--- license: llama3.2 language: - zh - en - it - de - fr - ja - ko base_model: lianghsun/Llama-3.2-Taiwan-3B-Instruct datasets: - lianghsun/tw-emergency-medicine-bench - lianghsun/tw-legal-nlp - lianghsun/tw-legal-synthetic-qa - lianghsun/tw-law-article-qa - lianghsun/tw-judgment-qa - lianghsun/tw-judgment-gist-chat - lianghsun/tw-bar-examination-2020-chat - lianghsun/tw-structured-law-article - lianghsun/tw-judgment-gist-chat - lianghsun/tw-contract-review-chat - lianghsun/reasoning-base-20k-chat - lianghsun/vulnerability-mitigation-qa-zh_tw - lianghsun/tw-instruct - rombodawg/Everything_Instruct_Multilingual - xzuyn/manythings-translations-alpaca - neural-bridge/rag-dataset-12000 - minyichen/glaive_toolcall_zh_tw pipeline_tag: text-generation library_name: transformers tags: - Taiwan - ROC - zh-tw - instruct - chat - llama3.2 - SLM - llama-cpp - gguf-my-repo widget: - text: 中華民國憲法第一條 metrics: - accuracy model-index: - name: Llama-3.2-Taiwan-3B-Instruct results: - task: type: text-generation name: Single Choice Question dataset: name: tw-legal-benchmark-v1 type: lianghsun/tw-legal-benchmark-v1 metrics: - type: accuracy value: 31.1 name: single choice - task: type: text-generation name: Single Choice Question dataset: name: (Society) Formosa Taiwan Knowledge Bench type: lianghsun/Formosa-bench config: society split: test revision: v2024.11.27 metrics: - type: accuracy value: 60.42 name: single choice - task: type: text-generation name: Single Choice Question dataset: name: (Governmnt) Formosa Taiwan Knowledge Bench type: lianghsun/Formosa-bench config: governmnt split: test revision: v2024.11.27 metrics: - type: accuracy value: 44.25 name: single choice - task: type: text-generation name: Single Choice Question dataset: name: (Geography) Formosa Taiwan Knowledge Bench type: lianghsun/Formosa-bench config: geography split: test revision: v2024.11.27 metrics: - type: accuracy value: 47.54 name: single choice - task: type: text-generation name: Single Choice Question dataset: name: (History) Formosa Taiwan Knowledge Bench type: lianghsun/Formosa-bench config: history split: test revision: v2024.11.27 metrics: - type: accuracy value: 60 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (geography_of_taiwan) tmmlu++ type: ikala/tmmluplus config: geography_of_taiwan split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 36.2 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (dentistry) tmmlu++ type: ikala/tmmluplus config: dentistry split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 33.83 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (technical) tmmlu++ type: ikala/tmmluplus config: technical split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 35.07 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (statistics_and_machine_learning) tmmlu++ type: ikala/tmmluplus config: statistics_and_machine_learning split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 28.57 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (clinical_psychology) tmmlu++ type: ikala/tmmluplus config: clinical_psychology split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 29.6 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (tve_design) tmmlu++ type: ikala/tmmluplus config: tve_design split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 38.54 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (three_principles_of_people) tmmlu++ type: ikala/tmmluplus config: three_principles_of_people split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 48.2 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (introduction_to_law) tmmlu++ type: ikala/tmmluplus config: introduction_to_law split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 29.96 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (linear_algebra) tmmlu++ type: ikala/tmmluplus config: linear_algebra split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 21.43 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (agriculture) tmmlu++ type: ikala/tmmluplus config: agriculture split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 24.5 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (jce_humanities) tmmlu++ type: ikala/tmmluplus config: jce_humanities split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 38.89 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (music) tmmlu++ type: ikala/tmmluplus config: music split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.9 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (secondary_physics) tmmlu++ type: ikala/tmmluplus config: secondary_physics split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 33.04 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (physics) tmmlu++ type: ikala/tmmluplus config: physics split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.84 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (advance_chemistry) tmmlu++ type: ikala/tmmluplus config: advance_chemistry split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.64 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (junior_science_exam) tmmlu++ type: ikala/tmmluplus config: junior_science_exam split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.05 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (veterinary_pathology) tmmlu++ type: ikala/tmmluplus config: veterinary_pathology split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.09 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (financial_analysis) tmmlu++ type: ikala/tmmluplus config: financial_analysis split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.13 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (national_protection) tmmlu++ type: ikala/tmmluplus config: national_protection split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 42.65 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (macroeconomics) tmmlu++ type: ikala/tmmluplus config: macroeconomics split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 26.76 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (politic_science) tmmlu++ type: ikala/tmmluplus config: politic_science split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.44 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (ttqav2) tmmlu++ type: ikala/tmmluplus config: ttqav2 split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 61.06 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (junior_chinese_exam) tmmlu++ type: ikala/tmmluplus config: junior_chinese_exam split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.86 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (traditional_chinese_medicine_clinical_medicine) tmmlu++ type: ikala/tmmluplus config: traditional_chinese_medicine_clinical_medicine split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.9 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (junior_math_exam) tmmlu++ type: ikala/tmmluplus config: junior_math_exam split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 21.71 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (auditing) tmmlu++ type: ikala/tmmluplus config: auditing split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 21.82 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (anti_money_laundering) tmmlu++ type: ikala/tmmluplus config: anti_money_laundering split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 37.31 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (pharmacology) tmmlu++ type: ikala/tmmluplus config: pharmacology split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.68 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (trust_practice) tmmlu++ type: ikala/tmmluplus config: trust_practice split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 28.18 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (tve_mathematics) tmmlu++ type: ikala/tmmluplus config: tve_mathematics split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 18.67 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (human_behavior) tmmlu++ type: ikala/tmmluplus config: human_behavior split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 32.04 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (pharmacy) tmmlu++ type: ikala/tmmluplus config: pharmacy split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 22.76 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (tve_chinese_language) tmmlu++ type: ikala/tmmluplus config: tve_chinese_language split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 36.65 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (optometry) tmmlu++ type: ikala/tmmluplus config: optometry split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.11 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (physical_education) tmmlu++ type: ikala/tmmluplus config: physical_education split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.73 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (organic_chemistry) tmmlu++ type: ikala/tmmluplus config: organic_chemistry split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 35.78 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (tve_natural_sciences) tmmlu++ type: ikala/tmmluplus config: tve_natural_sciences split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 33.73 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (education) tmmlu++ type: ikala/tmmluplus config: education split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 37.9 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (mechanical) tmmlu++ type: ikala/tmmluplus config: mechanical split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 42.37 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (taiwanese_hokkien) tmmlu++ type: ikala/tmmluplus config: taiwanese_hokkien split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 14.73 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (nautical_science) tmmlu++ type: ikala/tmmluplus config: nautical_science split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.49 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (business_management) tmmlu++ type: ikala/tmmluplus config: business_management split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 39.57 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (logic_reasoning) tmmlu++ type: ikala/tmmluplus config: logic_reasoning split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.34 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (marketing_management) tmmlu++ type: ikala/tmmluplus config: marketing_management split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 39.78 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (economics) tmmlu++ type: ikala/tmmluplus config: economics split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.95 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (basic_medical_science) tmmlu++ type: ikala/tmmluplus config: basic_medical_science split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 28.41 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (occupational_therapy_for_psychological_disorders) tmmlu++ type: ikala/tmmluplus config: occupational_therapy_for_psychological_disorders split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 35.73 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (general_principles_of_law) tmmlu++ type: ikala/tmmluplus config: general_principles_of_law split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 31.13 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (junior_chemistry) tmmlu++ type: ikala/tmmluplus config: junior_chemistry split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 24.88 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (veterinary_pharmacology) tmmlu++ type: ikala/tmmluplus config: veterinary_pharmacology split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 36.3 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (educational_psychology) tmmlu++ type: ikala/tmmluplus config: educational_psychology split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 33.52 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (finance_banking) tmmlu++ type: ikala/tmmluplus config: finance_banking split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 32.59 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (official_document_management) tmmlu++ type: ikala/tmmluplus config: official_document_management split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 32.43 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (fire_science) tmmlu++ type: ikala/tmmluplus config: fire_science split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 30.65 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (junior_social_studies) tmmlu++ type: ikala/tmmluplus config: junior_social_studies split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 47.62 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (accounting) tmmlu++ type: ikala/tmmluplus config: accounting split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 20.94 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (engineering_math) tmmlu++ type: ikala/tmmluplus config: engineering_math split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.18 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (education_(profession_level)) tmmlu++ type: ikala/tmmluplus config: education_(profession_level) split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 24.07 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (chinese_language_and_literature) tmmlu++ type: ikala/tmmluplus config: chinese_language_and_literature split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.64 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (management_accounting) tmmlu++ type: ikala/tmmluplus config: management_accounting split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 24.19 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (culinary_skills) tmmlu++ type: ikala/tmmluplus config: culinary_skills split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 39.38 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (administrative_law) tmmlu++ type: ikala/tmmluplus config: administrative_law split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 25.71 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (insurance_studies) tmmlu++ type: ikala/tmmluplus config: insurance_studies split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 33.42 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (real_estate) tmmlu++ type: ikala/tmmluplus config: real_estate split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 22.83 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (computer_science) tmmlu++ type: ikala/tmmluplus config: computer_science split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 31.61 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (taxation) tmmlu++ type: ikala/tmmluplus config: taxation split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 27.47 name: single choice - task: type: question-answering name: Single Choice Question dataset: name: (trade) tmmlu++ type: ikala/tmmluplus config: trade split: test revision: c0e8ae955997300d5dbf0e382bf0ba5115f85e8c metrics: - type: accuracy value: 20.32 name: single choice --- # itlwas/Llama-3.2-Taiwan-3B-Instruct-Q4_K_M-GGUF This model was converted to GGUF format from [`lianghsun/Llama-3.2-Taiwan-3B-Instruct`](https://huggingface.co./lianghsun/Llama-3.2-Taiwan-3B-Instruct) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co./spaces/ggml-org/gguf-my-repo) space. Refer to the [original model card](https://huggingface.co./lianghsun/Llama-3.2-Taiwan-3B-Instruct) for more details on the model. ## Use with llama.cpp Install llama.cpp through brew (works on Mac and Linux) ```bash brew install llama.cpp ``` Invoke the llama.cpp server or the CLI. ### CLI: ```bash llama-cli --hf-repo itlwas/Llama-3.2-Taiwan-3B-Instruct-Q4_K_M-GGUF --hf-file llama-3.2-taiwan-3b-instruct-q4_k_m.gguf -p "The meaning to life and the universe is" ``` ### Server: ```bash llama-server --hf-repo itlwas/Llama-3.2-Taiwan-3B-Instruct-Q4_K_M-GGUF --hf-file llama-3.2-taiwan-3b-instruct-q4_k_m.gguf -c 2048 ``` Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well. Step 1: Clone llama.cpp from GitHub. ``` git clone https://github.com/ggerganov/llama.cpp ``` Step 2: Move into the llama.cpp folder and build it with `LLAMA_CURL=1` flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux). ``` cd llama.cpp && LLAMA_CURL=1 make ``` Step 3: Run inference through the main binary. ``` ./llama-cli --hf-repo itlwas/Llama-3.2-Taiwan-3B-Instruct-Q4_K_M-GGUF --hf-file llama-3.2-taiwan-3b-instruct-q4_k_m.gguf -p "The meaning to life and the universe is" ``` or ``` ./llama-server --hf-repo itlwas/Llama-3.2-Taiwan-3B-Instruct-Q4_K_M-GGUF --hf-file llama-3.2-taiwan-3b-instruct-q4_k_m.gguf -c 2048 ```