models documentation - Azure/azureml-assets GitHub Wiki

Models

Categories

All models

ALLaM is a series of powerful language models designed to advance Arabic Language Technology (ALT) developed by the National Center for Artificial Intelligence (NCAI) at the Saudi Data and AI Authority (SDAIA). ALLaM-2-7b-instruct is traine...

  • analyze-conversations

    The "Analyze Conversations" is a standard model that utilizes Azure AI Language to perform various analyzes on text-based conversations. Azure AI language hosts pre-trained, task-oriented, and optimized conversation focused ML models, including various summarization aspects, PII entity extraction...

  • analyze-documents

    The "Analyze Documents" is a standard model that utilizes Azure AI Language to perform various analyzes on text-based documents. Azure AI language hosts pre-trained, task-oriented, and optimized document focused ML models, such as summarization, sentiment analysis, entity extraction, etc.

...

  • ask-wikipedia

    The "Ask Wikipedia" is a Q&A model that employs GPT3.5 to answer questions using information sourced from Wikipedia, ensuring more grounded responses. This process involves identifying the relevant Wikipedia link and extracting its contents. These contents are then used as an augmented prompt, en...

  • AutoML-Image-Classification

    Automated Machine Learning, or AutoML, is a process that automates the repetitive and time-consuming tasks involved in developing machine learning models. This helps data scientists, analysts, and developers to create models more efficiently and with higher quality, resulting in increased product...

  • AutoML-Image-Instance-Segmentation

    Automated Machine Learning, or AutoML, is a process that automates the repetitive and time-consuming tasks involved in developing machine learning models. This helps data scientists, analysts, and developers to create models more efficiently and with higher quality, resulting in increased product...

  • AutoML-Image-Object-Detection

    Automated Machine Learning, or AutoML, is a process that automates the repetitive and time-consuming tasks involved in developing machine learning models. This helps data scientists, analysts, and developers to create models more efficiently and with higher quality, resulting in increased product...

  • AutoML-Named-Entity-Recognition

    Automated Machine Learning, or AutoML, is a process that automates the repetitive and time-consuming tasks involved in developing machine learning models. This helps data scientists, analysts, and developers to create models more efficiently and with higher quality, resulting in increased product...

  • AutoML-Text-Classification

    Automated Machine Learning, or AutoML, is a process that automates the repetitive and time-consuming tasks involved in developing machine learning models. This helps data scientists, analysts, and developers to create models more efficiently and with higher quality, resulting in increased product...

  • bert-base-cased

    BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inpu...

  • bert-base-uncased

    BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labeling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate input...

  • bert-large-cased

    BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inpu...

  • bert-large-uncased

    BERT is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inpu...

  • BiomedCLIP-PubMedBERT_256-vit_base_patch16_224

    BiomedCLIP is a biomedical vision-language foundation model that is pretrained on PMC-15M, a dataset of 15 million figure-caption pairs extracted from biomedical research articles in PubMed Central, using contrastive learning. It uses PubMedBERT as the text encoder and Vision Transformer as the i...

  • Bleu-Score-Evaluator

    | | | | -- | -- | | Score range | Float [0-1]: higher means better quality. | | What is this metric? | BLEU (Bilingual Evaluation Understudy) score is commonly used in natural language processing (NLP) and machine translation. It measures how closely the generated text matches the reference text....

  • bring-your-own-data-chat-qna

    The "Bring Your Own Data Chat QnA" is a pre-trained chat model, enhanced by GPT3.5, that leverages your personally indexed data and chat history to deliver more concrete and relevant answers. It involves processing the raw query through an embedding procedure, followed by a "Vector Search" to pin...

  • bring-your-own-data-qna

    The "Bring your own data QnA" is a pre-trained Q&A model, enhanced by GPT3.5, that leverages your personally indexed data to deliver more concrete and relevant answers. It involves processing the raw query through an embedding procedure, followed by a "Vector Search" to pinpoint the most pertinen...

  • bytetrack_yolox_x_crowdhuman_mot17-private-half

    bytetrack_yolox_x_crowdhuman_mot17-private-half model is from OpenMMLab's MMTracking library. Multi-object tracking (MOT) aims at estimating bounding boxes and identities of objects in videos. Most methods obtai...

  • camembert-base

    CamemBERT is a state-of-the-art language model for French based on the RoBERTa model.

It is now available on Hugging Face in 6 different versions with varying number of parameters, amount of pretraining data and pretraining data source domains.

Training Details

Training Data

OSCAR or Open...

  • chat-quality-safety-eval

    The chat quality and safety evaluation flow will evaluate the chat systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your LLM responses . Utilizing GPT model to assist with measurements aims to achieve a high agreement with human evaluatio...

  • chat-with-wikipedia

    The "Chat with Wikipedia" is a pre-trained chat model with GPT3.5: it combines conversation history and information from Wikipedia to make the answer more grounded. It involves finding a relevant Wikipedia link and getting page contents for the question. It can remember previous interactions and ...

  • classification-accuracy-eval

    The "Classification Accuracy Evaluation" is a model designed to assess the effectiveness of a data classification system. It involves matching each prediction against the ground truth, subsequently assigning a "Correct" or "Incorrect" score. The cumulative results are then leveraged to generate p...

  • Coherence-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Measures how well the language model can produce output that flows smoothly, reads naturally, and resembles human-like language. | | How does it work? | The coherence...

  • compvis-stable-diffusion-v1-4

    Stable Diffusion is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input. The Stable-Diffusion-v1-4 checkpoint was initialized with the weights of the Stable-Diffusion-v1-2 checkpoint and subsequently fine-tuned on 225k steps at resolution 5...

  • count-cars

    The "Count Cars" is a model designed for accurately quantifying the number of specific vehicles – particularly red cars – in given images. Utilizing the advanced capabilities of Azure OpenAI GPT-4 Turbo with Vision, this system meticulously analyzes each image, identifies and counts red cars, out...

  • CxrReportGen

    Overview

The CXRReportGen model utilizes a multimodal architecture, integrating a BiomedCLIP image encoder with a Phi-3-Mini text encoder to help an application interpret complex medical imaging studies of chest X-rays. CXRReportGen follows the same framework as **[MAIRA-2](https://www.micros...

  • databricks-dolly-v2-12b

    Databricks' dolly-v2-12b, an instruction-following large language model trained on the Databricks machine learning platform that is licensed for commercial use. Based on pythia-12b, Dolly is trained on ~15k instruction/response fine tuning records [databricks-dolly-15k](https://github.com/d...

  • Deci-DeciCoder-1b

    The Model Card for DeciCoder 1B provides details about a 1 billion parameter decoder-only code completion model developed by Deci. The model was trained on Python, Java, and JavaScript subsets of Starcoder Training Dataset and uses Grouped Query Attention with a context window of 2048 tokens. It ...

  • deci-decidiffusion-v1-0

    DeciDiffusion 1.0 is an 820 million parameter latent diffusion model designed for text-to-image conversion. Trained initially on the LAION-v2 dataset and fine-tuned on the LAION-ART dataset, the model's training involved advanced techniques to improve speed, training performance, and achieve su...

  • Deci-DeciLM-7B

    DeciLM-7B is a decoder-only text generation model with 7.04 billion parameters, released by Deci under the Apache 2.0 license. It is the top-performing 7B base language model on the Open LLM Leaderboard and uses variable Grouped-Query Attention (GQA) to achieve a superior balance between accuracy...

  • Deci-DeciLM-7B-instruct

    DeciLM-7B-instruct is a model for short-form instruction following, built by LoRA fine-tuning on the SlimOrca dataset. It is a derivative of the recently released DeciLM-7B language model, a pre-trained, high-efficiency generative text model with 7 billion parameters. DeciLM-7B-instruct is one of...

  • deepset-minilm-uncased-squad2

    Training Details

Hyperparameters

seed=42
batch_size = 12
n_epochs = 4
base_LM_model = "microsoft/MiniLM-L12-H384-uncased"
max_seq_len = 384
learning_rate = 4e-5
lr_schedule = LinearWarmup
warmup_proportion = 0.2
doc_stride=128
max_query_length=64
grad_acc_steps=4

Evaluation Res...

Training Details

Hype...

Election Critical Information (ECI) refers to any content related to elections, including voting processes, candidate information, and election results. The ECI evaluator uses the Azure AI Safety Evaluation service to assess the generated responses for ECI without a disclaimer.

#...

  • F1Score-Evaluator

    | | | | -- | -- | | Score range | Float [0-1]: higher means better quality. | | What is this metric? | F1 score measures the similarity by shared tokens between the generated text and the ground truth, focusing on both precision and recall. | | How does it work? | The F1-score computes the ratio...

  • facebook-bart-large-cnn

    BART is a transformer model that combines a bidirectional encoder similar to BERT with an autoregressive decoder akin to GPT. It is trained using two main techniques: (1) corrupting text with a chosen noising function, and (2) training a model to reconstruct the original text.

When fine-tuned fo...

  • facebook-deit-base-patch16-224

    DeiT (Data-efficient image Transformers) is an image transformer that do not require very large amounts of data for training. This is achieved through a novel distillation procedure using teacher-student strategy, which results in high throughput and accuracy. DeiT is pre-trained and fine-tuned o...

  • facebook-dinov2-base-imagenet1k-1-layer

    Vision Transformer (base-sized model) trained using DINOv2

Vision Transformer (ViT) model trained using the DINOv2 method. It was introduced in the paper DINOv2: Learning Robust Visual Features without Supervision by Oquab et al. and first released...

  • Facebook-DinoV2-Image-Embeddings-ViT-Base

    The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrained on a large collection of images in a self-supervised fashion with the DinoV2 method.

Images are presented to the model as a sequence of fixed-size patches, which are linearly embedded. One also adds a [CLS] token ...

  • Facebook-DinoV2-Image-Embeddings-ViT-Giant

    The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrained on a large collection of images in a self-supervised fashion with the DinoV2 method.

Images are presented to the model as a sequence of fixed-size patches, which are linearly embedded. One also adds a [CLS] token ...

  • facebook-sam-vit-base

    The Segment Anything Model (SAM) produces high quality object masks from input prompts such as points or boxes, and it can be used to generate masks for all objects in an image. It has been trained on a dataset of 11 million images and 1.1 bi...

  • facebook-sam-vit-huge

    The Segment Anything Model (SAM) produces high quality object masks from input prompts such as points or boxes, and it can be used to generate masks for all objects in an image. It has been trained on a dataset of 11 million images and 1.1 bi...

  • facebook-sam-vit-large

    The Segment Anything Model (SAM) produces high quality object masks from input prompts such as points or boxes, and it can be used to generate masks for all objects in an image. It has been trained on a dataset of 11 million images and 1.1 bi...

  • finiteautomata-bertweet-base-sentiment-analysis

    Repository: https://github.com/finiteautomata/pysentimiento/

Model trained with SemEval 2017 corpus (around ~40k tweets). Base model is BERTweet, a RoBERTa model trained on English tweets.

Uses `POS...

  • Fluency-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Fluency measures the effectiveness and clarity of written communication, focusing on grammatical accuracy, vocabulary range, sentence complexity, coherence, and overa...

  • Gleu-Score-Evaluator

    | | | | -- | -- | | Score range | Float [0-1]: higher means better quality. | | What is this metric? | The GLEU (Google-BLEU) score measures the similarity by shared n-grams between the generated text and ground truth, similar to the BLEU score, focusing on both precision and recall. But it addre...

  • google-vit-base-patch16-224

    The Vision Transformer (ViT) model, as introduced in the paper "An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale" by Dosovitskiy et al., underwent pre-training on ImageNet-21k with a resolution of 224x224. Su...

  • gpt2

    GPT-2 is a transformers model pretrained on a very large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generat...

  • gpt2-large

    GPT-2 Large is the 774M parameter version of GPT-2, a transformer-based language model created and released by OpenAI. The model is a pretrained model on English language using a causal language modeling (CLM)

Training Details

See the [associated paper](https://d4mucfpksywv.cloudfront.net/bet...

  • gpt2-medium

    GPT-2 Medium is the 355M parameter version of GPT-2, a transformer-based language model created and released by OpenAI. The model is a pretrained model on English language using a causal language modeling (CLM) objective.

Training Details

See the [associated paper](https://d4mucfpksywv.c...

  • Groundedness-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Groundedness measures how well the generated response aligns with the given context in a retrieval-augmented generation scenario, focusing on its relevance and accura...

  • Groundedness-Pro-Evaluator

    | | | | -- | -- | | Score range | Boolean: [true, false]: false if response is ungrounded and true if it's grounded. | | What is this metric? | Groundedness Pro (powered by Azure AI Content Safety) detects whether the generated text response is consistent or accurate with respect to the given ...

  • Hate-and-Unfairness-Evaluator

    Definition

Hateful and unfair content refers to any language pertaining to hate toward or unfair representations of individuals and social groups along factors including but not limited to race, ethnicity, nationality, gender, sexual orientation, religion, immigration status, ability, persona...

  • how-to-use-functions-with-GPT-chat-API

    The "Use Functions with Chat Models" is a chat model illustrates how to employ the LLM tool's Chat API with external functions, thereby expanding the capabilities of GPT models. The Chat Completion API includes an optional 'functions' parameter, which can be used to stipulate function specificati...

  • Indirect-Attack-Evaluator

    Definition

Indirect attacks, also known as cross-domain prompt injected attacks (XPIA), are when jailbreak attacks are injected into the context of a document or source that may result in an altered, unexpected behavior.

Indirect attacks evaluations are broken down into three subcategories: ...

  • Jean-Baptiste-camembert-ner

    Summary: camembert-ner is a NER model fine-tuned from camemBERT on the Wikiner-fr dataset and was validated on email/chat data. It shows better performance on entities that do not start with an uppercase. The model has four classes: O, MISC, PER, ORG and LOC. The model can be loaded using Hugging...

  • Llama-2-13b

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • Llama-2-13b-chat

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • Llama-2-70b

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • Llama-2-70b-chat

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • Llama-2-7b

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • Llama-2-7b-chat

    Meta has developed and publicly released the Llama 2 family of large language models (LLMs), a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. Our fine-tuned LLMs, called Llama-2-Chat, are optimized for dialogue use cases. L...

  • mask_rcnn_swin-t-p4-w7_fpn_1x_coco

    This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual ...

  • MedImageInsight

    Most medical imaging AI today is narrowly built to detect a small set of individual findings on a single modality like chest X-rays. This training approach is data- and computationally inefficient, requiring ~6-12 months per finding1, and often fails to generalize in real world environments. By f...

  • MedImageParse

    Biomedical image analysis is fundamental for biomedical discovery in cell biology, pathology, radiology, and many other biomedical domains. MedImageParse is a biomedical foundation model for imaging parsing that can jointly conduct segmentation, detection, and recognition across 9 imaging modalit...

  • Meteor-Score-Evaluator

    | | | | -- | -- | | Score range | Float [0-1]: higher means better quality. | | What is this metric? | METEOR score measures the similarity by shared n-grams between the generated text and the ground truth, similar to the BLEU score, focusing on precision and recall. But it addresses limitations ...

  • microsoft-beit-base-patch16-224-pt22k-ft22k

    BEiT (Bidirectional Encoder representation from Image Transformers) is a vision transformer(ViT) pre-trained with Masked Image Modeling(MIM), which is a self-supervised pre-training inspired by BERT from NLP, followed by Intermediate fine-tuning using ImageNet-22k dataset. It is then fine-tuned f...

  • microsoft-deberta-base

    DeBERTa (Decoding-enhanced BERT with Disentangled Attention) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. With those two improvements, DeBERTa out perform RoBERTa on a majority of NLU tasks with 80GB training data...

  • microsoft-deberta-base-mnli

    DeBERTa (Decoding-enhanced BERT with Disentangled Attention) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. It outperforms BERT and RoBERTa on majority of NLU tasks with 80GB training data.

Please check the [offi...

  • microsoft-deberta-large

    DeBERTa (Decoding-enhanced BERT with Disentangled Attention) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. With those two improvements, DeBERTa out perform RoBERTa on a majority of NLU tasks with 80GB training data...

  • microsoft-deberta-large-mnli

    DeBERTa (Decoding-enhanced BERT with Disentangled Attention) improves the BERT and RoBERTa models using disentangled attention and enhanced mask decoder. It outperforms BERT and RoBERTa on majority of NLU tasks with 80GB training data.

Please check the [offi...

Large Language and Vision Assistant for bioMedicine (i.e., “LLaVA-Med”) is a large language and vision model trained using a curriculum l...

  • microsoft-Orca-2-13b

    Orca 2 is a finetuned version of LLAMA-2. Orca 2’s training data is a synthetic dataset that was created to enhance the small model’s reasoning abilities. All synthetic training data was moderated using the Microsoft Azure content filters. More details about the model can be found in the [Orca 2 ...

  • microsoft-Orca-2-7b

    Orca 2 is a finetuned version of LLAMA-2. Orca 2’s training data is a synthetic dataset that was created to enhance the small model’s reasoning abilities. All synthetic training data was moderated using the Microsoft Azure content filters. More details about the model can be found in the [Orca 2 ...

  • microsoft-phi-1-5

    Microsoft Phi-1.5

Phi-1.5 is a Transformer-based language model with 1.3 billion parameters. It was trained on a combination of data sources, including an additional source of NLP synthetic texts. Phi-1.5 performs exceptionally well on benchmarks testing common sense, language understandi...

The phi-2 is a language model with 2.7 billion parameters. The phi-2 model was trained using the same data sources as phi-1, augmented with a new data source that consists of various NLP synthetic texts and filtered websites (for safety and educational value). When assesse...

Mixtral-8x22B-v0.1 is a pretrained base model and therefore does not have any moderation mechanisms.

Evaluation Results

[Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/H...

Mistral-7B-v0.2 has the following changes compared to Mistral-7B-v0.1:

  • 32k context window (vs 8k context in v0.1)
  • Rope-theta = 1e6
  • No Sliding-Window Attention

For full details...

Mistral-7B-v0.3 has the following changes compared to Mistral-7B-v0.2

The Mistral-7B-Instruct-v0.1 Large Language Model (LLM) is a instruct fine-tuned version of the Mistral-7B-v0.1 generative text model using a variety of publicly available conversation datasets.

For full details of this mod...

The Mistral-7B-v0.1 Large Language Model (LLM) is a pretrained generative text model with 7 billion parameters. Mistral-7B-v0.1 outperforms Llama 2 13B on all benchmarks tested.

For full details of this model please read paper and [releas...

Inference samples

Inference type Python sample (Notebook) CLI with YAML
Real time <a href="https://aka.ms/...

Mixtral-8x22B-v0.1 is a pretrained base model and therefore does not have any moderation mechanisms.

Evaluation Results

[Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/H...

  • mistralai-Mixtral-8x7B-Instruct-v01

    The Mixtral-8x7B Large Language Model (LLM) is a pretrained generative Sparse Mixture of Experts. The Mixtral-8x7B outperforms Llama 2 70B on most benchmarks with 6x faster inference.

Mixtral-8x7B-v0.1 is a decoder-only model with 8 distinct groups or the "experts". At every layer, for every tok...

The Mixtral-8x7B-v0.1 Large Language Model (LLM) is a pretrained generative text model with 7 billion parameters. Mixtral-8x7B-v0.1 outperforms Llama 2 70B on most benchmarks with 6x faster inference.

For full details of this model please read [release blog post](https://mi...

It can be used for multi-class and multi-label multimodal classification tasks, and is capable of handling datasets with features from diverse modes, includ...

  • multi-index-rerank-qna

    This "Multi-Source Rerank Q&A" demonstrates Q&A application, enabled by reranking data from multiple sources and powered by GPT. It utilizes indexed files and the rerank tool from Azure Machine Learning to provide grounded answers. You can ask a wide range of questions and receive responses based...

  • ocsort_yolox_x_crowdhuman_mot17-private-half

    ocsort_yolox_x_crowdhuman_mot17-private-half model is from OpenMMLab's MMTracking library. Multi-Object Tracking (MOT) has rapidly progressed with the development of object detection and re-identification. Howev...

  • OpenAI-CLIP-Image-Text-Embeddings-vit-base-patch32

    OpenAI's CLIP (Contrastive Language–Image Pre-training) model was designed to investigate the factors that contribute to the robustness of computer vision tasks. It can seamlessly adapt to a range of image classification tasks without requiring specific training for each, demonstrating efficiency...

  • OpenAI-CLIP-Image-Text-Embeddings-ViT-Large-Patch14-336

    The CLIP model was developed by researchers at OpenAI to learn about what contributes to robustness in computer vision tasks. The model was also developed to test the ability of models to generalize to arbitrary image classification tasks in a zero-shot manner. It was not developed for general ...

  • openai-clip-vit-base-patch32

    OpenAI's CLIP (Contrastive Language–Image Pre-training) model was designed to investigate the factors that contribute to the robustness of computer vision tasks. It can seamlessly adapt to a range of image classification tasks without requiring specific training for each, demonstrating efficiency...

  • openai-clip-vit-large-patch14

    OpenAI's CLIP (Contrastive Language–Image Pre-training) model was designed to investigate the factors that contribute to the robustness of computer vision tasks. It can seamlessly adapt to a range of image classification tasks without requiring specific training for each, demonstrating efficiency...

  • openai-whisper-large

    Whisper is an OpenAI pre-trained speech recognition model with potential applications for ASR solutions for developers. However, due to weak supervision and large-scale noisy data, it should be used with caution in high-risk domains. The model has been trained on 680k hours of audio data represen...

  • openai-whisper-large-v3

    Whisper is a model that can recognize and translate speech using deep learning. It was trained on a large amount of data from different sources and languages. Whisper models can handle various tasks and domains without needing to adjust the model.

Whisper large-v3 is similar to the previous larg...

  • Phi-3-medium-128k-instruct

    The Phi-3-Medium-128K-Instruct is a 14B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model belongs to the Ph...

  • Phi-3-medium-4k-instruct

    The Phi-3-Medium-4K-Instruct is a 14B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model belongs to the Phi-...

  • Phi-3-mini-128k-instruct

    The Phi-3-Mini-128K-Instruct is a 3.8 billion-parameter, lightweight, state-of-the-art open model trained using the Phi-3 datasets. This dataset includes both synthetic data and filtered publicly available website data, with an emphasis on high-quality and reasoning-dense properties.

After initi...

  • Phi-3-mini-4k-instruct

    The Phi-3-Mini-4K-Instruct is a 3.8B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model belongs to the Phi-3...

  • Phi-3-small-128k-instruct

    The Phi-3-Small-128K-Instruct is a 7B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model supports 128K conte...

  • Phi-3-small-8k-instruct

    The Phi-3-Small-8K-Instruct is a 7B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model supports 8K context l...

  • Phi-3-vision-128k-instruct

    Model Summary

Phi-3 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 mo...

  • Phi-3.5-mini-instruct

    Phi-3.5-mini is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available websites - with a focus on very high-quality, reasoning dense data. The model belongs to the Phi-3 model family and supports 128K token context length. Th...

  • Phi-3.5-MoE-instruct

    Phi-3.5-MoE is a lightweight, state-of-the-art open model built upon datasets used for Phi-3 - synthetic data and filtered publicly available documents - with a focus on very high-quality, reasoning dense data. The model supports multilingual and comes with 128K context length (in tokens). The mo...

  • Phi-3.5-vision-instruct

    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 t...

  • playground-ayod-rag

    This flow template is an advanced RAG flow modeled on the implementation of Azure AI Playground - on Your Data. The flow consists of tools that rewrites user query input into one or more queries based on chat history context using LLM, retrieves data for rewritten queries from the data index and ...

  • Prism

    PRISM is a multi-modal generative foundation model for slide-level analysis of H&E-stained histopathology images. Utilizing Virchow tile embeddings and clinical report texts for pre-training, PRISM combines these embeddings into a single slide embedding and generates a text-based diagnostic repor...

  • projecte-aina-aguila-7b

    Model Description

Aguila-7b

Table of Contents

Click to expand

FLOR-1.3B

Table of Contents

Click to expand

FLOR-1.3B Instructed

Table of Contents

Click to expand

FLOR-6.3B

Table of Contents

Click to expand

FLOR-6.3B Instructed

Table of Contents

Click to expand

Protected material is any text that is under copyright, including song lyrics, recipes, and articles. Protected material evaluation leverages the Azure AI Content Safety Protected Material for Text service to perform the classification.

Labeling

Protected Material evaluations ...

Digital pathology poses unique computational challenges, as a standard gigapixel slide may comprise tens of thousands of image tiles[^1],[^2],[^3]. Previous models often rely predominantly on tile-level predictions, which can overlook critical slide-level context and spatial depen...

  • qna-ada-similarity-eval

    The "QnA Ada Similarity Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to...

  • qna-coherence-eval

    The "QnA Coherence Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to achi...

  • qna-f1-score-eval

    The "QnA F1 Score Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems using f1 score based on the word counts in predicted answer and ground truth.

Inference samples

Inference type CLI VS Code Extension
Real time <a href="https://microsoft.github.io...

  • qna-fluency-eval

    The "QnA Fluency Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to achiev...

  • qna-gpt-similarity-eval

    The "QnA GPT Similarity Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to...

  • qna-groundedness-eval

    The "QnA Groundedness Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to a...

  • qna-non-rag-metrics-eval

    The Q&A evaluation flow will evaluate the Q&A systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT and GPT embedding model to assist with measurements aims to achieve a high agreement with human evaluations compa...

  • qna-quality-safety-eval

    The Q&A quality and safety evaluation flow will evaluate the Q&A systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT and GPT embedding model to assist with measurements aims to achieve a high agreement with huma...

  • qna-rag-metrics-eval

    The Q&A RAG (Retrieval Augmented Generation) evaluation flow will evaluate the Q&A RAG systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses . Utilizing GPT model to assist with measurements aims to achieve a high agreement with...

  • qna-relevance-eval

    The "QnA Relevance Evaluation" is a model to evaluate the Q&A Retrieval Augmented Generation systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT-3.5 as the Language Model to assist with measurements aims to achi...

  • qna-with-your-own-data-using-faiss-index

    The "QnA with Your Own Data Using Faiss Index" is a Q&A model with GPT3.5 using information from vector search to make the answer more grounded. It involves embedding user's question with LLM, and then using Faiss Index Lookup to find relevant documents based on vectors. By utilizing vector searc...

  • rai-eval-ui-dag-flow

    The Q&A quality and safety evaluation flow will evaluate the Q&A systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT and GPT embedding model to assist with measurements aims to achieve a high agreement with huma...

  • rai-qna-quality-safety-eval

    The Q&A quality and safety evaluation flow will evaluate the Q&A systems by leveraging the state-of-the-art Large Language Models (LLM) to measure the quality and safety of your responses. Utilizing GPT and GPT embedding model to assist with measurements aims to achieve a high agreement with huma...

  • Relevance-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Coherence measures the logical and orderly presentation of ideas in a response, allowing the reader to easily follow and understand the writer's train of thought. A c...

  • rerank-qna

    This "Index Data Rerank Q&A" demonstrates Q&A application, enabled by reranking data from vector index stores and powered by GPT. It utilizes index stores and the rerank tool from Azure Machine Learning to provide grounded answers. You can ask a wide range of questions and receive responses based...

  • Retrieval-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Retrieval measures the quality of search without ground truth. It focuses on how relevant the context chunks (encoded as a string) are to address a query and how the ...

  • roberta-base

    RoBERTa is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate i...

  • roberta-base-openai-detector

    The RoBERTa base OpenAI Detector functions as a model designed to detect outputs generated by the GPT-2 model. It was created by refining a RoBERTa base model using the outputs of the 1.5B-parameter GPT-2 model. This detector is utilized to determine whether text was generated by a GPT-2 model. O...

  • roberta-large

    RoBERTa is a transformers model pretrained on a large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate i...

  • roberta-large-mnli

    roberta-large-mnli is the RoBERTa large model fine-tuned on the Multi-Genre Natural Language Inference (MNLI) corpus. The model is a pretrained model on English language text using a masked language modeling ...

  • roberta-large-openai-detector

    RoBERTa large OpenAI Detector is the GPT-2 output detector model, obtained by fine-tuning a RoBERTa large model with the outputs of the 1.5B-parameter GPT-2 model. The model can be used to predict if text was generated by a GPT-2 model. This model was released by OpenAI at the same time as Op...

  • Rouge-Score-Evaluator

    | | | | -- | -- | | Score range | Float [0-1]: higher means better quality. | | What is this metric? | ROUGE (Recall-Oriented Understudy for Gisting Evaluation) is a set of metrics used to evaluate automatic summarization and machine translation. It measures the overlap between generated text and...

  • runwayml-stable-diffusion-inpainting

    Stable Diffusion Inpainting is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input, with the extra capability of inpainting the pictures by using a mask.

The Stable-Diffusion-Inpainting was initialized with the weights of the Stable-Diffus...

  • runwayml-stable-diffusion-v1-5

    Stable Diffusion is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input. The Stable-Diffusion-v1-5 checkpoint was initialized with the weights of the Stable-Diffusion-v1-2 checkpoint and subsequently fine-tuned on 595k steps at resolution 5...

  • Salesforce-BLIP-2-opt-2-7b-image-to-text

The BLIP-2 model, utilizing OPT-2.7b (a large language model with 2.7 billion parameters), is presented in the paper titled "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models". T...

  • Salesforce-BLIP-2-opt-2-7b-vqa

    The BLIP-2 model, utilizing OPT-2.7b (a large language model with 2.7 billion parameters), is presented in the paper titled "BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models". Th...

  • Salesforce-BLIP-image-captioning-base

    BLIP (Bootstrapping Language-Image Pre-training) designed for unified vision-language understanding and generation is a new VLP framework that expands the scope of downstream tasks compared to existing methods. The framework encompasses two key contributions from both model and data perspective...

  • Salesforce-BLIP-vqa-base

    BLIP (Bootstrapping Language-Image Pre-training) designed for unified vision-language understanding and generation is a new VLP framework that expands the scope of downstream tasks compared to existing methods. The framework encompasses two key contributions from both model and data perspective...

  • Self-Harm-Related-Content-Evaluator

    Definition

Self-harm-related content includes language pertaining to actions intended to hurt, injure, or damage one's body or kill oneself.

Severity scale

Safety evaluations annotate self-harm-related content using a 0-7 scale.

Very Low (0-1) refers to

  • Content that contains self-...

Sexual content includes language pertaining to anatomical organs and genitals, romantic relationships, acts portrayed in erotic terms, pregnancy, physical sexual acts (including assault or sexual violence), prostitution, pornography, and sexual abuse.

Severity scale

Safety eva...

  • Similarity-Evaluator

    | | | | -- | -- | | Score range | Integer [1-5]: 1 is the lowest quality and 5 is the highest quality. | | What is this metric? | Similarity measures the degrees of similarity between the generated text and its ground truth with respect to a query. | | How does it work? | The similarity metric i...

  • snowflake-arctic-base

    Model Overview

Arctic is a dense-MoE Hybrid transformer architecture pre-trained from scratch by the Snowflake AI Research Team. We are releasing model checkpoints for both the base and instruct-tuned versions of Arctic under an Apache-2.0 license. This means you can use them freely in your ow...

Arctic is a dense-MoE Hybrid transformer architecture pre-trained from scratch by the Snowflake AI Research Team. We are releasing model checkpoints for both the base and instruct-tuned versions of Arctic under an Apache-2.0 license. This means you can use them freely in your ow...

The mod...

With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to B...

  • t5-large

    The developers of the Text-To-Text Transfer Transformer (T5) write:

With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to B...

  • t5-small

    The developers of the Text-To-Text Transfer Transformer (T5) write:

With T5, we propose reframing all NLP tasks into a unified text-to-text-format where the input and output are always text strings, in contrast to B...

  • template-chat-flow

    The "Template Chat Flow" is a chat model using GPT3.5 that generates the next message based on the conversation history and the latest chat content.

Inference samples

Inference type CLI VS Code Extension
Real time <a href="https://microsoft.github.io/promptflow/how-to-guides/dep...

  • template-eval-flow

    The "Template Evaluation Flow" is a evaluate model to measure how well the output matches the expected criteria and goals.

Inference samples

Inference type CLI VS Code Extension
Real time <a href="https://microsoft.github.io/promptflow/how-to-guides/deploy-a-flow/index.html" tar...

  • template-standard-flow

    The "Template Standard Flow" is a model using GPT3.5 to generate a joke based on user input.

Inference samples

Inference type CLI VS Code Extension
Real time deploy-promptflow...

Falcon-40B is a large language model (LLM) developed by the Technology Innovation Institute (TII) with 40 billion parameters. It is a causal decoder-only model trained on 1 trillion tokens from the RefinedWeb dataset, enhanced with curated corpora. Falcon-40B supports English, Germa...

Falcon-40B-Instruct is a large language model with 40 billion parameters, developed by TII. It is a causal decoder-only model fine-tuned on a mixture of Baize data and is released under the Apache 2.0 license. This model is optimized for inference and features FlashAttention and mul...

Falcon-7B is a large language model with 7 billion parameters. It is a causal decoder-only model developed by TII and trained on 1,500 billion tokens of RefinedWeb dataset, which was enhanced with curated corpora. The model is available under the Apache 2.0 license. It outperforms c...

Falcon-7B-Instruct is a large language model with 7 billion parameters, developed by TII. It is a causal decoder-only model and is released under the Apache 2.0 license. This model is optimized for inference and features FlashAttention and multiquery architectures. It is primarily d...

Violent content includes language pertaining to physical actions intended to hurt, injure, damage, or kill someone or something. It also includes descriptions of weapons and guns (and related entities such as manufacturers and associations).

Severity scale

Safety evaluations ...

  • Virchow

    Virchow is a self-supervised vision transformer pretrained using 1.5M whole slide histopathology images. The model can be used as a tile-level feature extractor (frozen or finetuned) to achieve state-of-the-art results for a wide variety of downstream computational pathology use cases.

Model ...

  • Virchow2

    Virchow2 is a self-supervised vision transformer pretrained using 3.1M whole slide histopathology images. The model can be used as a tile-level feature extractor (frozen or finetuned) to achieve state-of-the-art results for a wide variety of downstream computational pathology use cases.

Mode...

  • web-classification

    The "Web Classification" is a model demonstrating multi-class classification with LLM. Given an url, it will classify the url into one web category with just a few shots, simple summarization and classification prompts.

Inference samples

Inference type CLI VS Code Extension
Real...
⚠️ **GitHub.com Fallback** ⚠️