Zach996/Task1-only

1---2language:3- en4- de5- fr6- it7- pt8- hi9- es10- th11library_name: transformers12pipeline_tag: image-text-to-text13tags:14- facebook15- meta16- pytorch17- llama18- llama-319widget:20  - example_title: rococo art21    messages:22      - role: user23        content:24          - type: text25            text: Which era does this piece belong to? Give details about the era.26          - type: image_url27            image_url:28              url: https://huggingface.co/datasets/huggingface/release-assets/resolve/main/rococo.jpg29      - role: assistant30        content: |31          The piece is a painting of a woman in a red dress, surrounded by gold and white ornate details. The woman is depicted in mid-air, with her arms outstretched and her legs bent at the knees. She is holding a bouquet of flowers in her right hand and a fruit in her left hand.32          33          The background of the painting is a light blue sky with pink clouds, and there are also some pink flowers and green leaves surrounding the woman. The overall atmosphere of the painting is one of joy and celebration, as if the woman is dancing or celebrating something.34          35          This piece belongs to the Rococo era, which was a style of art and architecture that emerged in Europe in the 18th century. The Rococo style is characterized by its use of pastel colors, curved lines, and ornate details. It was popularized during the reign of King Louis XV of France, who ruled from 1715 to 1774.36  - example_title: diagram understanding37    messages:38      - role: user39        content:40          - type: text41            text: Where do the droughts happen according to this diagram?42          - type: image_url43            image_url:44              url: https://huggingface.co/datasets/huggingface/release-assets/resolve/main/weather_events.png45      - role: assistant46        content: |47          # Drought Regions in 201648          49          According to the diagram, the regions that experienced severe droughts in 2016 are:50          51          - **Eastern and Southern Africa**: This region was severely affected by droughts, with significant impacts on agriculture and water resources.52          - **Western and Southern Africa**: Droughts also occurred in this region, leading to food insecurity and other challenges.53          54          These regions were particularly vulnerable to droughts in 2016, highlighting the need for effective water management strategies and climate resilience measures.55  - example_title: document qa56    messages:57      - role: user58        content:59          - type: text60            text: How long does it take from invoice date to due date? Be short and concise.61          - type: image_url62            image_url:63              url: https://huggingface.co/datasets/huggingface/release-assets/resolve/main/invoice.png64      - role: assistant65        content: |66          To determine the time gap between the invoice date and the due date, we need to calculate the difference between these two dates.67          68          Invoice Date: February 11, 201969          Due Date: February 26, 201970          71          Step 1: Calculate the Difference in Days72          73          February 26 - February 11 = 15 days74          75          Therefore, the time gap between the invoice date and the due date is 15 days.76license: llama3.277extra_gated_prompt: >-78  ### LLAMA 3.2 COMMUNITY LICENSE AGREEMENT79 80 81  Llama 3.2 Version Release Date: September 25, 202482 83  84  “Agreement” means the terms and conditions for use, reproduction, distribution 85  and modification of the Llama Materials set forth herein.86 87  88  “Documentation” means the specifications, manuals and documentation accompanying Llama 3.289  distributed by Meta at https://llama.meta.com/doc/overview.90 91  92  “Licensee” or “you” means you, or your employer or any other person or entity (if you are 93  entering into this Agreement on such person or entity’s behalf), of the age required under94  applicable laws, rules or regulations to provide legal consent and that has legal authority95  to bind your employer or such other person or entity if you are entering in this Agreement96  on their behalf.97 98  99  “Llama 3.2” means the foundational large language models and software and algorithms, including100  machine-learning model code, trained model weights, inference-enabling code, training-enabling code,101  fine-tuning enabling code and other elements of the foregoing distributed by Meta at 102  https://www.llama.com/llama-downloads.103 104  105  “Llama Materials” means, collectively, Meta’s proprietary Llama 3.2 and Documentation (and 106  any portion thereof) made available under this Agreement.107 108  109  “Meta” or “we” means Meta Platforms Ireland Limited (if you are located in or, 110  if you are an entity, your principal place of business is in the EEA or Switzerland) 111  and Meta Platforms, Inc. (if you are located outside of the EEA or Switzerland). 112 113 114  By clicking “I Accept” below or by using or distributing any portion or element of the Llama Materials,115  you agree to be bound by this Agreement.116 117  118  1. License Rights and Redistribution.119  120  a. Grant of Rights. You are granted a non-exclusive, worldwide, 121  non-transferable and royalty-free limited license under Meta’s intellectual property or other rights 122  owned by Meta embodied in the Llama Materials to use, reproduce, distribute, copy, create derivative works 123  of, and make modifications to the Llama Materials.  124 125  b. Redistribution and Use.  126 127  i. If you distribute or make available the Llama Materials (or any derivative works thereof), 128  or a product or service (including another AI model) that contains any of them, you shall (A) provide129  a copy of this Agreement with any such Llama Materials; and (B) prominently display “Built with Llama”130  on a related website, user interface, blogpost, about page, or product documentation. If you use the131  Llama Materials or any outputs or results of the Llama Materials to create, train, fine tune, or132  otherwise improve an AI model, which is distributed or made available, you shall also include “Llama”133  at the beginning of any such AI model name.134 135  ii. If you receive Llama Materials, or any derivative works thereof, from a Licensee as part136  of an integrated end user product, then Section 2 of this Agreement will not apply to you. 137 138  iii. You must retain in all copies of the Llama Materials that you distribute the 139  following attribution notice within a “Notice” text file distributed as a part of such copies: 140  “Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms,141  Inc. All Rights Reserved.”142 143  iv. Your use of the Llama Materials must comply with applicable laws and regulations144  (including trade compliance laws and regulations) and adhere to the Acceptable Use Policy for145  the Llama Materials (available at https://www.llama.com/llama3_2/use-policy), which is hereby 146  incorporated by reference into this Agreement.147    148  2. Additional Commercial Terms. If, on the Llama 3.2 version release date, the monthly active users149  of the products or services made available by or for Licensee, or Licensee’s affiliates, 150  is greater than 700 million monthly active users in the preceding calendar month, you must request 151  a license from Meta, which Meta may grant to you in its sole discretion, and you are not authorized to152  exercise any of the rights under this Agreement unless or until Meta otherwise expressly grants you such rights.153  154  3. Disclaimer of Warranty. UNLESS REQUIRED BY APPLICABLE LAW, THE LLAMA MATERIALS AND ANY OUTPUT AND 155  RESULTS THEREFROM ARE PROVIDED ON AN “AS IS” BASIS, WITHOUT WARRANTIES OF ANY KIND, AND META DISCLAIMS156  ALL WARRANTIES OF ANY KIND, BOTH EXPRESS AND IMPLIED, INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES157  OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. YOU ARE SOLELY RESPONSIBLE158  FOR DETERMINING THE APPROPRIATENESS OF USING OR REDISTRIBUTING THE LLAMA MATERIALS AND ASSUME ANY RISKS ASSOCIATED159  WITH YOUR USE OF THE LLAMA MATERIALS AND ANY OUTPUT AND RESULTS.160  161  4. Limitation of Liability. IN NO EVENT WILL META OR ITS AFFILIATES BE LIABLE UNDER ANY THEORY OF LIABILITY, 162  WHETHER IN CONTRACT, TORT, NEGLIGENCE, PRODUCTS LIABILITY, OR OTHERWISE, ARISING OUT OF THIS AGREEMENT, 163  FOR ANY LOST PROFITS OR ANY INDIRECT, SPECIAL, CONSEQUENTIAL, INCIDENTAL, EXEMPLARY OR PUNITIVE DAMAGES, EVEN 164  IF META OR ITS AFFILIATES HAVE BEEN ADVISED OF THE POSSIBILITY OF ANY OF THE FOREGOING.165  166  5. Intellectual Property.167  168  a. No trademark licenses are granted under this Agreement, and in connection with the Llama Materials, 169  neither Meta nor Licensee may use any name or mark owned by or associated with the other or any of its affiliates, 170  except as required for reasonable and customary use in describing and redistributing the Llama Materials or as 171  set forth in this Section 5(a). Meta hereby grants you a license to use “Llama” (the “Mark”) solely as required 172  to comply with the last sentence of Section 1.b.i. You will comply with Meta’s brand guidelines (currently accessible 173  at https://about.meta.com/brand/resources/meta/company-brand/). All goodwill arising out of your use of the Mark 174  will inure to the benefit of Meta.175  176  b. Subject to Meta’s ownership of Llama Materials and derivatives made by or for Meta, with respect to any177  derivative works and modifications of the Llama Materials that are made by you, as between you and Meta,178  you are and will be the owner of such derivative works and modifications.179 180  c. If you institute litigation or other proceedings against Meta or any entity (including a cross-claim or181  counterclaim in a lawsuit) alleging that the Llama Materials or Llama 3.2 outputs or results, or any portion182  of any of the foregoing, constitutes infringement of intellectual property or other rights owned or licensable183  by you, then any licenses granted to you under this Agreement shall terminate as of the date such litigation or184  claim is filed or instituted. You will indemnify and hold harmless Meta from and against any claim by any third185  party arising out of or related to your use or distribution of the Llama Materials.186  187  6. Term and Termination. The term of this Agreement will commence upon your acceptance of this Agreement or access188  to the Llama Materials and will continue in full force and effect until terminated in accordance with the terms189  and conditions herein. Meta may terminate this Agreement if you are in breach of any term or condition of this190  Agreement. Upon termination of this Agreement, you shall delete and cease use of the Llama Materials. Sections 3,191  4 and 7 shall survive the termination of this Agreement. 192  193  7. Governing Law and Jurisdiction. This Agreement will be governed and construed under the laws of the State of 194  California without regard to choice of law principles, and the UN Convention on Contracts for the International195  Sale of Goods does not apply to this Agreement. The courts of California shall have exclusive jurisdiction of196  any dispute arising out of this Agreement. 197  198  ### Llama 3.2 Acceptable Use Policy199  200  Meta is committed to promoting safe and fair use of its tools and features, including Llama 3.2. 201  If you access or use Llama 3.2, you agree to this Acceptable Use Policy (“**Policy**”). 202  The most recent copy of this policy can be found at203  [https://www.llama.com/llama3_2/use-policy](https://www.llama.com/llama3_2/use-policy).204  205  #### Prohibited Uses206  207  We want everyone to use Llama 3.2 safely and responsibly. You agree you will not use, or allow others to use, Llama 3.2 to:208  209  1. Violate the law or others’ rights, including to:210      1. Engage in, promote, generate, contribute to, encourage, plan, incite, or further illegal or unlawful activity or content, such as:211          1. Violence or terrorism212          2. Exploitation or harm to children, including the solicitation, creation, acquisition, or dissemination of child exploitative content or failure to report Child Sexual Abuse Material213          3. Human trafficking, exploitation, and sexual violence214          4. The illegal distribution of information or materials to minors, including obscene materials, or failure to employ legally required age-gating in connection with such information or materials.215          5. Sexual solicitation216          6. Any other criminal activity217      1. Engage in, promote, incite, or facilitate the harassment, abuse, threatening, or bullying of individuals or groups of individuals218      2. Engage in, promote, incite, or facilitate discrimination or other unlawful or harmful conduct in the provision of employment, employment benefits, credit, housing, other economic benefits, or other essential goods and services219      3. Engage in the unauthorized or unlicensed practice of any profession including, but not limited to, financial, legal, medical/health, or related professional practices220      4. Collect, process, disclose, generate, or infer private or sensitive information about individuals, including information about individuals’ identity, health, or demographic information, unless you have obtained the right to do so in accordance with applicable law221      5. Engage in or facilitate any action or generate any content that infringes, misappropriates, or otherwise violates any third-party rights, including the outputs or results of any products or services using the Llama Materials222      6. Create, generate, or facilitate the creation of malicious code, malware, computer viruses or do anything else that could disable, overburden, interfere with or impair the proper working, integrity, operation or appearance of a website or computer system223      7. Engage in any action, or facilitate any action, to intentionally circumvent or remove usage restrictions or other safety measures, or to enable functionality disabled by Meta 224  2. Engage in, promote, incite, facilitate, or assist in the planning or development of activities that present a risk of death or bodily harm to individuals, including use of Llama 3.2 related to the following:225      8. Military, warfare, nuclear industries or applications, espionage, use for materials or activities that are subject to the International Traffic Arms Regulations (ITAR) maintained by the United States Department of State or to the U.S. Biological Weapons Anti-Terrorism Act of 1989 or the Chemical Weapons Convention Implementation Act of 1997226      9. Guns and illegal weapons (including weapon development)227      10. Illegal drugs and regulated/controlled substances228      11. Operation of critical infrastructure, transportation technologies, or heavy machinery229      12. Self-harm or harm to others, including suicide, cutting, and eating disorders230      13. Any content intended to incite or promote violence, abuse, or any infliction of bodily harm to an individual231  3. Intentionally deceive or mislead others, including use of Llama 3.2 related to the following:232      14. Generating, promoting, or furthering fraud or the creation or promotion of disinformation233      15. Generating, promoting, or furthering defamatory content, including the creation of defamatory statements, images, or other content234      16. Generating, promoting, or further distributing spam235      17. Impersonating another individual without consent, authorization, or legal right236      18. Representing that the use of Llama 3.2 or outputs are human-generated237      19. Generating or facilitating false online engagement, including fake reviews and other means of fake online engagement 238  4. Fail to appropriately disclose to end users any known dangers of your AI system239  5. Interact with third party tools, models, or software designed to generate unlawful content or engage in unlawful or harmful conduct and/or represent that the outputs of such tools, models, or software are associated with Meta or Llama 3.2240 241 242  With respect to any multimodal models included in Llama 3.2, the rights granted under Section 1(a) of the Llama 3.2 Community License Agreement are not being granted to you if you are an individual domiciled in, or a company with a principal place of business in, the European Union. This restriction does not apply to end users of a product or service that incorporates any such multimodal models.243 244 245  Please report any violation of this Policy, software “bug,” or other problems that could lead to a violation of this Policy through one of the following means:246 247 248  * Reporting issues with the model: [https://github.com/meta-llama/llama-models/issues](https://l.workplace.com/l.php?u=https%3A%2F%2Fgithub.com%2Fmeta-llama%2Fllama-models%2Fissues&h=AT0qV8W9BFT6NwihiOHRuKYQM_UnkzN_NmHMy91OT55gkLpgi4kQupHUl0ssR4dQsIQ8n3tfd0vtkobvsEvt1l4Ic6GXI2EeuHV8N08OG2WnbAmm0FL4ObkazC6G_256vN0lN9DsykCvCqGZ)249  250  * Reporting risky content generated by the model: [developers.facebook.com/llama_output_feedback](http://developers.facebook.com/llama_output_feedback)251  252  * Reporting bugs and security concerns: [facebook.com/whitehat/info](http://facebook.com/whitehat/info)253  254  * Reporting violations of the Acceptable Use Policy or unlicensed uses of Llama 3.2: LlamaUseReport@meta.com255extra_gated_fields:256  First Name: text257  Last Name: text258  Date of birth: date_picker259  Country: country260  Affiliation: text261  Job title:262    type: select263    options:264    - Student265    - Research Graduate266    - AI researcher267    - AI developer/engineer268    - Reporter269    - Other270  geo: ip_location271  By clicking Submit below I accept the terms of the license and acknowledge that the information I provide will be collected stored processed and shared in accordance with the Meta Privacy Policy: checkbox272extra_gated_description: >-273  The information you provide will be collected, stored, processed and shared in274  accordance with the [Meta Privacy275  Policy](https://www.facebook.com/privacy/policy/).276extra_gated_button_content: Submit277extra_gated_eu_disallowed: true278---279 280## Model Information281 282The Llama 3.2-Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes (text \+ images in / text out). The Llama 3.2-Vision instruction-tuned models are optimized for visual recognition, image reasoning, captioning, and answering general questions about an image. The models outperform many of the available open source and closed multimodal models on common industry benchmarks.283 284**Model Developer**: Meta285 286**Model Architecture:** Llama 3.2-Vision is built on top of Llama 3.1 text-only model, which is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety. To support image recognition tasks, the Llama 3.2-Vision model uses a separately trained vision adapter that integrates with the pre-trained Llama 3.1 language model. The adapter consists of a series of cross-attention layers that feed image encoder representations into the core LLM.287 288|  | Training Data | Params | Input modalities | Output modalities | Context length | GQA | Data volume | Knowledge cutoff |289| :---- | :---- | :---- | :---- | :---- | :---- | :---- | :---- | :---- |290| Llama 3.2-Vision  | (Image, text) pairs | 11B (10.6) | Text \+ Image | Text  | 128k | Yes | 6B (image, text) pairs | December 2023 |291| Llama 3.2-Vision | (Image, text) pairs | 90B (88.8) | Text \+ Image | Text  | 128k | Yes | 6B (image, text) pairs  | December 2023 |292 293**Supported Languages:** For text only tasks, English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai are officially supported. Llama 3.2 has been trained on a broader collection of languages than these 8 supported languages. Note for image+text applications, English is the only language supported. 294 295Developers may fine-tune Llama 3.2 models for languages beyond these supported languages, provided they comply with the Llama 3.2 Community License and the Acceptable Use Policy. Developers are always expected to ensure that their deployments, including those that involve additional languages, are completed safely and responsibly.296 297**Llama 3.2 Model Family:** Token counts refer to pretraining data only. All model versions use Grouped-Query Attention (GQA) for improved inference scalability.298 299**Model Release Date:** Sept 25, 2024300 301**Status:** This is a static model trained on an offline dataset. Future versions may be released that improve model capabilities and safety. 302 303**License:** Use of Llama 3.2 is governed by the [Llama 3.2 Community License](https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/LICENSE) (a custom, commercial license agreement).304 305**Feedback:** Where to send questions or comments about the model Instructions on how to provide feedback or comments on the model can be found in the model [README](https://github.com/meta-llama/llama-models/tree/main/models/llama3_2). For more technical information about generation parameters and recipes for how to use Llama 3.2-Vision in applications, please go [here](https://github.com/meta-llama/llama-recipes). 306 307## Intended Use308 309**Intended Use Cases:** Llama 3.2-Vision is intended for commercial and research use. Instruction tuned models are intended for visual recognition, image reasoning, captioning, and assistant-like chat with images, whereas pretrained models can be adapted for a variety of image reasoning tasks. Additionally, because of Llama 3.2-Vision’s ability to take images and text as inputs, additional use cases could include:310 3111. Visual Question Answering (VQA) and Visual Reasoning: Imagine a machine that looks at a picture and understands your questions about it.   3122. Document Visual Question Answering (DocVQA): Imagine a computer understanding both the text and layout of a document, like a map or contract, and then answering questions about it directly from the image.  3133. Image Captioning: Image captioning bridges the gap between vision and language, extracting details, understanding the scene, and then crafting a sentence or two that tells the story.  3144. Image-Text Retrieval: Image-text retrieval is like a matchmaker for images and their descriptions. Similar to a search engine but one that understands both pictures and words.  3155. Visual Grounding: Visual grounding is like connecting the dots between what we see and say. It’s about understanding how language references specific parts of an image, allowing AI models to pinpoint objects or regions based on natural language descriptions.  316   317 318The Llama 3.2 model collection also supports the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. The Llama 3.2 Community License allows for these use cases. 319 320**Out of Scope:** Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in any other way that is prohibited by the Acceptable Use Policy and Llama 3.2 Community License. Use in languages beyond those explicitly referenced as supported in this model card.321 322## How to use323 324This repository contains two versions of Llama-3.2-11B-Vision-Instruct, for use with transformers and with the original `llama` codebase.325 326### Use with transformers327 328Starting with transformers >= 4.45.0 onward, you can run inference using conversational messages that may include an image you can query about.329 330Make sure to update your transformers installation via `pip install --upgrade transformers`.331 332```python333import requests334import torch335from PIL import Image336from transformers import MllamaForConditionalGeneration, AutoProcessor337 338model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"339 340model = MllamaForConditionalGeneration.from_pretrained(341    model_id,342    torch_dtype=torch.bfloat16,343    device_map="auto",344)345processor = AutoProcessor.from_pretrained(model_id)346 347url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/0052a70beed5bf71b92610a43a52df6d286cd5f3/diffusers/rabbit.jpg"348image = Image.open(requests.get(url, stream=True).raw)349 350messages = [351    {"role": "user", "content": [352        {"type": "image"},353        {"type": "text", "text": "If I had to write a haiku for this one, it would be: "}354    ]}355]356input_text = processor.apply_chat_template(messages, add_generation_prompt=True)357inputs = processor(358    image,359    input_text,360    add_special_tokens=False,361    return_tensors="pt"362).to(model.device)363 364output = model.generate(**inputs, max_new_tokens=30)365print(processor.decode(output[0]))366```367 368### Use with `llama`369 370Please, follow the instructions in the [repository](https://github.com/meta-llama/llama).371 372To download the original checkpoints, you can use `huggingface-cli` as follows:373 374```375huggingface-cli download meta-llama/Llama-3.2-11B-Vision-Instruct --include "original/*" --local-dir Llama-3.2-11B-Vision-Instruct376```377 378## Hardware and Software379 380**Training Factors:** We used custom training libraries, Meta's custom built GPU cluster, and production infrastructure for pretraining. Fine-tuning, annotation, and evaluation were also performed on production infrastructure.381 382**Training Energy Use:** Training utilized a cumulative of **2.02M** GPU hours of computation on H100-80GB (TDP of 700W) type hardware, per the table below. Training time is the total GPU time required for training each model and power consumption is the peak power capacity per GPU device used, adjusted for power usage efficiency. 383 384## 385 386**Training Greenhouse Gas Emissions:** Estimated total location-based greenhouse gas emissions were **584** tons CO2eq for training. Since 2020, Meta has maintained net zero greenhouse gas emissions in its global operations and matched 100% of its electricity use with renewable energy, therefore the total market-based greenhouse gas emissions for training were 0 tons CO2eq.387 388|  | Training Time (GPU hours) | Training Power Consumption (W) | Training Location-Based Greenhouse Gas Emissions (tons CO2eq) | Training Market-Based Greenhouse Gas Emissions (tons CO2eq) |389| :---- | :---: | :---: | :---: | :---: |390| Llama 3.2-vision 11B | Stage 1 pretraining: 147K H100 hours Stage 2 annealing: 98K H100 hours SFT: 896 H100 hours RLHF: 224 H100 hours | 700 | 71 | 0 |391| Llama 3.2-vision 90B | Stage 1 pretraining: 885K H100 hours Stage 2 annealing: 885K H100 hours SFT: 3072 H100 hours RLHF: 2048 H100 hours | 700 | 513 | 0 |392| Total | 2.02M |  | 584 | 0 |393 394The methodology used to determine training energy use and greenhouse gas emissions can be found [here](https://arxiv.org/pdf/2204.05149).  Since Meta is openly releasing these models, the training energy use and greenhouse gas emissions will not be incurred by others.395 396## Training Data397 398**Overview:** Llama 3.2-Vision was pretrained on 6B image and text pairs. The instruction tuning data includes publicly available vision instruction datasets, as well as over 3M synthetically generated examples.399 400**Data Freshness:** The pretraining data has a cutoff of December 2023\.401 402## Benchmarks \- Image Reasoning403 404In this section, we report the results for Llama 3.2-Vision models on standard automatic benchmarks. For all these evaluations, we used our internal evaluations library.405 406### Base Pretrained Models407 408| Category | Benchmark | \# Shots | Metric | Llama 3.2 11B | Llama 3.2 90B |409| ----- | ----- | ----- | ----- | ----- | ----- |410| Image Understanding | VQAv2 (val) | 0 | Accuracy | 66.8 | 73.6 |411|  | Text VQA (val) | 0 | Relaxed accuracy | 73.1 | 73.5 |412|  | DocVQA (val, unseen) | 0 | ANLS | 62.3 | 70.7 |413| Visual Reasoning | MMMU (val, 0-shot) | 0 | Micro average accuracy | 41.7 | 49.3 |414|  | ChartQA (test) | 0 | Accuracy | 39.4 | 54.2 |415|  | InfographicsQA (val, unseen) | 0 | ANLS | 43.2 | 56.8 |416|  | AI2 Diagram (test) | 0 | Accuracy | 62.4 | 75.3 |417 418### Instruction Tuned Models419 420| Modality | Capability | Benchmark | \# Shots | Metric | Llama 3.2 11B | Llama 3.2 90B |421| ----- | :---: | ----- | :---: | :---: | ----- | ----- |422| Image | College-level Problems and Mathematical Reasoning | MMMU (val, CoT) | 0 | Micro average accuracy | 50.7 | 60.3 |423|  |  | MMMU-Pro, Standard (10 opts, test) | 0 | Accuracy | 33.0 | 45.2 |424|  |  | MMMU-Pro, Vision (test) | 0 | Accuracy | 23.7 | 33.8 |425|  |  | MathVista (testmini) | 0 | Accuracy | 51.5 | 57.3 |426|  | Charts and Diagram Understanding | ChartQA (test, CoT) | 0 | Relaxed accuracy | 83.4 | 85.5 |427|  |  | AI2 Diagram (test) | 0 | Accuracy | 91.1 | 92.3 |428|  |  | DocVQA (test) | 0 | ANLS | 88.4 | 90.1 |429|  | General Visual Question Answering | VQAv2 (test) | 0 | Accuracy | 75.2 | 78.1 |430|  |  |  |  |  |  |  |431| Text | General | MMLU (CoT) | 0 | Macro\_avg/acc | 73.0 | 86.0 |432|  | Math | MATH (CoT) | 0 | Final\_em | 51.9 | 68.0 |433|  | Reasoning | GPQA | 0 | Accuracy | 32.8 | 46.7 |434|  | Multilingual | MGSM (CoT) | 0 | em | 68.9 | 86.9 |435 436## Responsibility & Safety437 438As part of our Responsible release approach, we followed a three-pronged strategy to managing trust & safety risks:439 4401. Enable developers to deploy helpful, safe and flexible experiences for their target audience and for the use cases supported by Llama.   4412. Protect developers against adversarial users aiming to exploit Llama capabilities to potentially cause harm.  4423. Provide protections for the community to help prevent the misuse of our models.443 444### Responsible Deployment 445 446**Approach:** Llama is a foundational technology designed to be used in a variety of use cases, examples on how Meta’s Llama models have been responsibly deployed can be found in our [Community Stories webpage](https://llama.meta.com/community-stories/). Our approach is to build the most helpful models enabling the world to benefit from the technology power, by aligning our model safety for the generic use cases addressing a standard set of harms. Developers are then in the driver seat to tailor safety for their use case, defining their own policy and deploying the models with the necessary safeguards in their Llama systems. Llama 3.2 was developed following the best practices outlined in our Responsible Use Guide, you can refer to the [Responsible Use Guide](https://llama.meta.com/responsible-use-guide/) to learn more. 447 448#### Llama 3.2 Instruct 449 450**Objective:** Our main objectives for conducting safety fine-tuning are to provide the research community with a valuable resource for studying the robustness of safety fine-tuning, as well as to offer developers a readily available, safe, and powerful model for various applications to reduce the developer workload to deploy safe AI systems. We implemented the same set of safety mitigations as in Llama 3, and you can learn more about these in the Llama 3 [paper](https://ai.meta.com/research/publications/the-llama-3-herd-of-models/).451 452**Fine-Tuning Data:** We employ a multi-faceted approach to data collection, combining human-generated data from our vendors with synthetic data to mitigate potential safety risks. We’ve developed many large language model (LLM)-based classifiers that enable us to thoughtfully select high-quality prompts and responses, enhancing data quality control. 453 454**Refusals and Tone:** Building on the work we started with Llama 3, we put a great emphasis on model refusals to benign prompts as well as refusal tone. We included both borderline and adversarial prompts in our safety data strategy, and modified our safety data responses to follow tone guidelines.455 456#### Llama 3.2 Systems457 458**Safety as a System:** Large language models, including Llama 3.2, **are not designed to be deployed in isolation** but instead should be deployed as part of an overall AI system with additional safety guardrails as required. Developers are expected to deploy system safeguards when building agentic systems. Safeguards are key to achieve the right helpfulness-safety alignment as well as mitigating safety and security risks inherent to the system and any integration of the model or system with external tools. As part of our responsible release approach, we provide the community with [safeguards](https://llama.meta.com/trust-and-safety/) that developers should deploy with Llama models or other LLMs, including Llama Guard, Prompt Guard and Code Shield. All our [reference implementations](https://github.com/meta-llama/llama-agentic-system) demos contain these safeguards by default so developers can benefit from system-level safety out-of-the-box. 459 460### New Capabilities and Use Cases461 462**Technological Advancement:** Llama releases usually introduce new capabilities that require specific considerations in addition to the best practices that generally apply across all Generative AI use cases. For prior release capabilities also supported by Llama 3.2, see [Llama 3.1 Model Card](https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/MODEL_CARD.md), as the same considerations apply here as well., 463 464**Image Reasoning:** Llama 3.2-Vision models come with multimodal (text and image) input capabilities enabling image reasoning applications. As part of our responsible release process, we took dedicated measures including evaluations and mitigations to address the risk of the models uniquely identifying individuals in images. As with other LLM risks, models may not always be robust to adversarial prompts, and developers should evaluate identification and other applicable risks in the context of their applications as well as consider deploying Llama Guard 3-11B-Vision as part of their system or other mitigations as appropriate to detect and mitigate such risks.465 466### Evaluations467 468**Scaled Evaluations:** We built dedicated, adversarial evaluation datasets and evaluated systems composed of Llama models and Purple Llama safeguards to filter input prompt and output response. It is important to evaluate applications in context, and we recommend building dedicated evaluation dataset for your use case. 469 470**Red teaming:** We conducted recurring red teaming exercises with the goal of discovering risks via adversarial prompting and we used the learnings to improve our benchmarks and safety tuning datasets. We partnered early with subject-matter experts in critical risk areas to understand the nature of these real-world harms and how such models may lead to unintended harm for society. Based on these conversations, we derived a set of adversarial goals for the red team to attempt to achieve, such as extracting harmful information or reprogramming the model to act in a potentially harmful capacity. The red team consisted of experts in cybersecurity, adversarial machine learning, responsible AI, and integrity in addition to multilingual content specialists with background in integrity issues in specific geographic markets.471 472### Critical Risks 473 474In addition to our safety work above, we took extra care on measuring and/or mitigating the following critical risk areas:475 476**1\. CBRNE (Chemical, Biological, Radiological, Nuclear, and Explosive Weapons):** For Llama 3.1, to assess risks related to proliferation of chemical and biological weapons, we performed uplift testing designed to assess whether use of Llama 3.1 models could meaningfully increase the capabilities of malicious actors to plan or carry out attacks using these types of weapons. For Llama 3.2-Vision models, we conducted additional targeted evaluations and found that it was unlikely Llama 3.2 presented an increase in scientific capabilities due to its added image understanding capability as compared to Llama 3.1.477 478**2\. Child Safety:** Child Safety risk assessments were conducted using a team of experts, to assess the model’s capability to produce outputs that could result in Child Safety risks and inform on any necessary and appropriate risk mitigations via fine tuning. We leveraged those expert red teaming sessions to expand the coverage of our evaluation benchmarks through Llama 3 model development.  For Llama 3, we conducted new in-depth sessions using objective based methodologies to assess the model risks along multiple attack vectors including the additional languages Llama 3 is trained on. We also partnered with content specialists to perform red teaming exercises assessing potentially violating content while taking account of market specific nuances or experiences.  479 480**3\. Cyber Attacks:** For Llama 3.1 405B, our cyber attack uplift study investigated whether LLMs can enhance human capabilities in hacking tasks, both in terms of skill level and speed.481Our attack automation study focused on evaluating the capabilities of LLMs when used as autonomous agents in cyber offensive operations, specifically in the context of ransomware attacks. This evaluation was distinct from previous studies that considered LLMs as interactive assistants. The primary objective was to assess whether these models could effectively function as independent agents in executing complex cyber-attacks without human intervention. Because Llama 3.2’s vision capabilities are not generally germane to cyber uplift, we believe that the testing conducted for Llama 3.1 also applies to Llama 3.2.482 483### Community 484 485**Industry Partnerships:** Generative AI safety requires expertise and tooling, and we believe in the strength of the open community to accelerate its progress. We are active members of open consortiums, including the AI Alliance, Partnership on AI and MLCommons, actively contributing to safety standardization and transparency. We encourage the community to adopt taxonomies like the MLCommons Proof of Concept evaluation to facilitate collaboration and transparency on safety and content evaluations. Our Purple Llama tools are open sourced for the community to use and widely distributed across ecosystem partners including cloud service providers. We encourage community contributions to our [Github repository](https://github.com/meta-llama/PurpleLlama).486 487**Grants:** We also set up the [Llama Impact Grants](https://llama.meta.com/llama-impact-grants/) program to identify and support the most compelling applications of Meta’s Llama model for societal benefit across three categories: education, climate and open innovation. The 20 finalists from the hundreds of applications can be found [here](https://llama.meta.com/llama-impact-grants/#finalists). 488 489**Reporting:** Finally, we put in place a set of resources including an [output reporting mechanism](https://developers.facebook.com/llama_output_feedback) and [bug bounty program](https://www.facebook.com/whitehat) to continuously improve the Llama technology with the help of the community.490 491## Ethical Considerations and Limitations492 493**Values:** The core values of Llama 3.2 are openness, inclusivity and helpfulness. It is meant to serve everyone, and to work for a wide range of use cases. It is thus designed to be accessible to people across many different backgrounds, experiences and perspectives. Llama 3.2 addresses users and their needs as they are, without insertion unnecessary judgment or normativity, while reflecting the understanding that even content that may appear problematic in some cases can serve valuable purposes in others. It respects the dignity and autonomy of all users, especially in terms of the values of free thought and expression that power innovation and progress. 494 495**Testing:** But Llama 3.2 is a new technology, and like any new technology, there are risks associated with its use. Testing conducted to date has not covered, nor could it cover, all scenarios. For these reasons, as with all LLMs, Llama 3.2’s potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of Llama 3.2 models, developers should perform safety testing and tuning tailored to their specific applications of the model. Please refer to available resources including our [Responsible Use Guide](https://llama.meta.com/responsible-use-guide), [Trust and Safety](https://llama.meta.com/trust-and-safety/) solutions, and other [resources](https://llama.meta.com/docs/get-started/) to learn more about responsible development.   496