59:I[603186,["/_next/static/chunks/41aagivimsyys.js","/_next/static/chunks/1w6b9x4ox-2ab.js","/_next/static/chunks/2du4f4hdcx8vy.js","/_next/static/chunks/3lydxrl6slrf2.js","/_next/static/chunks/1la1pzs9j8idc.js","/_next/static/chunks/2hrea-dwwjjr_.js","/_next/static/chunks/28ougzph2g_t9.js","/_next/static/chunks/3r5yd4k4rms8-.js","/_next/static/chunks/125flzb3dd6q2.js","/_next/static/chunks/0wbn6hf2c6k2t.js","/_next/static/chunks/3pjc6r7w9qxan.js","/_next/static/chunks/39vtz3iby1cf_.js","/_next/static/chunks/2v5t_o330ua3s.js","/_next/static/chunks/3trxs7mom31q3.js","/_next/static/chunks/3nk0wrm9knx3m.js","/_next/static/chunks/1kxuebyfge68i.js","/_next/static/chunks/22e07luyn9bfp.js","/_next/static/chunks/1_lyn-lih218f.js","/_next/static/chunks/1dwajbnfjnvzx.js","/_next/static/chunks/0bvim9_g0hu68.js","/_next/static/chunks/3c7e5ojau5akn.js","/_next/static/chunks/0gf5i4fmb2vxz.js","/_next/static/chunks/3a57oltrmz5kh.js","/_next/static/chunks/3j9dwf14_qrhp.js","/_next/static/chunks/21k90yd6lrzsc.js","/_next/static/chunks/3-5mrnq0kjby0.js","/_next/static/chunks/1r40kzfox4d83.js","/_next/static/chunks/24_gg9g_4g22i.js","/_next/static/chunks/0l0c166aeafbb.js","/_next/static/chunks/0czpc-f8wjri8.js","/_next/static/chunks/3xaq4o65p_dw0.js","/_next/static/chunks/2dyt8rg18m736.js","/_next/static/chunks/43b9k5vjei3d1.js","/_next/static/chunks/3eatuedmz1636.js","/_next/static/chunks/2pqhypz2vo8mu.js","/_next/static/chunks/1nz6e-lpv7lb-.js","/_next/static/chunks/2l525havh4_yi.js","/_next/static/chunks/0fpx6rmsgi4um.js"],"PageMetadata"] 5a:I[831915,["/_next/static/chunks/41aagivimsyys.js","/_next/static/chunks/1w6b9x4ox-2ab.js","/_next/static/chunks/2du4f4hdcx8vy.js","/_next/static/chunks/3lydxrl6slrf2.js","/_next/static/chunks/1la1pzs9j8idc.js","/_next/static/chunks/2hrea-dwwjjr_.js","/_next/static/chunks/28ougzph2g_t9.js","/_next/static/chunks/3r5yd4k4rms8-.js","/_next/static/chunks/125flzb3dd6q2.js","/_next/static/chunks/0wbn6hf2c6k2t.js","/_next/static/chunks/3pjc6r7w9qxan.js","/_next/static/chunks/39vtz3iby1cf_.js","/_next/static/chunks/2v5t_o330ua3s.js","/_next/static/chunks/3trxs7mom31q3.js","/_next/static/chunks/3nk0wrm9knx3m.js","/_next/static/chunks/1kxuebyfge68i.js","/_next/static/chunks/22e07luyn9bfp.js","/_next/static/chunks/1_lyn-lih218f.js","/_next/static/chunks/1dwajbnfjnvzx.js","/_next/static/chunks/0bvim9_g0hu68.js","/_next/static/chunks/3c7e5ojau5akn.js","/_next/static/chunks/0gf5i4fmb2vxz.js","/_next/static/chunks/3a57oltrmz5kh.js","/_next/static/chunks/3j9dwf14_qrhp.js","/_next/static/chunks/21k90yd6lrzsc.js","/_next/static/chunks/3-5mrnq0kjby0.js","/_next/static/chunks/1r40kzfox4d83.js","/_next/static/chunks/24_gg9g_4g22i.js","/_next/static/chunks/0l0c166aeafbb.js","/_next/static/chunks/0czpc-f8wjri8.js","/_next/static/chunks/3xaq4o65p_dw0.js","/_next/static/chunks/2dyt8rg18m736.js","/_next/static/chunks/43b9k5vjei3d1.js","/_next/static/chunks/3eatuedmz1636.js","/_next/static/chunks/2pqhypz2vo8mu.js","/_next/static/chunks/1nz6e-lpv7lb-.js","/_next/static/chunks/2l525havh4_yi.js","/_next/static/chunks/0fpx6rmsgi4um.js"],"SearchResultsImage"] 5b:I[66106,["/_next/static/chunks/41aagivimsyys.js","/_next/static/chunks/1w6b9x4ox-2ab.js","/_next/static/chunks/2du4f4hdcx8vy.js","/_next/static/chunks/3lydxrl6slrf2.js","/_next/static/chunks/1la1pzs9j8idc.js","/_next/static/chunks/2hrea-dwwjjr_.js","/_next/static/chunks/28ougzph2g_t9.js","/_next/static/chunks/3r5yd4k4rms8-.js","/_next/static/chunks/125flzb3dd6q2.js","/_next/static/chunks/0wbn6hf2c6k2t.js","/_next/static/chunks/3pjc6r7w9qxan.js","/_next/static/chunks/39vtz3iby1cf_.js","/_next/static/chunks/2v5t_o330ua3s.js","/_next/static/chunks/3trxs7mom31q3.js","/_next/static/chunks/3nk0wrm9knx3m.js","/_next/static/chunks/1kxuebyfge68i.js","/_next/static/chunks/22e07luyn9bfp.js","/_next/static/chunks/1_lyn-lih218f.js","/_next/static/chunks/1dwajbnfjnvzx.js","/_next/static/chunks/0bvim9_g0hu68.js","/_next/static/chunks/3c7e5ojau5akn.js","/_next/static/chunks/0gf5i4fmb2vxz.js","/_next/static/chunks/3a57oltrmz5kh.js","/_next/static/chunks/3j9dwf14_qrhp.js","/_next/static/chunks/21k90yd6lrzsc.js","/_next/static/chunks/3-5mrnq0kjby0.js","/_next/static/chunks/1r40kzfox4d83.js","/_next/static/chunks/24_gg9g_4g22i.js","/_next/static/chunks/0l0c166aeafbb.js","/_next/static/chunks/0czpc-f8wjri8.js","/_next/static/chunks/3xaq4o65p_dw0.js","/_next/static/chunks/2dyt8rg18m736.js","/_next/static/chunks/43b9k5vjei3d1.js","/_next/static/chunks/3eatuedmz1636.js","/_next/static/chunks/2pqhypz2vo8mu.js","/_next/static/chunks/1nz6e-lpv7lb-.js","/_next/static/chunks/2l525havh4_yi.js","/_next/static/chunks/0fpx6rmsgi4um.js"],""] 5c:I[461459,["/_next/static/chunks/41aagivimsyys.js","/_next/static/chunks/1w6b9x4ox-2ab.js","/_next/static/chunks/2du4f4hdcx8vy.js","/_next/static/chunks/3lydxrl6slrf2.js","/_next/static/chunks/1la1pzs9j8idc.js","/_next/static/chunks/2hrea-dwwjjr_.js","/_next/static/chunks/28ougzph2g_t9.js","/_next/static/chunks/3r5yd4k4rms8-.js","/_next/static/chunks/125flzb3dd6q2.js","/_next/static/chunks/0wbn6hf2c6k2t.js","/_next/static/chunks/3pjc6r7w9qxan.js","/_next/static/chunks/39vtz3iby1cf_.js","/_next/static/chunks/2v5t_o330ua3s.js","/_next/static/chunks/3trxs7mom31q3.js","/_next/static/chunks/3nk0wrm9knx3m.js","/_next/static/chunks/1kxuebyfge68i.js","/_next/static/chunks/22e07luyn9bfp.js","/_next/static/chunks/1_lyn-lih218f.js","/_next/static/chunks/1dwajbnfjnvzx.js","/_next/static/chunks/0bvim9_g0hu68.js","/_next/static/chunks/3c7e5ojau5akn.js","/_next/static/chunks/0gf5i4fmb2vxz.js","/_next/static/chunks/3a57oltrmz5kh.js","/_next/static/chunks/3j9dwf14_qrhp.js","/_next/static/chunks/21k90yd6lrzsc.js","/_next/static/chunks/3-5mrnq0kjby0.js","/_next/static/chunks/1r40kzfox4d83.js","/_next/static/chunks/24_gg9g_4g22i.js","/_next/static/chunks/0l0c166aeafbb.js","/_next/static/chunks/0czpc-f8wjri8.js","/_next/static/chunks/3xaq4o65p_dw0.js","/_next/static/chunks/2dyt8rg18m736.js","/_next/static/chunks/43b9k5vjei3d1.js","/_next/static/chunks/3eatuedmz1636.js","/_next/static/chunks/2pqhypz2vo8mu.js","/_next/static/chunks/1nz6e-lpv7lb-.js","/_next/static/chunks/2l525havh4_yi.js","/_next/static/chunks/0fpx6rmsgi4um.js"],"ArtifactBadges"] 62:I[946424,["/_next/static/chunks/41aagivimsyys.js","/_next/static/chunks/1w6b9x4ox-2ab.js","/_next/static/chunks/2du4f4hdcx8vy.js","/_next/static/chunks/3lydxrl6slrf2.js","/_next/static/chunks/1la1pzs9j8idc.js","/_next/static/chunks/2hrea-dwwjjr_.js","/_next/static/chunks/28ougzph2g_t9.js","/_next/static/chunks/3r5yd4k4rms8-.js","/_next/static/chunks/125flzb3dd6q2.js","/_next/static/chunks/0wbn6hf2c6k2t.js","/_next/static/chunks/3pjc6r7w9qxan.js","/_next/static/chunks/39vtz3iby1cf_.js","/_next/static/chunks/2v5t_o330ua3s.js","/_next/static/chunks/3trxs7mom31q3.js","/_next/static/chunks/3nk0wrm9knx3m.js","/_next/static/chunks/1kxuebyfge68i.js","/_next/static/chunks/22e07luyn9bfp.js","/_next/static/chunks/1_lyn-lih218f.js","/_next/static/chunks/1dwajbnfjnvzx.js","/_next/static/chunks/0bvim9_g0hu68.js","/_next/static/chunks/3c7e5ojau5akn.js","/_next/static/chunks/0gf5i4fmb2vxz.js","/_next/static/chunks/3a57oltrmz5kh.js","/_next/static/chunks/3j9dwf14_qrhp.js","/_next/static/chunks/21k90yd6lrzsc.js","/_next/static/chunks/3-5mrnq0kjby0.js","/_next/static/chunks/1r40kzfox4d83.js","/_next/static/chunks/24_gg9g_4g22i.js","/_next/static/chunks/0l0c166aeafbb.js","/_next/static/chunks/0czpc-f8wjri8.js","/_next/static/chunks/3xaq4o65p_dw0.js","/_next/static/chunks/2dyt8rg18m736.js","/_next/static/chunks/43b9k5vjei3d1.js","/_next/static/chunks/3eatuedmz1636.js","/_next/static/chunks/2pqhypz2vo8mu.js","/_next/static/chunks/1nz6e-lpv7lb-.js","/_next/static/chunks/2l525havh4_yi.js","/_next/static/chunks/0fpx6rmsgi4um.js"],"default"] 56:["$","$L59",null,{"hasPartnerEndpoint":false}] 5d:T3ad9,### Model Summary | | | |------|------| | **Description** | Phi-3.5-vision is a lightweight, state-of-the-art open multimodal model built upon datasets which include - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data both on text and vision. The model belongs to the Phi-3 model family, and the multimodal version comes with 128K context length (in tokens) it can support. The model underwent a rigorous enhancement process, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures. This model is ready for commercial and research use. | **Release date** | August 20, 2024 | | **License** | [MIT](https://opensource.org/license/mit) | | **Architecture** | Phi-3.5-vision has 4.2B parameters and contains image encoder, connector, projector, and Phi-3 Mini language model. | | **Inputs** | Text and Image. It’s best suited for prompts using the chat format. | | **Context length** | 128K tokens | | **Outputs** | Generated text (string) in response to the input | | **Status** | This is a static model trained on an offline text dataset with cutoff date March 15, 2024. Future versions of the tuned models may be released as we improve models. | **Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.** ## Third-Party Community Consideration This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party’s requirements for this application and use case. ### Resources 🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3)
📰 [Phi-3 Microsoft Blog](https://aka.ms/phi3.5-techblog)
📖 [Phi-3 Technical Report](https://arxiv.org/abs/2404.14219)
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai)
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook)
## Release Notes In this release, the model enables multi-frame image understanding and reasoning which is based on valuable customer feedback. The hero example multi-frame capabilities include detailed image comparison, multi-image summarization/storytelling and video summarization, which have broad applications in many scenarios. We also observed performance improvement on most single image benchmarks, e.g., boosting MMMU performance from 40.2 to 43.0, MMBench performance from 80.5 to 81.9, document understanding benchmark TextVQA from 70.9 to 72.0. We believe most use cases will benefit from this release, but we encourage users to test the new model in their AI applications. We appreciate the enthusiastic adoption of the Phi-3 model family and continue to welcome all the feedback from the community. ## Intended Use ### Primary Use Cases The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications with visual and text input capabilities which require: 1) Memory/compute constrained environments 2) Latency bound scenarios 3) General image understanding 4) Optical character recognition 5) Chart and table understanding 6) Multiple image comparison 7) Multi-image or video clip summarization Our model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features. ### Out-of-Scope Use Cases Our models are not specifically designed or evaluated for all downstream purposes. Developers should consider common limitations of language models as they select use cases, and evaluate and mitigate for accuracy, safety, and fariness before using within a specific downstream use case, particularly for high risk scenarios. Developers should be aware of and adhere to applicable laws or regulations (including privacy, trade compliance laws, etc.) that are relevant to their use case. ## Usage ### Input Formats Given the nature of the training data, the Phi-3.5-vision model is best suited for prompts using the chat format as follows: Single image: ``` <|user|>\n<|image_1|>\n{prompt}<|end|>\n<|assistant|>\n ``` Multi-turn conversations: ``` <|user|>\n<|image_1|>\n{prompt_1}<|end|>\n<|assistant|>\n{response_1}<|end|>\n<|user|>\n{prompt_2}<|end|>\n<|assistant|>\n ``` For multi-image usage, add multiple image placeholders in the front of the prompts. <|image_{}|> index should start from 1. One example of prompt is shown as follows: ``` <|user|>\n<|image_1|>\n<|image_2|>\n<|image_3|>\n<|image_4|>\n{prompt}<|end|>\n<|assistant|>\n ``` ### Loading the model locally After obtaining the Phi-3.5-vision-instruct model checkpoints, users can use this sample code for inference. ```python from PIL import Image import requests from transformers import AutoModelForCausalLM from transformers import AutoProcessor model_id = "microsoft/Phi-3.5-vision-instruct" model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda", trust_remote_code=True, torch_dtype="auto", _attn_implementation='flash_attention_2') processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True, num_crops=4) images = [] placeholder = "" for i in range(1,20): url = f"https://image.slidesharecdn.com/azureintroduction-191206101932/75/Introduction-to-Microsoft-Azure-Cloud-{i}-2048.jpg" images.append(Image.open(requests.get(url, stream=True).raw)) placeholder += f"<|image_{i}|>\n" messages = [ {"role": "user", "content": placeholder+"Summarize the deck of slides."}, ] prompt = processor.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) inputs = processor(prompt, images, return_tensors="pt").to("cuda:0") generation_args = { "max_new_tokens": 1000, "temperature": 0.0, "do_sample": False, } generate_ids = model.generate(**inputs, eos_token_id=processor.tokenizer.eos_token_id, **generation_args) # remove input tokens generate_ids = generate_ids[:, inputs['input_ids'].shape[1]:] response = processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] print(response) ``` ## Model Version(s): [Use unique identifier for identifying the model in the future- this may be part of our internal naming, specifying variation like "pruned," "unpruned," "trained," or "deployable" or "quantized" where necessary and including a versioning number like vX.X along with short description differentiating if multiple versions are available]
## Datasets | | | |------|------| | **Dates** | Trained between July and August 2024 | | **Training time** | 6 days | | **Training data** | 500B tokens (vision tokens + text tokens) | Our training data includes a wide variety of sources, and is a combination of 1) publicly available documents filtered rigorously for quality, selected high-quality educational data and code; 2) selected high-quality image-text interleave data; 3) newly created synthetic, “textbook-like” data for the purpose of teaching math, coding, common sense reasoning, general knowledge of the world (science, daily activities, theory of mind, etc.), newly created image data, e.g., chart/table/diagram/slides, newly created multi-image and video data, e.g., short video clips/pair of two similar images; 4) high quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness. The data collection process involved sourcing information from publicly available documents, with a meticulous approach to filtering out undesirable documents and images. To safeguard privacy, we carefully filtered various image and text data sources to remove or scrub any potentially personal data from the training data. Below are the comparison results on existing multi-image benchmarks. On average, our model outperforms competitor models on the same size and competitive with much bigger models on multi-frame capabilities and video summarization. BLINK: a benchmark with 14 visual tasks that humans can solve very quickly but are still hard for current multimodal LLMs. | Benchmark | Phi-3.5-vision-instrust | LlaVA-Interleave-Qwen-7B | InternVL-2-4B | InternVL-2-8B | Gemini-1.5-Flash | GPT-4o-mini | Claude-3.5-Sonnet | Gemini-1.5-Pro | GPT-4o | |--|--|--|--|--|--|--|--|--|--| | Art Style | 87.2 | 62.4 | 55.6 | 52.1 | 64.1 | 70.1 | 59.8 | 70.9 | 73.3 | | Counting | 54.2 | 56.7 | 54.2 | 66.7 | 51.7 | 55.0 | 59.2 | 65.0 | 65.0 | | Forensic Detection | 92.4 | 31.1 | 40.9 | 34.1 | 54.5 | 38.6 | 67.4 | 60.6 | 75.8 | | Functional Correspondence | 29.2 | 34.6 | 24.6 | 24.6 | 33.1 | 26.9 | 33.8 | 31.5 | 43.8 | | IQ Test | 25.3 | 26.7 | 26.0 | 30.7 | 25.3 | 29.3 | 26.0 | 34.0 | 19.3 | | Jigsaw | 68.0 | 86.0 | 55.3 | 52.7 | 71.3 | 72.7 | 57.3 | 68.0 | 67.3 | | Multi-View Reasoning | 54.1 | 44.4 | 48.9 | 42.9 | 48.9 | 48.1 | 55.6 | 49.6 | 46.6 | | Object Localization | 49.2 | 54.9 | 53.3 | 54.1 | 57.3 | 57.4 | 62.3 | 65.6 | 68.0 | | Relative Depth | 69.4 | 77.4 | 63.7 | 67.7 | 32.8 | 58.1 | 71.8 | 76.6 | 71.0 | | Relative Reflectance | 37.3 | 34.3 | 32.8 | 38.8 | 32.8 | 27.6 | 36.6 | 38.8 | 40.3 | | Semantic Correspondence | 36.7 | 31.7 | 31.7 | 22.3 | 32.4 | 31.7 | 45.3 | 48.9 | 54.0 | | Spatial Relation | 65.7 | 75.5 | 78.3 | 78.3 | 55.9 | 81.1 | 60.1 | 79.0 | 84.6 | | Visual Correspondence | 53.5 | 40.7 | 34.9 | 33.1 | 29.7 | 52.9 | 72.1 | 81.4 | 86.0 | | Visual Similarity | 83.0 | 91.9 | 48.1 | 45.2 | 47.4 | 77.8 | 84.4 | 81.5 | 88.1 | | **Overall** | **57.0** | **53.1** | **45.9** | **45.4** | **45.1** | **51.9** | **56.5** | **61.0** | **63.2** | Video-MME: comprehensively assess the capabilities of MLLMs in processing video data, covering a wide range of visual domains, temporal durations, and data modalities. | Benchmark | Phi-3.5-vision-instrust | LlaVA-Interleave-Qwen-7B | InternVL-2-4B | InternVL-2-8B | Gemini-1.5-Flash | GPT-4o-mini | Claude-3.5-Sonnet | Gemini-1.5-Pro | GPT-4o | |--|--|--|--|--|--|--|--|--|--| | short (<2min) | 60.8 | 62.3 | 60.7 | 61.7 | 72.2 | 70.1 | 66.3 | 73.3 | 77.7 | | medium (4-15min) | 47.7 | 47.1 | 46.4 | 49.6 | 62.7 | 59.6 | 54.7 | 61.2 | 68.0 | | long (30-60) | 43.8 | 41.2 | 42.6 | 46.6 | 52.1 | 53.9 | 46.6 | 53.2 | 59.6 | | **Overall** | **50.8** | **50.2** | **49.9** | **52.6** | **62.3** | **61.2** | **55.9** | **62.6** | **68.4** | ## Inference: **Engine:** Tensor(RT)
**Test Hardware [Name the specific test hardware model]:**
* 256 A100-80G
## Responsible AI Considerations Like other models, the Phi family of models can potentially behave in ways that are unfair, unreliable, or offensive. Some of the limiting behaviors to be aware of include: * Quality of Service: The Phi models are trained primarily on English text. Languages other than English will experience worse performance. English language varieties with less representation in the training data might experience worse performance than standard American English. * Representation of Harms & Perpetuation of Stereotypes: These models can over- or under-represent groups of people, erase representation of some groups, or reinforce demeaning or negative stereotypes. Despite safety post-training, these limitations may still be present due to differing levels of representation of different groups or prevalence of examples of negative stereotypes in training data that reflect real-world patterns and societal biases. * Inappropriate or Offensive Content: These models may produce other types of inappropriate or offensive content, which may make it inappropriate to deploy for sensitive contexts without additional mitigations that are specific to the use case. * Information Reliability: Language models can generate nonsensical content or fabricate content that might sound reasonable but is inaccurate or outdated. * Limited Scope for Code: Majority of Phi-3 training data is based in Python and use common packages such as "typing, math, random, collections, datetime, itertools". If the model generates Python scripts that utilize other packages or scripts in other languages, we strongly recommend users manually verify all API uses. Developers should apply responsible AI best practices and are responsible for ensuring that a specific use case complies with relevant laws and regulations (e.g. privacy, trade, etc.). Important areas for consideration include: * Allocation: Models may not be suitable for scenarios that could have consequential impact on legal status or the allocation of resources or life opportunities (ex: housing, employment, credit, etc.) without further assessments and additional debiasing techniques. * High-Risk Scenarios: Developers should assess suitability of using models in high-risk scenarios where unfair, unreliable or offensive outputs might be extremely costly or lead to harm. This includes providing advice in sensitive or expert domains where accuracy and reliability are critical (ex: legal or health advice). Additional safeguards should be implemented at the application level according to the deployment context. * Misinformation: Models may produce inaccurate information. Developers should follow transparency best practices and inform end-users they are interacting with an AI system. At the application level, developers can build feedback mechanisms and pipelines to ground responses in use-case specific, contextual information, a technique known as Retrieval Augmented Generation (RAG). * Generation of Harmful Content: Developers should assess outputs for their context and use available safety classifiers or custom solutions appropriate for their use case. * Misuse: Other forms of misuse such as fraud, spam, or malware production may be possible, and developers should ensure that their applications do not violate applicable laws and regulations. * Identification of individuals: models with vision capabilities may have the potential to uniquely identify individuals in images. Safety post-training steers the model to refuse such requests, but developers should consider and implement, as appropriate, additional mitigations or user consent flows as required in their respective jurisdiction, (e.g., building measures to blur faces in image inputs before processing). ## Security and AI Concerns: Please report security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/). ## Trademarks This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow [Microsoft’s Trademark & Brand Guidelines](https://www.microsoft.com/en-us/legal/intellectualproperty/trademarks). Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party’s policies.57:[["$","div",".3",{"className":"relative","children":[["$","div",null,{"className":"relative flex min-h-[194px] w-full flex-col overflow-hidden p-md sm:p-ml bg-manitoulinLightBlack","children":[["$","$L5a",null,{"alt":"","className":"object-cover object-center !left-1/2 z-10 !w-1/2 brightness-20 lg:brightness-75","fill":true,"priority":true,"src":"https://assets.ngc.nvidia.com/products/api-catalog/images/phi-3_5-vision-instruct.jpg","style":{"maskImage":"linear-gradient(to left, rgba(0,0,0,1), rgba(0,0,0,0))"}}],["$","div",null,{"className":"z-30 mx-auto flex w-full max-w-screen-xl min-w-0 flex-1 flex-col items-start justify-between gap-md lg:flex-row lg:gap-ml","children":["$","section",null,{"aria-describedby":"resource-description","aria-labelledby":"resource-name","className":"flex w-full min-w-0 flex-col lg:flex-1","children":["$","div",null,{"className":"flex flex-col gap-1.5 whitespace-pre-wrap","children":[["$","$L5b",null,{"className":"w-fit text-ml text-n300 lowercase hover:text-n050","href":"/microsoft","children":"Microsoft"}],["$","div",null,{"className":"flex flex-col items-start gap-3 lg:flex-row lg:items-center","children":[["$","h1",null,{"className":"max-w-full text-lg leading-heading font-medium break-words text-primary","data-testid":"resource-name","id":"resource-name","children":"phi-3.5-vision-instruct"}],["$","$L5c",null,{"artifact":{"artifactType":"ENDPOINT","name":"phi-3_5-vision-instruct","displayName":"phi-3.5-vision-instruct","publisher":"microsoft","shortDescription":"Cutting-edge open multimodal model exceling in high-quality reasoning from images.","logo":"https://assets.ngc.nvidia.com/products/api-catalog/images/phi-3_5-vision-instruct.jpg","labels":["Language Generation","Vision Assistant","Visual Question Answering","nimType:endpoint:nim_type_preview","usecase:endpoint:usecase_image_to_text","usecase:endpoint:usecase_text_gen"],"attributes":[{"key":"DEPRECATION","value":"04/15/2026"},{"key":"PREVIEW","value":"true"}],"updatedDate":"2025-02-14T00:04:45.740Z","canGuestDownload":true,"createdDate":"2024-08-22T13:10:16.342Z","description":"$5d","isPublic":true,"isReadOnly":true,"orgName":"qc69jvmznzxy"}}],"$undefined"]}],"$L5e","$L5f"]}]}]}],"$L60"]}],"$L61"]}]] 63:T1d2a,<% // Catalog ejs locals replace JSON with {stringify} only (spread skips non-enumerable parse). const deepClone = (value) => { if (Array.isArray(value)) return value.map(deepClone); if (value && typeof value === "object") { const out = {}; for (const key of Object.keys(value)) out[key] = deepClone(value[key]); return out; } return value; }; const mediaConfig = {"image": {"mimeTypes": {}}, "video": {"mimeTypes": {}}, "audio": {"mimeTypes": {}, "defaultFormat": "wav"}}; const stripNulls = (value) => { if (Array.isArray(value)) return value.map(stripNulls); if (value && typeof value === "object") { const out = {}; for (const [key, item] of Object.entries(value)) { if (item === null || item === undefined) continue; out[key] = stripNulls(item); } return out; } return value; }; const payloadForSnippet = stripNulls(deepClone(request || {})); const hasStream = payloadForSnippet.stream !== undefined && payloadForSnippet.stream !== null; const pyStream = payloadForSnippet.stream ? "True" : "False"; if (hasStream) payloadForSnippet.stream = "__STREAM__"; const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0; const isInlineDataUrl = (url, kind) => ( typeof url === "string" && new RegExp(`^data:${kind}/[^;]+;base64,`, "i").test(url) ); const mediaUrl = (part, key) => { const slot = part?.[key]; return typeof slot === "string" ? slot : slot?.url; }; const setMediaUrl = (part, key, url) => { if (typeof part[key] === "string") part[key] = url; else { if (!part[key] || typeof part[key] !== "object") part[key] = {}; part[key].url = url; } }; let imageCount = 0; let videoCount = 0; let audioCount = 0; (payloadForSnippet.messages || []).forEach((message) => { if (!Array.isArray(message?.content)) return; message.content.forEach((part) => { const imageUrl = mediaUrl(part, "image_url"); const videoUrl = mediaUrl(part, "video_url"); const audioUrl = mediaUrl(part, "audio_url"); // File helpers only for inline base64; http(s) URLs stay as-is in the payload. if (hasMediaKind("image") && part?.type === "image_url" && isInlineDataUrl(imageUrl, "image")) { setMediaUrl(part, "image_url", `__IMAGE_DATA_URL_${imageCount}__`); imageCount += 1; } else if (hasMediaKind("video") && part?.type === "video_url" && isInlineDataUrl(videoUrl, "video")) { setMediaUrl(part, "video_url", `__VIDEO_DATA_URL_${videoCount}__`); videoCount += 1; } else if (hasMediaKind("audio") && part?.type === "audio_url" && isInlineDataUrl(audioUrl, "audio")) { setMediaUrl(part, "audio_url", `__AUDIO_DATA_URL_${audioCount}__`); audioCount += 1; } else if (hasMediaKind("audio") && part?.type === "input_audio" && part.input_audio?.data) { part.input_audio.data = `__AUDIO_B64_${audioCount}__`; part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat; audioCount += 1; } }); }); const pyRepr = (value, level = 0) => { if (value === null) return "None"; if (typeof value === "boolean") return value ? "True" : "False"; if (typeof value === "number") return String(value); if (typeof value === "string") return JSON.stringify(value); const pad = " ".repeat(level); const innerPad = " ".repeat(level + 2); if (Array.isArray(value)) { if (value.length === 0) return "[]"; return "[\n" + value.map((item) => `${innerPad}${pyRepr(item, level + 2)}`).join(",\n") + "\n" + pad + "]"; } if (typeof value === "object") { const entries = Object.entries(value); if (entries.length === 0) return "{}"; return "{\n" + entries.map(([key, item]) => `${innerPad}${JSON.stringify(key)}: ${pyRepr(item, level + 2)}`).join(",\n") + "\n" + pad + "}"; } return "None"; }; const replaceAll = (s, a, b) => s.split(a).join(b); let renderedPayload = pyRepr(payloadForSnippet); if (hasStream) renderedPayload = replaceAll(renderedPayload, '"__STREAM__"', "stream"); for (let i = 0; i < imageCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__IMAGE_DATA_URL_${i}__"`, `image_data_urls[${i}]`); } for (let i = 0; i < videoCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__VIDEO_DATA_URL_${i}__"`, `video_data_urls[${i}]`); } for (let i = 0; i < audioCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__AUDIO_DATA_URL_${i}__"`, `audio_data_urls[${i}]`); renderedPayload = replaceAll(renderedPayload, `"__AUDIO_B64_${i}__"`, `audio_b64s[${i}]`); } // Either file helpers XOR inline base64 - never emit unused helpers. const usesFileHelpers = imageCount + videoCount + audioCount > 0 && [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`image_data_urls[${i}]`)) && [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`video_data_urls[${i}]`)) && [...Array(audioCount).keys()].every( (i) => renderedPayload.includes(`audio_data_urls[${i}]`) || renderedPayload.includes(`audio_b64s[${i}]`) ); if (!usesFileHelpers) { imageCount = 0; videoCount = 0; audioCount = 0; const restored = stripNulls(deepClone(request || {})); const restoredStream = restored.stream !== undefined && restored.stream !== null; if (restoredStream) restored.stream = "__STREAM__"; renderedPayload = pyRepr(restored); if (restoredStream) renderedPayload = replaceAll(renderedPayload, '"__STREAM__"', "stream"); } const hasMedia = usesFileHelpers; const mimePy = (mimes) => "{" + Object.entries(mimes).map(([k, v]) => `${JSON.stringify(k)}: ${JSON.stringify(v)}`).join(", ") + "}"; %> <% if (hasMedia) { %>import base64 from pathlib import Path <% } %>import requests invoke_url = "https://integrate.api.nvidia.com/v1/chat/completions" stream = <%- pyStream %> <% if (hasMedia) { %> def media_data_url(kind, index, mime_types): for suffix, mime in mime_types.items(): path = Path(f"{kind}_{index}{suffix}") if path.exists(): return f"data:{mime};base64,{base64.b64encode(path.read_bytes()).decode()}" raise FileNotFoundError(f"Expected {kind}_{index} with suffix in {list(mime_types)}") <% if (imageCount > 0) { %> image_data_urls = [media_data_url("image", i, <%- mimePy(mediaConfig.image.mimeTypes) %>) for i in <%- imageCount === 1 ? "[1]" : ("range(1, " + (imageCount + 1) + ")") %>] <% } %><% if (videoCount > 0) { %> video_data_urls = [media_data_url("video", i, <%- mimePy(mediaConfig.video.mimeTypes) %>) for i in <%- videoCount === 1 ? "[1]" : ("range(1, " + (videoCount + 1) + ")") %>] <% } %><% if (audioCount > 0) { %> audio_data_urls = [media_data_url("audio", i, <%- mimePy(mediaConfig.audio.mimeTypes) %>) for i in <%- audioCount === 1 ? "[1]" : ("range(1, " + (audioCount + 1) + ")") %>] audio_b64s = [url.split(",", 1)[1] for url in audio_data_urls] <% } %><% } %> headers = { "Authorization": "Bearer $NVIDIA_API_KEY", "Accept": "text/event-stream" if stream else "application/json", } payload = <%- renderedPayload %> response = requests.post(invoke_url, headers=headers, json=payload, stream=stream) if stream: for line in response.iter_lines(): if line: print(line.decode("utf-8")) else: print(response.json()) 64:T1647,<% const deepClone = (value) => { if (Array.isArray(value)) return value.map(deepClone); if (value && typeof value === "object") { const out = {}; for (const key of Object.keys(value)) out[key] = deepClone(value[key]); return out; } return value; }; const mediaConfig = {"image": {"mimeTypes": {}}, "video": {"mimeTypes": {}}, "audio": {"mimeTypes": {}, "defaultFormat": "wav"}}; const payloadForSnippet = deepClone(request || {}); const messagesForSnippet = deepClone(payloadForSnippet.messages || []); const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0; const isInlineDataUrl = (url, kind) => ( typeof url === "string" && new RegExp(`^data:${kind}/[^;]+;base64,`, "i").test(url) ); const mediaUrl = (part, key) => { const slot = part?.[key]; return typeof slot === "string" ? slot : slot?.url; }; const setMediaUrl = (part, key, url) => { if (typeof part[key] === "string") part[key] = url; else { if (!part[key] || typeof part[key] !== "object") part[key] = {}; part[key].url = url; } }; let imageCount = 0; let hasUnsupportedMedia = false; messagesForSnippet.forEach((message) => { if (!Array.isArray(message?.content)) return; message.content.forEach((part) => { const imageUrl = mediaUrl(part, "image_url"); if (hasMediaKind("image") && part?.type === "image_url" && isInlineDataUrl(imageUrl, "image")) { setMediaUrl(part, "image_url", `__IMAGE_DATA_URL_${imageCount}__`); imageCount += 1; } else if (part?.type === "video_url" && isInlineDataUrl(mediaUrl(part, "video_url"), "video")) { // LangChain cannot ship local file helpers for video; reject inline only. hasUnsupportedMedia = true; setMediaUrl(part, "video_url", "__UNSUPPORTED_MEDIA_DATA__"); } else if (part?.type === "audio_url" && isInlineDataUrl(mediaUrl(part, "audio_url"), "audio")) { hasUnsupportedMedia = true; setMediaUrl(part, "audio_url", "__UNSUPPORTED_MEDIA_DATA__"); } else if (part?.type === "input_audio" && part.input_audio?.data) { hasUnsupportedMedia = true; part.input_audio.data = "__UNSUPPORTED_MEDIA_DATA__"; } }); }); const pyRepr = (value, level = 0) => { if (value === null) return "None"; if (typeof value === "boolean") return value ? "True" : "False"; if (typeof value === "number") return String(value); if (typeof value === "string") return JSON.stringify(value); const pad = " ".repeat(level); const innerPad = " ".repeat(level + 2); if (Array.isArray(value)) { if (value.length === 0) return "[]"; return "[\n" + value.map((item) => `${innerPad}${pyRepr(item, level + 2)},`).join("\n") + "\n" + pad + "]"; } if (typeof value === "object") { const entries = Object.entries(value); if (entries.length === 0) return "{}"; return "{\n" + entries.map(([key, item]) => `${innerPad}${JSON.stringify(key)}: ${pyRepr(item, level + 2)},`).join("\n") + "\n" + pad + "}"; } return "None"; }; const replaceAll = (s, a, b) => s.split(a).join(b); let renderedMessages = pyRepr(messagesForSnippet); for (let i = 0; i < imageCount; i++) { renderedMessages = replaceAll(renderedMessages, `"__IMAGE_DATA_URL_${i}__"`, `image_data_urls[${i}]`); } // Either file helpers XOR inline base64 - never emit unused helpers. const usesFileHelpers = imageCount > 0 && [...Array(imageCount).keys()].every((i) => renderedMessages.includes(`image_data_urls[${i}]`)); if (!usesFileHelpers) { imageCount = 0; renderedMessages = pyRepr(deepClone((request || {}).messages || [])); } %> from langchain_nvidia_ai_endpoints import ChatNVIDIA <% if (imageCount > 0) { %>from pathlib import Path import base64 <% } %> client = ChatNVIDIA( model="<%- request.model %>", api_key="$NVIDIA_API_KEY", temperature=<%- request.temperature %>, top_p=<%- request.top_p %>, max_completion_tokens=<%- request.max_tokens %>, ) <% if (imageCount > 0) { %> def find_image_path(index): image_mime_types = <%- pyRepr(mediaConfig.image.mimeTypes) %> for suffix in image_mime_types: path = Path(f"image_{index}{suffix}") if path.exists(): return path expected = "/".join(f".{suffix.lstrip('.')}" for suffix in image_mime_types) raise FileNotFoundError(f"Expected image_{index}{expected}") def read_image_data_url(path): image_mime_types = <%- pyRepr(mediaConfig.image.mimeTypes) %> with open(path, "rb") as f: image_b64 = base64.b64encode(f.read()).decode() return f"data:{image_mime_types[path.suffix.lower()]};base64,{image_b64}" image_data_urls = [read_image_data_url(find_image_path(i)) for i in range(1, <%- imageCount + 1 %>)] <% } %><% if (hasUnsupportedMedia) { %> raise ValueError("LangChain snippet currently supports text and image input. Use Python, Node.js, or Shell for video or audio input.") <% } %> lc_messages = <%- renderedMessages %> <% if (request.stream) { %> for chunk in client.stream(lc_messages<%= request.chat_template_kwargs ? ', chat_template_kwargs=' + pyRepr(request.chat_template_kwargs) : '' %>): if chunk.additional_kwargs and "reasoning_content" in chunk.additional_kwargs: print(chunk.additional_kwargs["reasoning_content"], end="") print(chunk.content, end="") <% } else { %> response = client.invoke(lc_messages<%= request.chat_template_kwargs ? ', chat_template_kwargs=' + pyRepr(request.chat_template_kwargs) : '' %>) if response.additional_kwargs and "reasoning_content" in response.additional_kwargs: print(response.additional_kwargs["reasoning_content"]) print(response.content) <% } %> 65:T1a86,<% const deepClone = (value) => { if (Array.isArray(value)) return value.map(deepClone); if (value && typeof value === "object") { const out = {}; for (const key of Object.keys(value)) out[key] = deepClone(value[key]); return out; } return value; }; const mediaConfig = {"image": {"mimeTypes": {}}, "video": {"mimeTypes": {}}, "audio": {"mimeTypes": {}, "defaultFormat": "wav"}}; const payloadForSnippet = deepClone(request || {}); const hasStream = payloadForSnippet.stream !== undefined && payloadForSnippet.stream !== null; if (hasStream) payloadForSnippet.stream = "__STREAM__"; const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0; const isInlineDataUrl = (url, kind) => ( typeof url === "string" && new RegExp(`^data:${kind}/[^;]+;base64,`, "i").test(url) ); let imageCount = 0; let videoCount = 0; let audioCount = 0; (payloadForSnippet.messages || []).forEach((message) => { if (!Array.isArray(message?.content)) return; message.content.forEach((part) => { if (hasMediaKind("image") && part?.type === "image_url" && isInlineDataUrl(part.image_url?.url, "image")) { part.image_url.url = `__IMAGE_DATA_URL_${imageCount}__`; imageCount += 1; } else if (hasMediaKind("video") && part?.type === "video_url" && isInlineDataUrl(part.video_url?.url, "video")) { part.video_url.url = `__VIDEO_DATA_URL_${videoCount}__`; videoCount += 1; } else if (hasMediaKind("audio") && part?.type === "audio_url" && isInlineDataUrl(part.audio_url?.url, "audio")) { part.audio_url.url = `__AUDIO_DATA_URL_${audioCount}__`; audioCount += 1; } else if (hasMediaKind("audio") && part?.type === "input_audio" && part.input_audio?.data) { part.input_audio.data = `__AUDIO_B64_${audioCount}__`; part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat; audioCount += 1; } }); }); const replaceAll = (s, a, b) => s.split(a).join(b); let renderedPayload = JSON.stringify(payloadForSnippet, null, 2); if (hasStream) renderedPayload = replaceAll(renderedPayload, '"__STREAM__"', "stream"); for (let i = 0; i < imageCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__IMAGE_DATA_URL_${i}__"`, `imageDataUrls[${i}]`); } for (let i = 0; i < videoCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__VIDEO_DATA_URL_${i}__"`, `videoDataUrls[${i}]`); } for (let i = 0; i < audioCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__AUDIO_DATA_URL_${i}__"`, `audioDataUrls[${i}]`); renderedPayload = replaceAll(renderedPayload, `"__AUDIO_B64_${i}__"`, `audioB64s[${i}]`); } // Either file helpers XOR inline base64 - never emit unused helpers. const usesFileHelpers = imageCount + videoCount + audioCount > 0 && [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`imageDataUrls[${i}]`)) && [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`videoDataUrls[${i}]`)) && [...Array(audioCount).keys()].every( (i) => renderedPayload.includes(`audioDataUrls[${i}]`) || renderedPayload.includes(`audioB64s[${i}]`) ); if (!usesFileHelpers) { imageCount = 0; videoCount = 0; audioCount = 0; const restored = deepClone(request || {}); const restoredStream = restored.stream !== undefined && restored.stream !== null; if (restoredStream) restored.stream = "__STREAM__"; renderedPayload = JSON.stringify(restored, null, 2); if (restoredStream) renderedPayload = replaceAll(renderedPayload, '"__STREAM__"', "stream"); } const hasMedia = usesFileHelpers; %> import axios from 'axios'; <% if (hasMedia) { %>import { existsSync } from 'node:fs'; import { readFile } from 'node:fs/promises'; import { extname } from 'node:path'; <% } %> const invokeUrl = "https://integrate.api.nvidia.com/v1/chat/completions"; const stream = <%- request.stream %>; const headers = { "Authorization": "Bearer $NVIDIA_API_KEY", "Accept": stream ? "text/event-stream" : "application/json" }; <% if (hasMedia) { %> function findMediaPath(kind, index, mimeTypes) { for (const suffix of Object.keys(mimeTypes)) { const path = `${kind}_${index}${suffix}`; if (existsSync(path)) { return path; } } const expected = Object.keys(mimeTypes).join("/"); throw new Error(`Expected ${kind}_${index}${expected}`); } async function readMediaB64(path) { return Buffer.from(await readFile(path)).toString('base64'); } function mediaDataUrl(path, mediaB64, mimeTypes) { return `data:${mimeTypes[extname(path).toLowerCase()]};base64,${mediaB64}`; } <% } %> async function main() { <% if (imageCount > 0) { %> const imageMimeTypes = <%- JSON.stringify(mediaConfig.image.mimeTypes, null, 2) %>; const imagePaths = [<% for (let i = 1; i <= imageCount; i++) { %>findMediaPath("image", <%- i %>, imageMimeTypes)<%- i < imageCount ? ", " : "" %><% } %>]; const imageB64s = await Promise.all(imagePaths.map(readMediaB64)); const imageDataUrls = imagePaths.map((path, index) => mediaDataUrl(path, imageB64s[index], imageMimeTypes)); <% } %><% if (videoCount > 0) { %> const videoMimeTypes = <%- JSON.stringify(mediaConfig.video.mimeTypes, null, 2) %>; const videoPaths = [<% for (let i = 1; i <= videoCount; i++) { %>findMediaPath("video", <%- i %>, videoMimeTypes)<%- i < videoCount ? ", " : "" %><% } %>]; const videoB64s = await Promise.all(videoPaths.map(readMediaB64)); const videoDataUrls = videoPaths.map((path, index) => mediaDataUrl(path, videoB64s[index], videoMimeTypes)); <% } %><% if (audioCount > 0) { %> const audioMimeTypes = <%- JSON.stringify(mediaConfig.audio.mimeTypes, null, 2) %>; const audioPaths = [<% for (let i = 1; i <= audioCount; i++) { %>findMediaPath("audio", <%- i %>, audioMimeTypes)<%- i < audioCount ? ", " : "" %><% } %>]; const audioB64s = await Promise.all(audioPaths.map(readMediaB64)); const audioDataUrls = audioPaths.map((path, index) => mediaDataUrl(path, audioB64s[index], audioMimeTypes)); <% } %> const payload = <%- renderedPayload %>; const response = await axios.post(invokeUrl, payload, { headers: headers, responseType: stream ? 'stream' : 'json' }); if (stream) { response.data.on('data', (chunk) => { console.log(chunk.toString()); }); } else { console.log(JSON.stringify(response.data)); } } main().catch(error => { if (error.response) { console.error(`HTTP ${error.response.status}`); if (error.response.data?.on) { error.response.data.on('data', (chunk) => console.error(chunk.toString())); } else { console.error(error.response.data); } } else { console.error(error); } }); 66:T153d,<% const deepClone = (value) => { if (Array.isArray(value)) return value.map(deepClone); if (value && typeof value === "object") { const out = {}; for (const key of Object.keys(value)) out[key] = deepClone(value[key]); return out; } return value; }; const mediaConfig = {"image": {"mimeTypes": {}}, "video": {"mimeTypes": {}}, "audio": {"mimeTypes": {}, "defaultFormat": "wav"}}; const payloadForSnippet = deepClone(request || {}); const hasMediaKind = (kind) => Object.keys(mediaConfig[kind].mimeTypes || {}).length > 0; const isInlineDataUrl = (url, kind) => ( typeof url === "string" && new RegExp(`^data:${kind}/[^;]+;base64,`, "i").test(url) ); let imageCount = 0; let videoCount = 0; let audioCount = 0; (payloadForSnippet.messages || []).forEach((message) => { if (!Array.isArray(message?.content)) return; message.content.forEach((part) => { if (hasMediaKind("image") && part?.type === "image_url" && isInlineDataUrl(part.image_url?.url, "image")) { part.image_url.url = `__IMAGE_DATA_URL_${imageCount}__`; imageCount += 1; } else if (hasMediaKind("video") && part?.type === "video_url" && isInlineDataUrl(part.video_url?.url, "video")) { part.video_url.url = `__VIDEO_DATA_URL_${videoCount}__`; videoCount += 1; } else if (hasMediaKind("audio") && part?.type === "audio_url" && isInlineDataUrl(part.audio_url?.url, "audio")) { part.audio_url.url = `__AUDIO_DATA_URL_${audioCount}__`; audioCount += 1; } else if (hasMediaKind("audio") && part?.type === "input_audio" && part.input_audio?.data) { part.input_audio.data = `__AUDIO_B64_${audioCount}__`; part.input_audio.format = part.input_audio.format || mediaConfig.audio.defaultFormat; audioCount += 1; } }); }); const replaceAll = (s, a, b) => s.split(a).join(b); let renderedPayload = JSON.stringify(payloadForSnippet, null, 2); for (let i = 0; i < imageCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__IMAGE_DATA_URL_${i}__"`, `"${"${"}image_${i + 1}_data_url}"`); } for (let i = 0; i < videoCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__VIDEO_DATA_URL_${i}__"`, `"${"${"}video_${i + 1}_data_url}"`); } for (let i = 0; i < audioCount; i++) { renderedPayload = replaceAll(renderedPayload, `"__AUDIO_DATA_URL_${i}__"`, `"${"${"}audio_${i + 1}_data_url}"`); renderedPayload = replaceAll(renderedPayload, `"__AUDIO_B64_${i}__"`, `"${"${"}audio_${i + 1}_b64}"`); } // Either file helpers XOR inline base64 - never emit unused helpers. const usesFileHelpers = imageCount + videoCount + audioCount > 0 && [...Array(imageCount).keys()].every((i) => renderedPayload.includes(`image_${i + 1}_data_url`)) && [...Array(videoCount).keys()].every((i) => renderedPayload.includes(`video_${i + 1}_data_url`)) && [...Array(audioCount).keys()].every( (i) => renderedPayload.includes(`audio_${i + 1}_data_url`) || renderedPayload.includes(`audio_${i + 1}_b64`) ); if (!usesFileHelpers) { imageCount = 0; videoCount = 0; audioCount = 0; renderedPayload = JSON.stringify(deepClone(request || {}), null, 2); } const hasMedia = usesFileHelpers; %> stream=<%- request.stream %> if [ "$stream" = true ]; then accept_header='Accept: text/event-stream' else accept_header='Accept: application/json' fi <% if (hasMedia) { %> mime_type_for() { case "$1" in <% for (const [suffix, mimeType] of Object.entries({...mediaConfig.image.mimeTypes, ...mediaConfig.video.mimeTypes, ...mediaConfig.audio.mimeTypes})) { %> *<%- suffix %>) echo "<%- mimeType %>" ;; <% } %> *) echo "Unsupported media file: $1" >&2; return 1 ;; esac } find_media_path() { local kind="$1" local index="$2" shift 2 local ext path for ext in "$@"; do path="${kind}_${index}.${ext}" if [ -f "$path" ]; then echo "$path" return 0 fi done echo "Expected ${kind}_${index}.$*" >&2 return 1 } media_b64() { base64 < "$1" | tr -d '\n' } media_data_url() { local path="$1" local mime_type media_base64 mime_type=$(mime_type_for "$path") || return 1 media_base64=$(media_b64 "$path") || return 1 printf 'data:%s;base64,%s' "$mime_type" "$media_base64" } <% for (let i = 1; i <= imageCount; i++) { %> image_<%- i %>_path=$(find_media_path image <%- i %><% for (const suffix of Object.keys(mediaConfig.image.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>) image_<%- i %>_data_url=$(media_data_url "$image_<%- i %>_path") <% } %><% for (let i = 1; i <= videoCount; i++) { %> video_<%- i %>_path=$(find_media_path video <%- i %><% for (const suffix of Object.keys(mediaConfig.video.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>) video_<%- i %>_data_url=$(media_data_url "$video_<%- i %>_path") <% } %><% for (let i = 1; i <= audioCount; i++) { %> audio_<%- i %>_path=$(find_media_path audio <%- i %><% for (const suffix of Object.keys(mediaConfig.audio.mimeTypes)) { %> <%- suffix.slice(1) %><% } %>) audio_<%- i %>_b64=$(media_b64 "$audio_<%- i %>_path") audio_<%- i %>_data_url=$(media_data_url "$audio_<%- i %>_path") <% } %><% } %> cat > payload.json < JSON curl https://integrate.api.nvidia.com/v1/chat/completions \ -H "Authorization: Bearer $NVIDIA_API_KEY" \ -H "Content-Type: application/json" \ -H "$accept_header" \ -d @payload.json58:["$","$L62",null,{"spec":{"openAPISpec":{"openapi":"3.1.0","info":{"title":"NVIDIA NIM API for microsoft/phi-3.5-vision-instruct","description":"The NVIDIA NIM REST API. Please see https://docs.api.nvidia.com/nim/reference/microsoft-phi-3_5-vision-instruct for more details.","version":"1.0.0","termsOfService":"https://nvidia.com/legal/terms-of-use","contact":{"name":"NVIDIA Support","url":"https://help.nvidia.com/"},"license":{"name":"Apache 2.0","url":"https://choosealicense.com/licenses/apache-2.0/"}},"servers":[{"url":"https://integrate.api.nvidia.com/v1"}],"tags":[{"name":"Multimodal API","description":"This API performs inference using visual language understanding models."}],"paths":{"/microsoft/phi-3.5-vision-instruct":{"post":{"tags":["Multimodal API"],"summary":"Request response from the model","description":"Invokes inference using the model chat parameters.","operationId":"invokeFunction","requestBody":{"content":{"application/json":{"schema":{"$ref":"#/components/schemas/NIMLLMChatCompletionRequest"}}},"required":true},"responses":{"200":{"description":"Invocation is fulfilled","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ChatCompletionResponse"}},"text/event-stream":{"schema":{"$ref":"#/components/schemas/ChatCompletionStreamResponse"}}}},"202":{"description":"Result is pending. Client should poll using the requestId.\n","content":{"application/json":{"example":{},"schema":{}}},"headers":{"NVCF-REQID":{"description":"requestId required for pooling","schema":{"type":"string","format":"uuid","maxLength":36}},"NVCF-STATUS":{"description":"Invocation status","schema":{"type":"string","format":"^[a-zA-Z-]{1,64}$","maxLength":64}}}},"422":{"description":"Validation failed, provided entity could not be processed.","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ErrorResponse"},"example":{"type":"urn:nvcf-worker-service:problem-details:unprocessable-entity","title":"Unprocessable Entity","status":422,"detail":"string","instance":"/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48","requestId":"3fa85f64-5717-4562-b3fc-2c963f66afa6"}}}},"500":{"description":"The invocation ended with an error.","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ErrorResponse"},"example":{"type":"urn:nvcf-worker-service:problem-details:internal-server-error","title":"Internal Server Error","status":500,"detail":"string","instance":"/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48","requestId":"3fa85f64-5717-4562-b3fc-2c963f66afa6"}}}}},"x-nvai-meta":{"name":"Create chat completion","returns":"Returns a [chat completion](/docs/api-reference/chat/object) object, or a streamed sequence of [chat completion chunk](/docs/api-reference/chat/streaming) objects if the request is streamed.\n","path":"create","examples":[{"name":"image_captioning","input":{"text":"Describe the image.","image":"https://assets.ngc.nvidia.com/products/api-catalog/phi-3-5-vision/example1a.jpg"},"output":{"text":"The image displays a red stop sign in the foreground with a Chinese character '停' above it, indicating a stop sign in a Chinese-speaking area. In the background, there is a traditional Chinese building with red and gold colors, and a black vehicle parked in front of it. There are also various storefronts and a clear sky."},"requestJson":"{ \"max_tokens\": 512, \"temperature\": 0.2, \"top_p\": 0.7, \"model\": \"microsoft/phi-3.5-vision-instruct\", \"messages\": [ { \"role\": \"user\", \"content\": [ { \"type\": \"text\", \"text\": \"<%- input.text %>\" }, { \"type\": \"image_url\", \"image_url\": { \"url\": \"<%- input.image %>\" } } ] } ], \"stream\": true }","responseJson":"{ \"id\": \"b52fc961-3c88-41cd-91c7-eab3eef21015\", \"choices\": [ { \"index\": 0, \"model\": \"microsoft/phi-3.5-vision-instruct\", \"message\": { \"role\": \"assistant\", \"content\": \"<%- output.text %>\" }, \"finish_reason\": \"stop\" } ], \"usage\": { \"completion_tokens\": 76, \"prompt_tokens\": 770, \"total_tokens\": 846 } }"},{"name":"visual_qa","input":{"text":"How many dogs are in the image?","image":"https://assets.ngc.nvidia.com/products/api-catalog/phi-3-5-vision/example2b.jpg"},"output":{"text":"There are two dogs in the image."},"requestJson":"{ \"max_tokens\": 512, \"temperature\": 0.2, \"top_p\": 0.7, \"model\": \"microsoft/phi-3.5-vision-instruct\", \"messages\": [ { \"role\": \"user\", \"content\": [ { \"type\": \"text\", \"text\": \"<%- input.text %>\" }, { \"type\": \"image_url\", \"image_url\": { \"url\": \"<%- input.image %>\" } } ] } ], \"stream\": false }","responseJson":"{ \"id\": \"7bda3f6a-f5e1-48c4-a67b-fc77f362dc3c\", \"choices\": [ { \"index\": 0, \"model\": \"microsoft/phi-3.5-vision-instruct\", \"message\": { \"role\": \"assistant\", \"content\": \"<%- output.text %>\" }, \"finish_reason\": \"stop\" } ], \"usage\": { \"completion_tokens\": 10, \"prompt_tokens\": 773, \"total_tokens\": 783 } }"}],"templates":[{"title":"Default","response":"{\n \"choices\": [{\n \"index\": 0,\n \"model\": \"microsoft/phi-3.5-vision-instruct\",\n \"message\": {\n \"role\": \"assistant\",\n \"content\": \"...\",\n },\n \"finish_reason\": \"stop\"\n }],\n \"usage\": {\n \"prompt_tokens\": 9,\n \"completion_tokens\": 12,\n \"total_tokens\": 21\n }\n}\n","requestEjs":{"python":"$63","langChain":"$64","node.js":"$65","curl":"$66"}}]}}},"/status/{requestId}":{"get":{"tags":["Multimodal API"],"summary":"Gets the result of an earlier function invocation request that returned a status of 202.","operationId":"getFunctionInvocationResult","parameters":[{"name":"requestId","in":"path","description":"requestId to poll results","required":true,"schema":{"type":"string","format":"uuid","maxLength":36}}],"responses":{"200":{"description":"Invocation is fulfilled","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ChatCompletionResponse"}}}},"202":{"description":"Result is pending. Client should poll using the requestId.\n","content":{"application/json":{"example":{},"schema":{}}},"headers":{"NVCF-REQID":{"description":"requestId required for pooling","schema":{"type":"string","format":"uuid","maxLength":36}},"NVCF-STATUS":{"description":"Invocation status","schema":{"type":"string","format":"^[a-zA-Z-]{1,64}$","maxLength":64}}}},"422":{"description":"The invocation ended with an error.","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ErrorResponse"},"example":{"type":"urn:nvcf-worker-service:problem-details:unprocessable-entity","title":"Unprocessable Entity","status":422,"detail":"string","instance":"/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48","requestId":"3fa85f64-5717-4562-b3fc-2c963f66afa6"}}}},"500":{"description":"The invocation ended with an error.","content":{"application/json":{"schema":{"$ref":"#/components/schemas/ErrorResponse"},"example":{"type":"urn:nvcf-worker-service:problem-details:internal-server-error","title":"Internal Server Error","status":500,"detail":"string","instance":"/v2/nvcf/pexec/functions/e598bfc1-b058-41af-869d-556d3c7e1b48","requestId":"3fa85f64-5717-4562-b3fc-2c963f66afa6"}}}}}}}},"security":[{"Token":[]}],"components":{"securitySchemes":{"Token":{"type":"http","scheme":"bearer"}},"schemas":{"ChatCompletionResponse":{"properties":{"id":{"type":"string","title":"Id"},"object":{"type":"string","enum":["chat.completion"],"const":"chat.completion","title":"Object","default":"chat.completion"},"created":{"type":"integer","title":"Created"},"model":{"type":"string","title":"Model"},"choices":{"items":{"$ref":"#/components/schemas/ChatCompletionResponseChoice"},"type":"array","title":"Choices"},"usage":{"$ref":"#/components/schemas/UsageInfo"}},"additionalProperties":false,"type":"object","required":["model","choices","usage"],"title":"ChatCompletionResponse"},"ChatCompletionResponseChoice":{"properties":{"index":{"type":"integer","title":"Index"},"message":{"$ref":"#/components/schemas/ChatMessage"},"finish_reason":{"anyOf":[{"type":"string","enum":["stop","length"]},{"type":"null"}],"title":"Finish Reason"},"stop_reason":{"anyOf":[{"type":"integer"},{"type":"string"},{"type":"null"}],"title":"Stop Reason"}},"additionalProperties":false,"type":"object","required":["index","message"],"title":"ChatCompletionResponseChoice"},"ChatCompletionResponseStreamChoice":{"properties":{"index":{"type":"integer","title":"Index"},"delta":{"$ref":"#/components/schemas/DeltaMessage"},"finish_reason":{"anyOf":[{"type":"string","enum":["stop","length"]},{"type":"null"}],"title":"Finish Reason"},"stop_reason":{"anyOf":[{"type":"integer"},{"type":"string"},{"type":"null"}],"title":"Stop Reason"}},"additionalProperties":false,"type":"object","required":["index","delta"],"title":"ChatCompletionResponseStreamChoice"},"ChatCompletionStreamResponse":{"properties":{"id":{"type":"string","title":"Id"},"object":{"type":"string","enum":["chat.completion.chunk"],"const":"chat.completion.chunk","title":"Object","default":"chat.completion.chunk"},"created":{"type":"integer","title":"Created"},"model":{"type":"string","title":"Model"},"choices":{"items":{"$ref":"#/components/schemas/ChatCompletionResponseStreamChoice"},"type":"array","title":"Choices"},"usage":{"anyOf":[{"$ref":"#/components/schemas/UsageInfo"},{"type":"null"}]}},"additionalProperties":false,"type":"object","required":["model","choices"],"title":"ChatCompletionStreamResponse"},"ChatMessage":{"properties":{"role":{"type":"string","title":"Role"},"content":{"anyOf":[{"type":"string"},{"type":"null"}],"title":"Content"}},"additionalProperties":false,"type":"object","required":["role"],"title":"ChatMessage"},"DeltaMessage":{"properties":{"role":{"anyOf":[{"type":"string"},{"type":"null"}],"title":"Role"},"content":{"anyOf":[{"type":"string"},{"type":"null"}],"title":"Content"}},"additionalProperties":false,"type":"object","title":"DeltaMessage"},"ErrorResponse":{"properties":{"object":{"type":"string","title":"Object","default":"error"},"message":{"type":"string","title":"Message"},"type":{"type":"string","title":"Type"},"param":{"anyOf":[{"type":"string"},{"type":"null"}],"title":"Param"},"code":{"type":"integer","title":"Code"}},"additionalProperties":false,"type":"object","required":["message","type","code"],"title":"ErrorResponse"},"HTTPValidationError":{"properties":{"detail":{"items":{"$ref":"#/components/schemas/ValidationError"},"type":"array","title":"Detail"}},"type":"object","title":"HTTPValidationError"},"ImageURL":{"properties":{"url":{"type":"string","title":"Url"}},"type":"object","required":["url"],"title":"ImageURL"},"NIMLLMChatCompletionMessage":{"properties":{"role":{"allOf":[{"$ref":"#/components/schemas/Role"}],"description":"The role of the message's author."},"content":{"anyOf":[{"type":"string"},{"items":{"anyOf":[{"$ref":"#/components/schemas/NIMVLMChatCompletionContentPartImage"},{"$ref":"#/components/schemas/NIMVLMChatCompletionContentPartText"}]},"type":"array"}],"title":"Content","description":"The contents of the message.\n
To pass media (only with role=`user`):\n
- Use content parts. Text is passed with `type`=`text`.\n
- Images are passed with `type`=`image_url`; set `image_url.url` to an image URL or a base64 data URI like `data:image/{format};base64,{base64encodedimage}`.\n
For `system` and `assistant` roles, the object list format is not supported.\n"}},"type":"object","required":["role","content"],"title":"NIMLLMChatCompletionMessage"},"NIMLLMChatCompletionRequest":{"properties":{"messages":{"items":{"$ref":"#/components/schemas/NIMLLMChatCompletionMessage"},"type":"array","minItems":1,"title":"Messages","description":"A list of messages comprising the conversation so far."},"model":{"type":"string","title":"Model","description":"The model to use."},"frequency_penalty":{"anyOf":[{"type":"number","maximum":2,"minimum":-2},{"type":"null"}],"title":"Frequency Penalty","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim.","default":0},"max_tokens":{"anyOf":[{"type":"integer","maximum":8192,"minimum":1},{"type":"null"}],"title":"Max Tokens","description":"The maximum number of tokens that can be generated.","default":512},"presence_penalty":{"anyOf":[{"type":"number","maximum":2,"minimum":-2},{"type":"null"}],"title":"Presence Penalty","description":"Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far, increasing the model's likelihood to talk about new topics.","default":0},"seed":{"anyOf":[{"type":"integer","maximum":9223372036854776000,"minimum":-9223372036854776000},{"type":"null"}],"title":"Seed","description":"Changing the seed will produce a different response with similar characteristics. Fixing the seed will reproduce the same results if all other parameters are also kept constant."},"stop":{"anyOf":[{"type":"string"},{"items":{"type":"string"},"type":"array"},{"type":"null"}],"title":"Stop","description":"Sequences where the API will stop generating further tokens."},"stream":{"anyOf":[{"type":"boolean"},{"type":"null"}],"title":"Stream","description":"If set, partial message deltas will be sent, like in ChatGPT. Tokens will be sent as data-only [server-sent events](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format) as they become available, with the stream terminated by a `data: [DONE]`","default":false},"temperature":{"anyOf":[{"type":"number","maximum":2,"minimum":0},{"type":"null"}],"title":"Temperature","description":"What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.","default":0.2},"top_p":{"anyOf":[{"type":"number","maximum":1,"exclusiveMinimum":0},{"type":"null"}],"title":"Top P","description":"An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered. We generally recommend altering this or `temperature` but not both.","default":0.7}},"additionalProperties":false,"type":"object","required":["messages","model"],"title":"NIMLLMChatCompletionRequest"},"NIMVLMChatCompletionContentPartImage":{"properties":{"image_url":{"$ref":"#/components/schemas/ImageURL"},"type":{"type":"string","enum":["image_url"],"const":"image_url","title":"Type"}},"type":"object","required":["image_url","type"],"title":"NIMVLMChatCompletionContentPartImage"},"NIMVLMChatCompletionContentPartText":{"properties":{"text":{"type":"string","title":"Text"},"type":{"type":"string","enum":["text"],"const":"text","title":"Type"}},"type":"object","required":["text","type"],"title":"NIMVLMChatCompletionContentPartText"},"Role":{"type":"string","enum":["assistant","user"],"title":"Role"},"UsageInfo":{"properties":{"prompt_tokens":{"type":"integer","title":"Prompt Tokens","default":0},"total_tokens":{"type":"integer","title":"Total Tokens","default":0},"completion_tokens":{"anyOf":[{"type":"integer"},{"type":"null"}],"title":"Completion Tokens","default":0}},"additionalProperties":false,"type":"object","title":"UsageInfo"},"ValidationError":{"properties":{"loc":{"items":{"anyOf":[{"type":"string"},{"type":"integer"}]},"type":"array","title":"Location"},"msg":{"type":"string","title":"Message"},"type":{"type":"string","title":"Error Type"}},"type":"object","required":["loc","msg","type"],"title":"ValidationError"}}}},"namespace":"qc69jvmznzxy","updatedDate":"2026-07-16T02:10:02.828Z","nvcfFunctionId":"None","createdDate":"2024-08-22T13:10:16.664Z","attributes":{"apiDocsUrl":"https://docs.api.nvidia.com/nim/reference/microsoft-phi-3_5-vision-instruct","termsOfUse":"GOVERNING TERMS: The trial service is governed by the NVIDIA API Trial Terms of Service; and the use of this model is governed by the MIT License.\n","deprecationNotice":"This API will be deprecated on 04/15/2026. It will no longer be supported after 04/15/2026. Please use qwen/qwen3.5-122b-a10b instead. For more information, visit our API Reference.","showUnavailableBanner":false,"playground":{"type":"custom","input":{"items":[{"key":"image_b64","type":"upload","optional":true,"formats":["jpg","jpeg","png"],"title":"Upload Image"},{"key":"prompt","type":"text-area","title":"Input"}]},"output":{"items":[{"key":"choices[0].message.content","type":"code"}]},"requestBody":"{ \"model\": \"<%- request.model %>\", \"max_tokens\": <%- request.max_tokens %>, \"temperature\": <%- request.temperature %>, \"top_p\": <%- request.top_p %>, \"frequency_penalty\": <%- request.frequency_penalty %>, \"presence_penalty\": <%- request.presence_penalty %>, <% if (request.stop) { %>\"stop\": \"<%- request.stop %>\",<% } %> <% if (request.seed) { %>\"seed\": <%- request.seed %>,<% } %> \"stream\": <%- request.stream %>, \"messages\": [ { \"role\": \"user\", \"content\": [ { \"type\": \"text\", \"text\": \"<%- request.prompt %>\" }<% if (request.image_b64) { %>, { \"type\": \"image_url\", \"image_url\": { \"url\": \"<%- request.image_b64 %>\" } }<% } %> ] } ] }"},"specifications":{"contextLength":131072,"parameters":4146621440,"inputModalities":["Text","Image"],"outputModalities":["Text"]},"modelCapability":{"functionCalling":false,"structuredOutput":false,"reasoning":false}},"artifactName":"phi-3_5-vision-instruct"}}]