. It can be used for a wide range of natural language processing tasks, such as:

  • Educational question answering
  • Text summarization
  • Content generation for educational purposes
  • Code understanding and generation
  • Mathematical problem solving

The model's exposure to the diverse content in the UltraTextbooks-2.0 dataset makes it particularly well-suited for applications in educational technology and research.

Training Data

tau-0.5B was further pre-trained on the UltraTextbooks-2.0 dataset, which is an expanded version of the original UltraTextbooks dataset. UltraTextbooks-2.0 incorporates additional high-quality synthetic and human-written textbooks from various sources on the Hugging Face platform, with a focus on increasing the diversity of content in the domains of machine learning, mathematics, and coding.

For more details on the dataset, please refer to the UltraTextbooks-2.0 Dataset Card.

Performance and Limitations

Refer to Evaluation for evaluations. It is essential to note that the model may still exhibit biases or inaccuracies present in the training data. Users are encouraged to critically evaluate the model's outputs and report any issues to facilitate continuous improvement.

Environmental Impact

The training of tau-0.5B required computational resources that contribute to the model's overall environmental impact. However, efforts were made to optimize the training process and minimize the carbon footprint.

Ethical Considerations

tau-0.5B was trained on a diverse dataset that may contain biases and inaccuracies. Users should be aware of these potential limitations and use the model responsibly. The model should not be used for tasks that could cause harm or discriminate against individuals or groups.

Evaluation

Tasks Version Filter n-shot Metric Value Stderr
agieval_nous N/A none 0 acc 0.2235 ± 0.0434
none 0 acc_norm 0.2141 ± 0.0498
- agieval_aqua_rat 1 none 0 acc 0.1417 ± 0.0219
none 0 acc_norm 0.1535 ± 0.0227
- agieval_logiqa_en 1 none 0 acc 0.2796 ± 0.0176
none 0 acc_norm 0.3118 ± 0.0182
- agieval_lsat_ar 1 none 0 acc 0.2000 ± 0.0264
none 0 acc_norm 0.1696 ± 0.0248
- agieval_lsat_lr 1 none 0 acc 0.2275 ± 0.0186
none 0 acc_norm 0.2020 ± 0.0178
- agieval_lsat_rc 1 none 0 acc 0.1487 ± 0.0217
none 0 acc_norm 0.1561 ± 0.0222
- agieval_sat_en 1 none 0 acc 0.2330 ± 0.0295
none 0 acc_norm 0.2039 ± 0.0281
- agieval_sat_en_without_passage 1 none 0 acc 0.2524 ± 0.0303
none 0 acc_norm 0.1942 ± 0.0276
- agieval_sat_math 1 none 0 acc 0.2227 ± 0.0281
none 0 acc_norm 0.1682 ± 0.0253
Tasks Version Filter n-shot Metric Value Stderr
truthfulqa 2 none 0 acc 0.3931 ± 0.0143
mmlu N/A none 0 acc 0.3642 ± 0.0040
- humanities N/A none 5 acc 0.3320 ± 0.0068
- formal_logic 0 none 5 acc 0.2619 ± 0.0393
- high_school_european_history 0 none 5 acc 0.4909 ± 0.0390
- high_school_us_history 0 none 5 acc 0.4167 ± 0.0346
- high_school_world_history 0 none 5 acc 0.4641 ± 0.0325
- international_law 0 none 5 acc 0.5537 ± 0.0454
- jurisprudence 0 none 5 acc 0.4167 ± 0.0477
- logical_fallacies 0 none 5 acc 0.2638 ± 0.0346
- moral_disputes 0 none 5 acc 0.3757 ± 0.0261
- moral_scenarios 0 none 5 acc 0.2402 ± 0.0143
- philosophy 0 none 5 acc 0.3794 ± 0.0276
- prehistory 0 none 5 acc 0.3426 ± 0.0264
- professional_law 0 none 5 acc 0.3103 ± 0.0118
- world_religions 0 none 5 acc 0.2807 ± 0.0345
- other N/A none 5 acc 0.4071 ± 0.0088
- business_ethics 0 none 5 acc 0.4200 ± 0.0496
- clinical_knowledge 0 none 5 acc 0.4491 ± 0.0306
- college_medicine 0 none 5 acc 0.3873 ± 0.0371
- global_facts 0 none 5 acc 0.3600 ± 0.0482
- human_aging 0 none 5 acc 0.3498 ± 0.0320
- management 0 none 5 acc 0.4854 ± 0.0495
- marketing 0 none 5 acc 0.5470 ± 0.0326
- medical_genetics 0 none 5 acc 0.4000 ± 0.0492
- miscellaneous 0 none 5 acc 0.4291 ± 0.0177
- nutrition 0 none 5 acc 0.4183 ± 0.0282
- professional_accounting 0 none 5 acc 0.3582 ± 0.0286
- professional_medicine 0 none 5 acc 0.3015 ± 0.0279
- virology 0 none 5 acc 0.3494 ± 0.0371
- social_sciences N/A none 5 acc 0.4075 ± 0.0088
- econometrics 0 none 5 acc 0.2719 ± 0.0419
- high_school_geography 0 none 5 acc 0.5000 ± 0.0356
- high_school_government_and_politics 0 none 5 acc 0.4611 ± 0.0360
- high_school_macroeconomics 0 none 5 acc 0.4051 ± 0.0249
- high_school_microeconomics 0 none 5 acc 0.3908 ± 0.0317
- high_school_psychology 0 none 5 acc 0.4239 ± 0.0212
- human_sexuality 0 none 5 acc 0.3893 ± 0.0428
- professional_psychology 0 none 5 acc 0.3399 ± 0.0192
- public_relations 0 none 5 acc 0.4455 ± 0.0476
- security_studies 0 none 5 acc 0.3510 ± 0.0306
- sociology 0 none 5 acc 0.5174 ± 0.0353
- us_foreign_policy 0 none 5 acc 0.5500 ± 0.0500
- stem N/A none 5 acc 0.3276 ± 0.0083
- abstract_algebra 0 none 5 acc 0.3000 ± 0.0461
- anatomy 0 none 5 acc 0.2889 ± 0.0392
- astronomy 0 none 5 acc 0.3487 ± 0.0388
- college_biology 0 none 5 acc 0.3403 ± 0.0396
- college_chemistry 0 none 5 acc 0.2600 ± 0.0441
- college_computer_science 0 none 5 acc 0.3800 ± 0.0488
- college_mathematics 0 none 5 acc 0.3300 ± 0.0473
- college_physics 0 none 5 acc 0.2745 ± 0.0444
- computer_security 0 none 5 acc 0.4300 ± 0.0498
- conceptual_physics 0 none 5 acc 0.3447 ± 0.0311
- electrical_engineering 0 none 5 acc 0.3931 ± 0.0407
- elementary_mathematics 0 none 5 acc 0.3095 ± 0.0238
- high_school_biology 0 none 5 acc 0.4161 ± 0.0280
- high_school_chemistry 0 none 5 acc 0.2759 ± 0.0314
- high_school_computer_science 0 none 5 acc 0.3100 ± 0.0465
- high_school_mathematics 0 none 5 acc 0.3185 ± 0.0284
- high_school_physics 0 none 5 acc 0.2517 ± 0.0354
- high_school_statistics 0 none 5 acc 0.3009 ± 0.0313
- machine_learning 0 none 5 acc 0.3036 ± 0.0436
medqa_4options Yaml none 5 acc 0.2687 ± 0.0124
none 5 acc_norm 0.2687 ± 0.0124
logieval 0 get-answer 5 exact_match 0.3505 ± 0.0120
gsm8k_cot 3 strict-match 8 exact_match 0.0690 ± 0.0070
flexible-extract 8 exact_match 0.1365 ± 0.0095
Tasks Version Filter n-shot Metric Value Stderr
arc_easy 1 none 25 acc 0.5981 ± 0.0101
none 25 acc_norm 0.5939 ± 0.0101
arc_challenge 1 none 25 acc 0.2688 ± 0.0130
none 25 acc_norm 0.2969 ± 0.0134

Usage Rights

Make sure to read Qwen's license before using this model.

Downloads last month
48
GGUF
Model size
620M params
Architecture
qwen2

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for afrideva/tau-0.5B-GGUF

Base model

M4-ai/tau-0.5B
Quantized
(4)
this model

Dataset used to train afrideva/tau-0.5B-GGUF