Azure
/
azureml-assets
зеркало из https://github.com/Azure/azureml-assets.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

fix (#3270) 

Co-authored-by: Ali Soylemezoglu <alisoy@microsoft.com>
This commit is contained in:
Ali Soylemezoglu 2024-08-16 20:13:51 -04:00 коммит произвёл GitHub
Родитель 13262bcd98
Коммит a602ada5fa
Не найден ключ, соответствующий данной подписи
Идентификатор ключа GPG: B5690EEEBB952194
24 изменённых файлов: 655 добавлений и 1141 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

193
assets/models/system/phi-3-medium-128k-instruct/description.md
Просмотреть файл

@ -1,189 +1,26 @@
## Model Summary
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 Phi-3 family with the Medium version in two variants [4k](https://huggingface.co/microsoft/Phi-3-medium-4k-instruct) and [128K](https://huggingface.co/microsoft/Phi-3-medium-128k-instruct) which is the context length (in tokens) that it can support.
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 supports 128K context length (in tokens).
The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
The model underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3-Medium-128K-Instruct showcased a robust and state-of-the-art performance among models of the same-size and next-size-up.
Resources and Technical Documentation:
## Resources
+ [Phi-3 Microsoft Blog](https://aka.ms/phi3blog-april)
+ [Phi-3 Technical Report](https://aka.ms/phi3-tech-report)
🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3) <br>
📰 [Phi-3 Microsoft Blog](https://aka.ms/Phi-3Build2024) <br>
📖 [Phi-3 Technical Report](https://aka.ms/phi3-tech-report) <br>
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai) <br>
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook) <br>
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3-Medium-128k-Instruct has 14B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications which require :
## Training Datasets
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
The model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.
**Use case considerations**
The model is 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.
Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.
## Training
### Model
* Architecture: Phi-3-Medium-128k-Instruct has 14B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 128k tokens
* GPUs: 512 H100-80G
* Training time: 42 days
* Training data: 4.8T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between February and April 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates: The model weight is released on May 21, 2024.
### Datasets
The training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
Our training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results for Phi-3-Medium-128k-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x22b, Gemini-Pro, Command R+ 104B, Llama-3-70B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106(Chat).
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Medium-128k-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct<br>8b|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|AGI Eval<br>5-shot|49.7|50.1|54.0|56.9|48.4|49.0|59.6|
|MMLU<br>5-shot|76.6|73.8|76.2|80.2|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|77.9|74.1|81.8|80.4|68.3|75.6|87.7|
|ANLI<br>7-shot|57.3|63.4|65.2|68.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|81.6|78.0|79.0|82.6|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|91.0|86.9|91.3|93.0|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.6|95.7|96.9|98.2|96.3|96.1|98.8|
|BoolQ<br>2-shot|86.5|86.1|82.7|89.1|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|82.2|82.0|82.0|84.4|79.6|81.8|86.7|
|MedQA<br>2-shot|67.6|59.2|67.9|78.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.2|86.8|88.6|91.8|86.0|86.4|93.4|
|PIQA<br>5-shot|87.8|86.4|85.0|85.3|86.6|86.2|90.1|
|Social IQA<br>5-shot|79.0|75.3|78.2|81.1|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|74.3|57.8|67.4|81.9|67.7|72.6|85.2|
|WinoGrande<br>5-shot|78.9|77.0|75.3|83.3|68.8|72.2|86.7|
|TriviaQA<br>5-shot|73.9|82.8|84.5|78.5|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|87.5|78.3|83.8|93.5|78.1|80.4|94.2|
|HumanEval<br>0-shot|58.5|61.6|39.6|78.7|62.2|64.4|79.9|
|MBPP<br>3-shot|73.8|68.9|70.7|81.3|77.8|73.2|86.7|
|Average|77.3|75.0|76.3|82.5|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Medium-128k-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct<br>8b|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
| Popular aggregated benchmark | 72.3 | 69.9 | 73.4 | 76.3 | 67.0 | 67.5 | 80.5 |
| Reasoning | 83.2 | 79.3 | 81.5 | 86.7 | 78.3 | 80.4 | 89.3 |
| Language understanding | 75.3 | 75.7 | 78.7 | 77.9 | 70.4 | 75.3 | 81.6 |
| Code generation | 64.2 | 68.6 | 60.0 | 69.3 | 70.4 | 66.7 | 76.1 |
| Math | 52.9 | 45.3 | 52.5 | 59.7 | 52.8 | 50.9 | 67.1 |
| Factual knowledge | 47.5 | 60.3 | 60.6 | 52.4 | 63.4 | 54.6 | 45.9 |
| Multilingual | 62.2 | 67.8 | 69.8 | 62.0 | 67.0 | 73.4 | 78.2 |
| Robustness | 70.2 | 57.9 | 65.5 | 78.7 | 69.3 | 69.7 | 84.6 |
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [DeepSpeed](https://github.com/microsoft/DeepSpeed)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
## Hardware
Note that by default, the Phi-3-Medium model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
+ Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [128k](https://huggingface.co/microsoft/Phi-3-medium-128k-instruct-onnx)
## Cross Platform Support
ONNX runtime ecosystem now supports Phi3 Medium models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 Medium across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via RTN
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

45
assets/models/system/phi-3-medium-128k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,45 @@
We report the results for Phi-3-Medium-128k-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x22b, Gemini-Pro, Command R+ 104B, Llama-3-70B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106(Chat).
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Medium-128k-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|AGI Eval<br>5-shot|49.7|50.1|54.0|56.9|48.4|49.0|59.6|
|MMLU<br>5-shot|76.6|73.8|76.2|80.2|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|77.9|74.1|81.8|80.4|68.3|75.6|87.7|
|ANLI<br>7-shot|57.3|63.4|65.2|68.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|81.6|78.0|79.0|82.6|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|91.0|86.9|91.3|93.0|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.6|95.7|96.9|98.2|96.3|96.1|98.8|
|BoolQ<br>2-shot|86.5|86.1|82.7|89.1|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|82.2|82.0|82.0|84.4|79.6|81.8|86.7|
|MedQA<br>2-shot|67.6|59.2|67.9|78.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.2|86.8|88.6|91.8|86.0|86.4|93.4|
|PIQA<br>5-shot|87.8|86.4|85.0|85.3|86.6|86.2|90.1|
|Social IQA<br>5-shot|79.0|75.3|78.2|81.1|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|74.3|57.8|67.4|81.9|67.7|72.6|85.2|
|WinoGrande<br>5-shot|78.9|77.0|75.3|83.3|68.8|72.2|86.7|
|TriviaQA<br>5-shot|73.9|82.8|84.5|78.5|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|87.5|78.3|83.8|93.5|78.1|80.4|94.2|
|HumanEval<br>0-shot|58.5|61.6|39.6|78.7|62.2|64.4|79.9|
|MBPP<br>3-shot|73.8|68.9|70.7|81.3|77.8|73.2|86.7|
|Average|77.3|75.0|76.3|82.5|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Medium-128k-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
| Popular aggregated benchmark | 72.3 | 69.9 | 73.4 | 76.3 | 67.0 | 67.5 | 80.5 |
| Reasoning | 83.2 | 79.3 | 81.5 | 86.7 | 78.3 | 80.4 | 89.3 |
| Language understanding | 75.3 | 75.7 | 78.7 | 77.9 | 70.4 | 75.3 | 81.6 |
| Code generation | 64.2 | 68.6 | 60.0 | 69.3 | 70.4 | 66.7 | 76.1 |
| Math | 52.9 | 45.3 | 52.5 | 59.7 | 52.8 | 50.9 | 67.1 |
| Factual knowledge | 47.5 | 60.3 | 60.6 | 52.4 | 63.4 | 54.6 | 45.9 |
| Multilingual | 62.2 | 67.8 | 69.8 | 62.0 | 67.0 | 73.4 | 78.2 |
| Robustness | 70.2 | 57.9 | 65.5 | 78.7 | 69.3 | 69.7 | 84.6 |

49
assets/models/system/phi-3-medium-128k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-medium-128k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 96|4|880|256
finetune-recommended-sku: Standard_NC96ads_A100_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 medium instruct (128k)"
summary: "Same Phi-3-medium model, but with a larger context size for RAG or few shot prompting."

193
assets/models/system/phi-3-medium-4k-instruct/description.md
Просмотреть файл

@ -1,189 +1,26 @@
## Model Summary
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-3 family with the Medium version in two variants [4K](https://huggingface.co/microsoft/Phi-3-medium-4k-instruct) and [128K](https://huggingface.co/microsoft/Phi-3-medium-128k-instruct) which is the context length (in tokens) that it can support.
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 supports 4K context length (in tokens).
The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
The model underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3-Medium-4K-Instruct showcased a robust and state-of-the-art performance among models of the same-size and next-size-up.
Resources and Technical Documentation:
## Resources
+ [Phi-3 Microsoft Blog](https://aka.ms/phi3blog-april)
+ [Phi-3 Technical Report](https://aka.ms/phi3-tech-report)
🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3) <br>
📰 [Phi-3 Microsoft Blog](https://aka.ms/Phi-3Build2024) <br>
📖 [Phi-3 Technical Report](https://aka.ms/phi3-tech-report) <br>
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai) <br>
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook) <br>
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3-Medium-4K-Instruct has 14B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications which require:
## Training Datasets
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
The model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.
**Use case considerations**
The 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.
Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.
## Training
### Model
* Architecture: Phi-3-Medium-4K-Instruct has 14B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 4K tokens
* GPUs: 512 H100-80G
* Training time: 42 days
* Training data: 4.8T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between February and April 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates: The model weight is released on May 21, 2024.
### Datasets
The training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
Our training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results for Phi-3-Medium-4K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x22b, Gemini-Pro, Command R+ 104B, Llama-3-70B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106(Chat).
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Medium-4K-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct<br>8b|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|AGI Eval<br>5-shot|50.2|50.1|54.0|56.9|48.4|49.0|59.6|
|MMLU<br>5-shot|78.0|73.8|76.2|80.2|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|81.4|74.1|81.8|80.4|68.3|75.6|87.7|
|ANLI<br>7-shot|55.8|63.4|65.2|68.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|82.4|78.0|79.0|82.6|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|91.6|86.9|91.3|93.0|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.7|95.7|96.9|98.2|96.3|96.1|98.8|
|BoolQ<br>2-shot|86.5|86.1|82.7|89.1|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|82.8|82.0|82.0|84.4|79.6|81.8|86.7|
|MedQA<br>2-shot|69.9|59.2|67.9|78.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.4|86.8|88.6|91.8|86.0|86.4|93.4|
|PIQA<br>5-shot|87.9|86.4|85.0|85.3|86.6|86.2|90.1|
|Social IQA<br>5-shot|80.2|75.3|78.2|81.1|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|75.1|57.8|67.4|81.9|67.7|72.6|85.2|
|WinoGrande<br>5-shot|81.5|77.0|75.3|83.3|68.8|72.2|86.7|
|TriviaQA<br>5-shot|73.9|82.8|84.5|78.5|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|91.0|78.3|83.8|93.5|78.1|80.4|94.2|
|HumanEval<br>0-shot|62.2|61.6|39.6|78.7|62.2|64.4|79.9|
|MBPP<br>3-shot|75.2|68.9|70.7|81.3|77.8|73.2|86.7|
|Average|78.5|75.0|76.3|82.5|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Medium-4K-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct<br>8b|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|75.4|69.9|73.4|76.3|67.0|67.5|80.5|
|Reasoning|84.1|79.3|81.5|86.7|78.3|80.4|89.3|
|Language understanding|73.9|75.6|78.1|76.9|68.7|76.2|80.7|
|Code generation|66.1|68.6|60.0|69.3|70.4|66.7|76.1|
|Math|52.8|45.3|52.5|59.7|52.8|50.9|67.1|
|Factual knowledge|48.3|60.3|60.6|52.4|63.4|54.6|45.9|
|Multilingual|62.9|67.8|69.8|62.0|67.0|73.4|78.2|
|Robustness|66.5|57.9|65.5|78.7|69.3|69.7|84.6|
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [DeepSpeed](https://github.com/microsoft/DeepSpeed)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
## Hardware
Note that by default, the Phi-3-Medium model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
+ Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [4K](https://huggingface.co/microsoft/Phi-3-medium-4k-instruct-onnx)
## Cross Platform Support
ONNX runtime ecosystem now supports Phi3 Medium models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 Medium across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via RTN
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

45
assets/models/system/phi-3-medium-4k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,45 @@
We report the results for Phi-3-Medium-4K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x22b, Gemini-Pro, Command R+ 104B, Llama-3-70B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106(Chat).
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Medium-4K-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|AGI Eval<br>5-shot|50.2|50.1|54.0|56.9|48.4|49.0|59.6|
|MMLU<br>5-shot|78.0|73.8|76.2|80.2|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|81.4|74.1|81.8|80.4|68.3|75.6|87.7|
|ANLI<br>7-shot|55.8|63.4|65.2|68.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|82.4|78.0|79.0|82.6|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|91.6|86.9|91.3|93.0|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.7|95.7|96.9|98.2|96.3|96.1|98.8|
|BoolQ<br>2-shot|86.5|86.1|82.7|89.1|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|82.8|82.0|82.0|84.4|79.6|81.8|86.7|
|MedQA<br>2-shot|69.9|59.2|67.9|78.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.4|86.8|88.6|91.8|86.0|86.4|93.4|
|PIQA<br>5-shot|87.9|86.4|85.0|85.3|86.6|86.2|90.1|
|Social IQA<br>5-shot|80.2|75.3|78.2|81.1|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|75.1|57.8|67.4|81.9|67.7|72.6|85.2|
|WinoGrande<br>5-shot|81.5|77.0|75.3|83.3|68.8|72.2|86.7|
|TriviaQA<br>5-shot|73.9|82.8|84.5|78.5|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|91.0|78.3|83.8|93.5|78.1|80.4|94.2|
|HumanEval<br>0-shot|62.2|61.6|39.6|78.7|62.2|64.4|79.9|
|MBPP<br>3-shot|75.2|68.9|70.7|81.3|77.8|73.2|86.7|
|Average|78.5|75.0|76.3|82.5|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Medium-4K-Instruct<br>14b|Command R+<br>104B|Mixtral<br>8x22B|Llama-3-70B-Instruct|GPT3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|75.4|69.9|73.4|76.3|67.0|67.5|80.5|
|Reasoning|84.1|79.3|81.5|86.7|78.3|80.4|89.3|
|Language understanding|73.9|75.6|78.1|76.9|68.7|76.2|80.7|
|Code generation|66.1|68.6|60.0|69.3|70.4|66.7|76.1|
|Math|52.8|45.3|52.5|59.7|52.8|50.9|67.1|
|Factual knowledge|48.3|60.3|60.6|52.4|63.4|54.6|45.9|
|Multilingual|62.9|67.8|69.8|62.0|67.0|73.4|78.2|
|Robustness|66.5|57.9|65.5|78.7|69.3|69.7|84.6|

49
assets/models/system/phi-3-medium-4k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-medium-4k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 96|4|880|256
finetune-recommended-sku: Standard_NC96ads_A100_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 medium instruct (4k)"
summary: "A 14B parameters model, proves better quality than Phi-3-mini, with a focus on high-quality, reasoning-dense data."

235
assets/models/system/phi-3-mini-128k-instruct/description.md
Просмотреть файл

@ -1,245 +1,26 @@
## Model Summary
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.
The model belongs to the Phi-3 family with the Mini version in two variants 4K and 128K which is the context length (in tokens) that it can support.
After initial training, the model underwent a post-training process that involved supervised fine-tuning and direct preference optimization to enhance its ability to follow instructions and adhere to safety measures.
When evaluated against benchmarks that test common sense, language understanding, mathematics, coding, long-term context, and logical reasoning, the Phi-3 Mini-128K-Instruct demonstrated robust and state-of-the-art performance among models with fewer than 13 billion parameters.
Resources and Technical Documentation:
## Resources
🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3) <br>
📰 [Phi-3 Microsoft Blog](https://aka.ms/Phi-3Build2024) <br>
📖 [Phi-3 Technical Report](https://aka.ms/phi3-tech-report) <br>
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai) <br>
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook) <br>
🖥️ [Try It](https://aka.ms/try-phi3)
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3 Mini-128K-Instruct has 3.8B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for commercial and research use in English. The model provides uses for applications which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
Our model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.
**Use case considerations**
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.
***Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.***
## Release Notes
This is an update over the original instruction-tuned Phi-3-mini release based on valuable customer feedback.
The model used additional post-training data leading to substantial gains on long-context understanding, instruction following, and structure output.
We also improve multi-turn conversation quality, explicitly support <|system|> tag, and significantly improve reasoning capability.
We believe most use cases will benefit from this release, but we encourage users to test in their particular AI applications.
We appreciate the enthusiastic adoption of the Phi-3 model family, and continue to welcome all feedback from the community.
These tables below highlights improvements on instruction following, structure output, reasoning, and long-context understanding of the new release on our public and internal benchmark datasets.
| Benchmarks | Original | June 2024 Update |
|------------|----------|------------------|
| Instruction Extra Hard | 5.7 | 5.9 |
| Instruction Hard | 5.0 | 5.2 |
| JSON Structure Output | 1.9 | 60.1 |
| XML Structure Output | 47.8 | 52.9 |
| GPQA | 25.9 | 29.7 |
| MMLU | 68.1 | 69.7 |
| **Average** | **25.7** | **37.3** |
RULER: a retrieval-based benchmark for long context understanding
| Model | 4K | 8K | 16K | 32K | 64K | 128K | Average |
|-------------------|------|------|------|------|------|------|---------|
| Original | 86.7 | 78.1 | 75.6 | 70.3 | 58.9 | 43.3 | **68.8** |
| June 2024 Update | 92.4 | 91.1 | 90.8 | 87.9 | 79.8 | 65.6 | **84.6** |
RepoQA: a benchmark for long context code understanding
| Model | Python | C++ | Rust | Java | TypeScript | Average |
|-------------------|--------|-----|------|------|------------|---------|
| Original | 27 | 29 | 40 | 33 | 33 | **32.4** |
| June 2024 Update | 85 | 63 | 72 | 93 | 72 | **77** |
### Tokenizer
Phi-3 Mini-128K-Instruct supports a vocabulary size of up to `32064` tokens.
## Training
### Model
* Architecture: Phi-3 Mini-128K-Instruct has 3.8B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 128K tokens
* GPUs: 512 H100-80G
* Training time: 10 days
* Training data: 4.9T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between May and June 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates: June, 2024.
### Datasets
## Training Datasets
Our training data includes a wide variety of sources, totaling 4.9 trillion tokens, and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results under completion format for Phi-3-Mini-128K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mistral-7b-v0.1, Mixtral-8x7b, Gemma 7B, Llama-3-8B-Instruct, and GPT-3.5.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
| Category | Benchmark | Phi-3-Mini-128K-Ins | Gemma-7B | Mistral-7B | Mixtral-8x7B | Llama-3-8B-Ins | GPT3.5-Turbo-1106 |
|----------|-----------|---------------------|----------|------------|--------------|----------------|-------------------|
| Popular aggregated benchmark | AGI Eval <br>5-shot| 39.5 | 42.1 | 35.1 | 45.2 | 42 | 48.4 |
| | MMLU <br>5-shot | 69.7 | 63.6 | 61.7 | 70.5 | 66.5 | 71.4 |
| | BigBench Hard <br>3-shot | 72.1 | 59.6 | 57.3 | 69.7 | 51.5 | 68.3 |
| Language Understanding | ANLI <br>7-shot | 52.3 | 48.7 | 47.1 | 55.2 | 57.3 | 58.1 |
| | HellaSwag <br>5-shot | 70.5 | 49.8 | 58.5 | 70.4 | 71.1 | 78.8 |
| Reasoning | ARC Challenge <br>10-shot | 85.5 | 78.3 | 78.6 | 87.3 | 82.8 | 87.4 |
| | BoolQ <br>0-shot | 77.1 | 66 | 72.2 | 76.6 | 80.9 | 79.1 |
| | MedQA <br>2-shot | 56.4 | 49.6 | 50 | 62.2 | 60.5 | 63.4 |
| | OpenBookQA <br>10-shot | 78.8 | 78.6 | 79.8 | 85.8 | 82.6 | 86 |
| | PIQA <br>5-shot | 80.1 | 78.1 | 77.7 | 86 | 75.7 | 86.6 |
| | GPQA <br>0-shot | 29.7 | 2.9 | 15 | 6.9 | 32.4 | 29.9 |
| | Social IQA <br>5-shot | 74.7 | 65.5 | 74.6 | 75.9 | 73.9 | 68.3 |
| | TruthfulQA (MC2) <br>10-shot | 64.8 | 52.1 | 53 | 60.1 | 63.2 | 67.7 |
| | WinoGrande <br>5-shot | 71.0 | 55.6 | 54.2 | 62 | 65 | 68.8 |
| Factual Knowledge | TriviaQA <br>5-shot | 57.8 | 72.3 | 75.2 | 82.2 | 67.7 | 85.8 |
| Math | GSM8K CoTT <br>8-shot | 85.3 | 59.8 | 46.4 | 64.7 | 77.4 | 78.1 |
| Code Generation | HumanEval <br>0-shot | 60.4 | 34.1 | 28.0 | 37.8 | 60.4 | 62.2 |
| | MBPP <br>3-shot | 70.0 | 51.5 | 50.8 | 60.2 | 67.7 | 77.8 |
| **Average** | | **66.4** | **56.0** | **56.4** | **64.4** | **65.5** | **70.3** |
**Long Context**: Phi-3 Mini-128K-Instruct supports 128K context length, therefore the model is capable of several long context tasks including long document/meeting summarization, long document QA.
| Benchmark | Phi-3 Mini-128K-Instruct | Mistral-7B | Mixtral 8x7B | LLaMA-3-8B-Instruct |
|---------------|--------------------------|------------|--------------|---------------------|
| GovReport | 25.3 | 4.9 | 20.3 | 10.3 |
| QMSum | 21.9 | 15.5 | 20.6 | 2.9 |
| Qasper | 41.6 | 23.5 | 26.6 | 8.1 |
| SQuALITY | 24.1 | 14.7 | 16.2 | 25 |
| SummScreenFD | 16.8 | 9.3 | 11.3 | 5.1 |
| **Average** | **25.9** | **13.6** | **19.0** | **10.3** |
We take a closer look at different categories across 100 public benchmark datasets at the table below:
| Category | Phi-3-Mini-128K-Instruct | Gemma-7B | Mistral-7B | Mixtral 8x7B | Llama-3-8B-Instruct | GPT-3.5-Turbo |
|----------|--------------------------|----------|------------|--------------|---------------------|---------------|
| Popular aggregated benchmark | 60.6 | 59.4 | 56.5 | 66.2 | 59.9 | 67.0 |
| Reasoning | 69.4 | 60.3 | 62.8 | 68.1 | 69.6 | 71.7 |
| Language understanding | 57.5 | 57.6 | 52.5 | 66.1 | 63.2 | 67.7 |
| Code generation | 61.0 | 45.6 | 42.9 | 52.7 | 56.4 | 70.4 |
| Math | 51.6 | 35.8 | 25.4 | 40.3 | 41.1 | 52.8 |
| Factual knowledge | 35.8 | 46.7 | 49.8 | 58.6 | 43.1 | 63.4 |
| Multilingual | 56.4 | 66.5 | 57.4 | 66.7 | 66.6 | 71.0 |
| Robustness | 61.1 | 38.4 | 40.6 | 51.0 | 64.5 | 69.3 |
Overall, the model with only 3.8B-param achieves a similar level of language understanding and reasoning ability as much larger models. However, it is still fundamentally limited by its size for certain tasks. The model simply does not have the capacity to store too much world knowledge, which can be seen for example with low performance on TriviaQA. However, we believe such weakness can be resolved by augmenting Phi-3-Mini with a search engine.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Cross Platform Support
[ONNX runtime](https://onnxruntime.ai/blogs/accelerating-phi-3) now supports Phi-3 mini models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 mini across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via RTN
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
## Hardware
Note that by default, the Phi-3 Mini-128K-Instruct model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
* NVIDIA V100 or earlier generation GPUs: call AutoModelForCausalLM.from_pretrained() with attn_implementation="eager"
* Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [128K](https://aka.ms/phi3-mini-128k-instruct-onnx)
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

57
assets/models/system/phi-3-mini-128k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,57 @@
We report the results under completion format for Phi-3-Mini-128K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mistral-7b-v0.1, Mixtral-8x7b, Gemma 7B, Llama-3-8B-Instruct, and GPT-3.5.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
| Category | Benchmark | Phi-3-Mini-128K-Ins | Gemma-7B | Mistral-7B | Mixtral-8x7B | Llama-3-8B-Ins | GPT3.5-Turbo-1106 |
| :----------| :-----------| :---------------------| :----------| :------------| :--------------| :----------------| :-------------------|
| Popular aggregated benchmark | AGI Eval <br>5-shot| 39.5 | 42.1 | 35.1 | 45.2 | 42 | 48.4 |
| | MMLU <br>5-shot | 69.7 | 63.6 | 61.7 | 70.5 | 66.5 | 71.4 |
| | BigBench Hard <br>3-shot | 72.1 | 59.6 | 57.3 | 69.7 | 51.5 | 68.3 |
| Language Understanding | ANLI <br>7-shot | 52.3 | 48.7 | 47.1 | 55.2 | 57.3 | 58.1 |
| | HellaSwag <br>5-shot | 70.5 | 49.8 | 58.5 | 70.4 | 71.1 | 78.8 |
| Reasoning | ARC Challenge <br>10-shot | 85.5 | 78.3 | 78.6 | 87.3 | 82.8 | 87.4 |
| | BoolQ <br>0-shot | 77.1 | 66 | 72.2 | 76.6 | 80.9 | 79.1 |
| | MedQA <br>2-shot | 56.4 | 49.6 | 50 | 62.2 | 60.5 | 63.4 |
| | OpenBookQA <br>10-shot | 78.8 | 78.6 | 79.8 | 85.8 | 82.6 | 86 |
| | PIQA <br>5-shot | 80.1 | 78.1 | 77.7 | 86 | 75.7 | 86.6 |
| | GPQA <br>0-shot | 29.7 | 2.9 | 15 | 6.9 | 32.4 | 29.9 |
| | Social IQA <br>5-shot | 74.7 | 65.5 | 74.6 | 75.9 | 73.9 | 68.3 |
| | TruthfulQA (MC2) <br>10-shot | 64.8 | 52.1 | 53 | 60.1 | 63.2 | 67.7 |
| | WinoGrande <br>5-shot | 71.0 | 55.6 | 54.2 | 62 | 65 | 68.8 |
| Factual Knowledge | TriviaQA <br>5-shot | 57.8 | 72.3 | 75.2 | 82.2 | 67.7 | 85.8 |
| Math | GSM8K CoTT <br>8-shot | 85.3 | 59.8 | 46.4 | 64.7 | 77.4 | 78.1 |
| Code Generation | HumanEval <br>0-shot | 60.4 | 34.1 | 28.0 | 37.8 | 60.4 | 62.2 |
| | MBPP <br>3-shot | 70.0 | 51.5 | 50.8 | 60.2 | 67.7 | 77.8 |
| **Average** | | **66.4** | **56.0** | **56.4** | **64.4** | **65.5** | **70.3** |
**Long Context**: Phi-3 Mini-128K-Instruct supports 128K context length, therefore the model is capable of several long context tasks including long document/meeting summarization, long document QA.
| Benchmark | Phi-3 Mini-128K-Instruct | Mistral-7B | Mixtral 8x7B | LLaMA-3-8B-Instruct |
| :---------------| :--------------------------|:------------|:--------------|:---------------------|
| GovReport | 25.3 | 4.9 | 20.3 | 10.3 |
| QMSum | 21.9 | 15.5 | 20.6 | 2.9 |
| Qasper | 41.6 | 23.5 | 26.6 | 8.1 |
| SQuALITY | 24.1 | 14.7 | 16.2 | 25 |
| SummScreenFD | 16.8 | 9.3 | 11.3 | 5.1 |
| **Average** | **25.9** | **13.6** | **19.0** | **10.3** |
We take a closer look at different categories across 100 public benchmark datasets at the table below:
| Category | Phi-3-Mini-128K-Instruct | Gemma-7B | Mistral-7B | Mixtral 8x7B | Llama-3-8B-Instruct | GPT-3.5-Turbo |
|:----------|:--------------------------|:----------|:------------|:--------------|:---------------------|:---------------|
| Popular aggregated benchmark | 60.6 | 59.4 | 56.5 | 66.2 | 59.9 | 67.0 |
| Reasoning | 69.4 | 60.3 | 62.8 | 68.1 | 69.6 | 71.7 |
| Language understanding | 57.5 | 57.6 | 52.5 | 66.1 | 63.2 | 67.7 |
| Code generation | 61.0 | 45.6 | 42.9 | 52.7 | 56.4 | 70.4 |
| Math | 51.6 | 35.8 | 25.4 | 40.3 | 41.1 | 52.8 |
| Factual knowledge | 35.8 | 46.7 | 49.8 | 58.6 | 43.1 | 63.4 |
| Multilingual | 56.4 | 66.5 | 57.4 | 66.7 | 66.6 | 71.0 |
| Robustness | 61.1 | 38.4 | 40.6 | 51.0 | 64.5 | 69.3 |
Overall, the model with only 3.8B-param achieves a similar level of language understanding and reasoning ability as much larger models. However, it is still fundamentally limited by its size for certain tasks. The model simply does not have the capacity to store too much world knowledge, which can be seen for example with low performance on TriviaQA. However, we believe such weakness can be resolved by augmenting Phi-3-Mini with a search engine.

49
assets/models/system/phi-3-mini-128k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-mini-128k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 24|1|220|64
finetune-recommended-sku: Standard_NC24ads_A100_v4, Standard_NC48ads_A100_v4, Standard_NC96ads_A100_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 mini instruct (128k)"
summary: "Same Phi-3-mini model, but with a larger context size for RAG or few shot prompting."

216
assets/models/system/phi-3-mini-4k-instruct/description.md
Просмотреть файл

@ -1,224 +1,26 @@
## Model Summary
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 family with the Mini version in two variants 4K and 128K which is the context length (in tokens) that it can support.
The model belongs to the Phi-3 family with the Mini version in two variants [4K](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) and [128K](https://huggingface.co/microsoft/Phi-3-mini-128k-instruct) which is the context length (in tokens) that it can support.
The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
The model underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3 Mini-4K-Instruct showcased a robust and state-of-the-art performance among models with less than 13 billion parameters.
Resources and Technical Documentation:
## Resources
🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3) <br>
📰 [Phi-3 Microsoft Blog](https://aka.ms/Phi-3Build2024) <br>
📖 [Phi-3 Technical Report](https://aka.ms/phi3-tech-report) <br>
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai) <br>
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook) <br>
🖥️ [Try It](https://aka.ms/try-phi3)
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3 Mini-4K-Instruct has 3.8B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications which require
1) memory/compute constrained environments;
2) latency bound scenarios;
3) strong reasoning (especially math and logic).
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 fairness 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.
***Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.***
## Release Notes
This is an update over the original instruction-tuned Phi-3-mini release based on valuable customer feedback.
The model used additional post-training data leading to substantial gains on instruction following and structure output.
We also improve multi-turn conversation quality, explicitly support <|system|> tag, and significantly improve reasoning capability.
We believe most use cases will benefit from this release, but we encourage users to test in their particular AI applications.
We appreciate the enthusiastic adoption of the Phi-3 model family, and continue to welcome all feedback from the community.
The table below highlights improvements on instruction following, structure output, and reasoning of the new release on our internal benchmark datasets.
| Benchmarks | Original | June 2024 Update |
|------------|----------|------------------|
| Instruction Extra Hard | 5.7 | 6.0 |
| Instruction Hard | 4.9 | 5.1 |
| Instructions Challenge | 24.6 | 42.3 |
| JSON Structure Output | 11.5 | 52.3 |
| XML Structure Output | 14.4 | 49.8 |
| GPQA | 23.7 | 30.6 |
| MMLU | 68.8 | 70.9 |
| **Average** | **21.9** | **36.7** |
### Tokenizer
Phi-3 Mini-4K-Instruct supports a vocabulary size of up to `32064` tokens. The tokenizer files already provide placeholder tokens that can be used for downstream fine-tuning, but they can also be extended up to the model's vocabulary size.
## Training
### Model
* Architecture: Phi-3 Mini-4K-Instruct has 3.8B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 4K tokens
* GPUs: 512 H100-80G
* Training time: 10 days
* Training data: 4.9T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between May and June 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates: June, 2024.
### Datasets
## Training Datasets
Our training data includes a wide variety of sources, totaling 4.9 trillion tokens, and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results under completion format for Phi-3-Mini-4K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mistral-7b-v0.1, Mixtral-8x7b, Gemma 7B, Llama-3-8B-Instruct, and GPT3.5-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
| Category | Benchmark | Phi-3-Mini-4K-Ins | Gemma-7B | Mistral-7b | Mixtral-8x7b | Llama-3-8B-Ins | GPT3.5-Turbo-1106 |
|----------|-----------|-------------------|----------|------------|--------------|----------------|-------------------|
| Popular aggregated benchmark | AGI Eval <br>5-shot| 39.0 | 42.1 | 35.1 | 45.2 | 42 | 48.4 |
| | MMLU <br>5-shot | 70.9 | 63.6 | 61.7 | 70.5 | 66.5 | 71.4 |
| | BigBench Hard CoT<br>3-shot| 73.5 | 59.6 | 57.3 | 69.7 | 51.5 | 68.3 |
| Language Understanding | ANLI <br>7-shot | 53.6 | 48.7 | 47.1 | 55.2 | 57.3 | 58.1 |
| | HellaSwag <br>5-shot| 75.3 | 49.8 | 58.5 | 70.4 | 71.1 | 78.8 |
| Reasoning | ARC Challenge <br>10-shot | 86.3 | 78.3 | 78.6 | 87.3 | 82.8 | 87.4 |
| | BoolQ <br>0-shot | 78.1 | 66 | 72.2 | 76.6 | 80.9 | 79.1 |
| | MedQA <br>2-shot| 56.5 | 49.6 | 50 | 62.2 | 60.5 | 63.4 |
| | OpenBookQA <br>10-shot| 82.2 | 78.6 | 79.8 | 85.8 | 82.6 | 86 |
| | PIQA <br>5-shot| 83.5 | 78.1 | 77.7 | 86 | 75.7 | 86.6 |
| | GPQA <br>0-shot| 30.6 | 2.9 | 15 | 6.9 | 32.4 | 30.8 |
| | Social IQA <br>5-shot| 77.6 | 65.5 | 74.6 | 75.9 | 73.9 | 68.3 |
| | TruthfulQA (MC2) <br>10-shot| 64.7 | 52.1 | 53 | 60.1 | 63.2 | 67.7 |
| | WinoGrande <br>5-shot| 71.6 | 55.6 | 54.2 | 62 | 65 | 68.8 |
| Factual Knowledge | TriviaQA <br>5-shot| 61.4 | 72.3 | 75.2 | 82.2 | 67.7 | 85.8 |
| Math | GSM8K CoT <br>8-shot| 85.7 | 59.8 | 46.4 | 64.7 | 77.4 | 78.1 |
| Code Generation | HumanEval <br>0-shot| 57.3 | 34.1 | 28.0 | 37.8 | 60.4 | 62.2 |
| | MBPP <br>3-shot| 69.8 | 51.5 | 50.8 | 60.2 | 67.7 | 77.8 |
| **Average** | | **67.6** | **56.0** | **56.4** | **64.4** | **65.5** | **70.4** |
We take a closer look at different categories across 100 public benchmark datasets at the table below:
| Category | Phi-3-Mini-4K-Instruct | Gemma-7B | Mistral-7B | Mixtral 8x7B | Llama-3-8B-Instruct | GPT-3.5-Turbo |
|----------|------------------------|----------|------------|--------------|---------------------|---------------|
| Popular aggregated benchmark | 61.1 | 59.4 | 56.5 | 66.2 | 59.9 | 67.0 |
| Reasoning | 70.8 | 60.3 | 62.8 | 68.1 | 69.6 | 71.8 |
| Language understanding | 60.5 | 57.6 | 52.5 | 66.1 | 63.2 | 67.7 |
| Code generation | 60.7 | 45.6 | 42.9 | 52.7 | 56.4 | 70.4 |
| Math | 50.6 | 35.8 | 25.4 | 40.3 | 41.1 | 52.8 |
| Factual knowledge | 38.4 | 46.7 | 49.8 | 58.6 | 43.1 | 63.4 |
| Multilingual | 56.7 | 66.5 | 57.4 | 66.7 | 66.6 | 71.0 |
| Robustness | 61.1 | 38.4 | 40.6 | 51.0 | 64.5 | 69.3 |
Overall, the model with only 3.8B-param achieves a similar level of language understanding and reasoning ability as much larger models. However, it is still fundamentally limited by its size for certain tasks. The model simply does not have the capacity to store too much world knowledge, which can be seen for example with low performance on TriviaQA. However, we believe such weakness can be resolved by augmenting Phi-3-Mini with a search engine.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
[ONNX runtime](https://onnxruntime.ai/blogs/accelerating-phi-3) now supports Phi-3 mini models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 mini across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via R
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
## Hardware
Note that by default, the Phi-3 Mini-4K-Instruct model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
* NVIDIA V100 or earlier generation GPUs: call AutoModelForCausalLM.from_pretrained() with attn_implementation="eager"
* CPU: use the **GGUF** quantized models [4K](https://aka.ms/Phi3-mini-4k-instruct-gguf)
+ Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [4K](https://aka.ms/Phi3-mini-4k-instruct-onnx)
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

48
assets/models/system/phi-3-mini-4k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,48 @@
We report the results under completion format for Phi-3-Mini-4K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mistral-7b-v0.1, Mixtral-8x7b, Gemma 7B, Llama-3-8B-Instruct, and GPT3.5-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
| Category | Benchmark | Phi-3-Mini-4K-Ins | Gemma-7B | Mistral-7b | Mixtral-8x7b | Llama-3-8B-Ins | GPT3.5-Turbo-1106 |
|:----------|:-----------|:-------------------|:----------|:------------|:--------------|:----------------|:-------------------|
| Popular aggregated benchmark | AGI Eval <br>5-shot| 39.0 | 42.1 | 35.1 | 45.2 | 42 | 48.4 |
| | MMLU <br>5-shot | 70.9 | 63.6 | 61.7 | 70.5 | 66.5 | 71.4 |
| | BigBench Hard CoT<br>3-shot| 73.5 | 59.6 | 57.3 | 69.7 | 51.5 | 68.3 |
| Language Understanding | ANLI <br>7-shot | 53.6 | 48.7 | 47.1 | 55.2 | 57.3 | 58.1 |
| | HellaSwag <br>5-shot| 75.3 | 49.8 | 58.5 | 70.4 | 71.1 | 78.8 |
| Reasoning | ARC Challenge <br>10-shot | 86.3 | 78.3 | 78.6 | 87.3 | 82.8 | 87.4 |
| | BoolQ <br>0-shot | 78.1 | 66 | 72.2 | 76.6 | 80.9 | 79.1 |
| | MedQA <br>2-shot| 56.5 | 49.6 | 50 | 62.2 | 60.5 | 63.4 |
| | OpenBookQA <br>10-shot| 82.2 | 78.6 | 79.8 | 85.8 | 82.6 | 86 |
| | PIQA <br>5-shot| 83.5 | 78.1 | 77.7 | 86 | 75.7 | 86.6 |
| | GPQA <br>0-shot| 30.6 | 2.9 | 15 | 6.9 | 32.4 | 30.8 |
| | Social IQA <br>5-shot| 77.6 | 65.5 | 74.6 | 75.9 | 73.9 | 68.3 |
| | TruthfulQA (MC2) <br>10-shot| 64.7 | 52.1 | 53 | 60.1 | 63.2 | 67.7 |
| | WinoGrande <br>5-shot| 71.6 | 55.6 | 54.2 | 62 | 65 | 68.8 |
| Factual Knowledge | TriviaQA <br>5-shot| 61.4 | 72.3 | 75.2 | 82.2 | 67.7 | 85.8 |
| Math | GSM8K CoT <br>8-shot| 85.7 | 59.8 | 46.4 | 64.7 | 77.4 | 78.1 |
| Code Generation | HumanEval <br>0-shot| 57.3 | 34.1 | 28.0 | 37.8 | 60.4 | 62.2 |
| | MBPP <br>3-shot| 69.8 | 51.5 | 50.8 | 60.2 | 67.7 | 77.8 |
| **Average** | | **67.6** | **56.0** | **56.4** | **64.4** | **65.5** | **70.4** |
We take a closer look at different categories across 100 public benchmark datasets at the table below:
| Category | Phi-3-Mini-4K-Instruct | Gemma-7B | Mistral-7B | Mixtral 8x7B | Llama-3-8B-Instruct | GPT-3.5-Turbo |
|:----------|:------------------------|:----------|:------------|:--------------|:---------------------|:---------------|
| Popular aggregated benchmark | 61.1 | 59.4 | 56.5 | 66.2 | 59.9 | 67.0 |
| Reasoning | 70.8 | 60.3 | 62.8 | 68.1 | 69.6 | 71.8 |
| Language understanding | 60.5 | 57.6 | 52.5 | 66.1 | 63.2 | 67.7 |
| Code generation | 60.7 | 45.6 | 42.9 | 52.7 | 56.4 | 70.4 |
| Math | 50.6 | 35.8 | 25.4 | 40.3 | 41.1 | 52.8 |
| Factual knowledge | 38.4 | 46.7 | 49.8 | 58.6 | 43.1 | 63.4 |
| Multilingual | 56.7 | 66.5 | 57.4 | 66.7 | 66.6 | 71.0 |
| Robustness | 61.1 | 38.4 | 40.6 | 51.0 | 64.5 | 69.3 |
Overall, the model with only 3.8B-param achieves a similar level of language understanding and reasoning ability as much larger models. However, it is still fundamentally limited by its size for certain tasks. The model simply does not have the capacity to store too much world knowledge, which can be seen for example with low performance on TriviaQA. However, we believe such weakness can be resolved by augmenting Phi-3-Mini with a search engine.

49
assets/models/system/phi-3-mini-4k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-mini-4k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 24|1|220|64
finetune-recommended-sku: Standard_NC24ads_A100_v4, Standard_NC48ads_A100_v4, Standard_NC96ads_A100_v4, Standard_ND96asr_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 mini instruct (4k)"
summary: "Tiniest member of the Phi-3 family. Optimized for both quality and low latency."

181
assets/models/system/phi-3-small-128k-instruct/description.md
Просмотреть файл

@ -1,191 +1,22 @@
## Model Summary
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 context length (in tokens).
The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
The model underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3-Small-128K-Instruct showcased a robust and state-of-the-art performance among models of the same-size and next-size-up.
Resources and Technical Documentation:
## Resources
+ [Phi-3 Microsoft Blog](https://aka.ms/phi3blog-april)
+ [Phi-3 Technical Report](https://aka.ms/phi3-tech-report)
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3 Small-128K-Instruct has 7B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications which require :
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
Our model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.
**Use case considerations**
The model is 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.
Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.
## Training
### Model
* Architecture: Phi-3 Small-128K-Instruct has 7B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 128K tokens
* GPUs: 1024 H100-80G
* Training time: 18 days
* Training data: 4.8T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between February and April 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates The model weight is released on May 21, 2024.
### Datasets
## Training Datasets
Our training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results for Phi-3-Small-128K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x7b, Gemini-Pro, Gemma 7B, Llama-3-8B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Small-128K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-------------------------------|----------|-------------|-------------------------|---------------------------|------------|--------------------------------|
|AGI Eval<br>5-shot|43.9|42.1|45.2|42.0|48.4|49.0|59.6|
|MMLU<br>5-shot|75.5|63.6|70.5|66.5|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|77.6|59.6|69.7|51.5|68.3|75.6|87.7|
|ANLI<br>7-shot|55.8|48.7|55.2|57.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|79.6|49.8|70.4|71.1|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|90.8|78.3|87.3|82.8|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.3|91.4|95.6|93.4|96.3|96.1|98.8|
|BoolQ<br>2-shot|83.7|66.0|76.6|80.9|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|80.8|76.2|78.1|79.0|79.6|81.8|86.7|
|MedQA<br>2-shot|46.3|49.6|62.2|60.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.8|78.6|85.8|82.6|86.0|86.4|93.4|
|PIQA<br>5-shot|88.1|78.1|86.0|75.7|86.6|86.2|90.1|
|Social IQA<br>5-shot|78.7|65.5|75.9|73.9|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|69.6|52.1|60.1|63.2|67.7|72.6|85.2|
|WinoGrande<br>5-shot|80.1|55.6|62.0|65.0|68.8|72.2|86.7|
|TriviaQA<br>5-shot|66.0|72.3|82.2|67.7|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|87.3|59.8|64.7|77.4|78.1|80.4|94.2|
|HumanEval<br>0-shot|59.1|34.1|37.8|60.4|62.2|64.4|79.9|
|MBPP<br>3-shot|70.3|51.5|60.2|67.7|77.8|73.2|86.7|
|Average|74.6|61.8|69.8|69.4|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Small-128K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|--------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|70.6|59.4|66.2|59.9|67.0|67.5|80.5|
|Reasoning|80.3|69.1|77.0|75.7|78.3|80.4|89.3|
|Language understanding|67.4|58.4|64.9|65.4|70.4|75.3|81.6|
|Code generation|60.0|45.6|52.7|56.4|70.4|66.7|76.1|
|Math|48.1|35.8|40.3|41.1|52.8|50.9|67.1|
|Factual knowledge|41.7|46.7|58.6|43.1|63.4|54.6|45.9|
|Multilingual|62.6|63.2|63.4|65.0|69.1|76.5|82.0|
|Robustness|68.7|38.4|51.0|64.5|69.3|69.7|84.6|
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [DeepSpeed](https://github.com/microsoft/DeepSpeed)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
* [Tiktoken](https://github.com/openai/tiktoken)
* [Triton](https://github.com/openai/triton)
## Hardware
Note that by default, the Phi-3-Small model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
+ Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [128K](https://huggingface.co/microsoft/Phi-3-small-128k-instruct-onnx)
## Cross Platform Support
ONNX runtime ecosystem now supports Phi3 small models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 Small across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via RTN
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

45
assets/models/system/phi-3-small-128k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,45 @@
We report the results for Phi-3-Small-128K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x7b, Gemini-Pro, Gemma 7B, Llama-3-8B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Small-128K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-------------------------------|----------|-------------|-------------------------|---------------------------|------------|--------------------------------|
|AGI Eval<br>5-shot|43.9|42.1|45.2|42.0|48.4|49.0|59.6|
|MMLU<br>5-shot|75.5|63.6|70.5|66.5|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|77.6|59.6|69.7|51.5|68.3|75.6|87.7|
|ANLI<br>7-shot|55.8|48.7|55.2|57.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|79.6|49.8|70.4|71.1|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|90.8|78.3|87.3|82.8|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.3|91.4|95.6|93.4|96.3|96.1|98.8|
|BoolQ<br>2-shot|83.7|66.0|76.6|80.9|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|80.8|76.2|78.1|79.0|79.6|81.8|86.7|
|MedQA<br>2-shot|46.3|49.6|62.2|60.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|87.8|78.6|85.8|82.6|86.0|86.4|93.4|
|PIQA<br>5-shot|88.1|78.1|86.0|75.7|86.6|86.2|90.1|
|Social IQA<br>5-shot|78.7|65.5|75.9|73.9|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|69.6|52.1|60.1|63.2|67.7|72.6|85.2|
|WinoGrande<br>5-shot|80.1|55.6|62.0|65.0|68.8|72.2|86.7|
|TriviaQA<br>5-shot|66.0|72.3|82.2|67.7|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|87.3|59.8|64.7|77.4|78.1|80.4|94.2|
|HumanEval<br>0-shot|59.1|34.1|37.8|60.4|62.2|64.4|79.9|
|MBPP<br>3-shot|70.3|51.5|60.2|67.7|77.8|73.2|86.7|
|Average|74.6|61.8|69.8|69.4|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Small-128K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|--------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|70.6|59.4|66.2|59.9|67.0|67.5|80.5|
|Reasoning|80.3|69.1|77.0|75.7|78.3|80.4|89.3|
|Language understanding|67.4|58.4|64.9|65.4|70.4|75.3|81.6|
|Code generation|60.0|45.6|52.7|56.4|70.4|66.7|76.1|
|Math|48.1|35.8|40.3|41.1|52.8|50.9|67.1|
|Factual knowledge|41.7|46.7|58.6|43.1|63.4|54.6|45.9|
|Multilingual|62.6|63.2|63.4|65.0|69.1|76.5|82.0|
|Robustness|68.7|38.4|51.0|64.5|69.3|69.7|84.6|

49
assets/models/system/phi-3-small-128k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-small-128k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 96|4|880|256
finetune-recommended-sku: Standard_NC96ads_A100_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 small instruct (128k)"
summary: "Same Phi-3-small model, but with a larger context size for RAG or few shot prompting."

187
assets/models/system/phi-3-small-8k-instruct/description.md
Просмотреть файл

@ -1,190 +1,25 @@
## Model Summary
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 length (in tokens).
The model has underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
The model underwent a post-training process that incorporates both supervised fine-tuning and direct preference optimization for the instruction following and safety measures.
When assessed against benchmarks testing common sense, language understanding, math, code, long context and logical reasoning, Phi-3-Small-8K-Instruct showcased a robust and state-of-the-art performance among models of the same-size and next-size-up.
Resources and Technical Documentation:
## Resources
+ [Phi-3 Microsoft Blog](https://aka.ms/phi3blog-april)
+ [Phi-3 Technical Report](https://aka.ms/phi3-tech-report)
🏡 [Phi-3 Portal](https://azure.microsoft.com/en-us/products/phi-3) <br>
📰 [Phi-3 Microsoft Blog](https://aka.ms/Phi-3Build2024) <br>
📖 [Phi-3 Technical Report](https://aka.ms/phi3-tech-report) <br>
🛠️ [Phi-3 on Azure AI Studio](https://aka.ms/phi3-azure-ai) <br>
👩‍🍳 [Phi-3 Cookbook](https://github.com/microsoft/Phi-3CookBook) <br>
## Intended Uses
## Model Architecture
**Primary use cases**
Phi-3 Small-8K-Instruct has 7B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidelines.
The model is intended for broad commercial and research use in English. The model provides uses for general purpose AI systems and applications which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
Our model is designed to accelerate research on language and multimodal models, for use as a building block for generative AI powered features.
**Use case considerations**
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.
Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.
## Training
### Model
* Architecture: Phi-3 Small-8K-Instruct has 7B parameters and is a dense decoder-only Transformer model. The model is fine-tuned with Supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) to ensure alignment with human preferences and safety guidlines.
* Inputs: Text. It is best suited for prompts using chat format.
* Context length: 8K tokens
* GPUs: 1024 H100-80G
* Training time: 18 days
* Training data: 4.8T tokens
* Outputs: Generated text in response to the input
* Dates: Our models were trained between February and April 2024
* Status: This is a static model trained on an offline dataset with cutoff date October 2023. Future versions of the tuned models may be released as we improve models.
* Release dates The model weight is released on May 21, 2024.
### Datasets
## Training Datasets
Our training data includes a wide variety of sources, totaling 4.8 trillion tokens (including 10% multilingual), and is a combination of
1) Publicly available documents filtered rigorously for quality, selected high-quality educational data, and code;
2) 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.);
3) High quality chat format supervised data covering various topics to reflect human preferences on different aspects such as instruct-following, truthfulness, honesty and helpfulness.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).
## Benchmarks
We report the results for Phi-3-Small-8K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x7b, Gemini-Pro, Gemma 7B, Llama-3-8B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Small-8K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-----------------|----------|------------------------|
|AGI Eval<br>5-shot|45.1|42.1|45.2|42.0|48.4|49.0|59.6|
|MMLU<br>5-shot|75.7|63.6|70.5|66.5|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|79.1|59.6|69.7|51.5|68.3|75.6|87.7|
|ANLI<br>7-shot|58.1|48.7|55.2|57.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|77.0|49.8|70.4|71.1|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|90.7|78.3|87.3|82.8|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.0|91.4|95.6|93.4|96.3|96.1|98.8|
|BoolQ<br>2-shot|84.8|66.0|76.6|80.9|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|80.0|76.2|78.1|79.0|79.6|81.8|86.7|
|MedQA<br>2-shot|65.4|49.6|62.2|60.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|88.0|78.6|85.8|82.6|86.0|86.4|93.4|
|PIQA<br>5-shot|86.9|78.1|86.0|75.7|86.6|86.2|90.1|
|Social IQA<br>5-shot|79.2|65.5|75.9|73.9|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|70.2|52.1|60.1|63.2|67.7|72.6|85.2|
|WinoGrande<br>5-shot|81.5|55.6|62.0|65.0|68.8|72.2|86.7|
|TriviaQA<br>5-shot|58.1|72.3|82.2|67.7|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|89.6|59.8|64.7|77.4|78.1|80.4|94.2|
|HumanEval<br>0-shot|61.0|34.1|37.8|60.4|62.2|64.4|79.9|
|MBPP<br>3-shot|71.7|51.5|60.2|67.7|77.8|73.2|86.7|
|Average|75.7|61.8|69.8|69.4|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Small-8K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|71.1|59.4|66.2|59.9|67.0|67.5|80.5|
|Reasoning|82.4|69.1|77.0|75.7|78.3|80.4|89.3|
|Language understanding|70.6|58.4|64.9|65.4|70.4|75.3|81.6|
|Code generation|60.7|45.6|52.7|56.4|70.4|66.7|76.1|
|Math|51.6|35.8|40.3|41.1|52.8|50.9|67.1|
|Factual knowledge|38.6|46.7|58.6|43.1|63.4|54.6|45.9|
|Multilingual|62.5|63.2|63.4|65.0|69.1|76.5|82.0|
|Robustness|72.9|38.4|51.0|64.5|69.3|69.7|84.6|
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Finetuning samples
Task|Dataset|Python sample (Notebook)
|--|--|--|
Chat completion|<a href="https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k" target="_blank">Ultrachat 200K</a>|<a href="https://aka.ms/phi3ftnotebook" target="_blank">chat-completion.ipynb</a>
## Sample inputs and outputs (for real-time inference)
### **Sample input**
```json
{
"input_data": {
"input_string": [
{
"role": "user",
"content": "I am going to Paris, give me a list of 10 places to visit"
}
],
"parameters": {
"temperature": 0.7,
"top_p": 0.9,
"do_sample": true,
"max_new_tokens": 1000
}
}
}
```
### **Sample output**
```json
{
"output": " 1. Eiffel Tower: Visit the iconic symbol of Paris, offering breathtaking views of the city.\n\n2. Louvre Museum: Explore one of the world's largest and most visited museums, home to thousands of works of art, including the Mona Lisa.\n\n3. Notre-Dame Cathedral: Marvel at the stunning Gothic architecture of this famous cathedral, although note that it is currently under renovation due to the 2019 fire.\n\n4. Montmartre: Discover this historic and artistic neighborhood, famous for its bohemian past and the stunning Sacré-Cœur Basilica.\n\n5. Seine River Cruise: Take a relaxing cruise on the Seine River, seeing some of the city's most famous landmarks like the Louvre, Notre-Dame, and the Eiffel Tower from a unique perspective.\n\n6. Champs-Élysées: Visit this famous avenue lined with shops, cafes, and theaters. Don't forget to check out the Arc de Triomphe at its end.\n\n7. Palace of Versailles: Take a day trip from Paris to explore the opulent palace and gardens of Versailles, a UNESCO World Heritage site.\n\n8. Sacré-Cœur Basilica: Located at the highest point in the city, this basilica offers panoramic views of Paris.\n\n9. Latin Quarter: Stroll through this historic and vibrant neighborhood, famous for its student life, lively atmosphere, and cafes.\n\n10. Musée d'Orsay: Visit this museum, housing an impressive collection of Impressionist and Post-Impressionist art, including works by Monet, Degas, Renoir, and Van Gogh."
}
```
## Software
* [PyTorch](https://github.com/pytorch/pytorch)
* [DeepSpeed](https://github.com/microsoft/DeepSpeed)
* [Transformers](https://github.com/huggingface/transformers)
* [Flash-Attention](https://github.com/HazyResearch/flash-attention)
* [Tiktoken](https://github.com/openai/tiktoken)
* [Triton](https://github.com/openai/triton)
## Hardware
Note that by default, the Phi-3-Small model uses flash attention, which requires certain types of GPU hardware to run. We have tested on the following GPU types:
* NVIDIA A100
* NVIDIA A6000
* NVIDIA H100
If you want to run the model on:
+ Optimized inference on GPU, CPU, and Mobile: use the **ONNX** models [8K](https://huggingface.co/microsoft/Phi-3-medium-4k-instruct-onnx)
## Cross Platform Support
ONNX runtime ecosystem now supports Phi3 small models across platforms and hardware.
Optimized phi-3 models are also published here in ONNX format, to run with ONNX Runtime on CPU and GPU across devices, including server platforms, Windows, Linux and Mac desktops, and mobile CPUs, with the precision best suited to each of these targets. DirectML GPU acceleration is supported for Windows desktops GPUs (AMD, Intel, and NVIDIA).
Along with DML, ONNX Runtime provides cross platform support for Phi3 Small across a range of devices CPU, GPU, and mobile.
Here are some of the optimized configurations we have added:
1. ONNX models for int4 DML: Quantized to int4 via AWQ
2. ONNX model for fp16 CUDA
3. ONNX model for int4 CUDA: Quantized to int4 via RTN
4. ONNX model for int4 CPU and Mobile: Quantized to int4 via RTN
## License
The model is licensed under the MIT license.
## 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.
We are focusing on the quality of data that could potentially improve the reasoning ability for the model, and we filter the publicly available documents to contain the correct level of knowledge. As an example, the result of a game in premier league in a particular day might be good training data for frontier models, but we need to remove such information to leave more model capacity for reasoning for the small size models. More details about data can be found in the [Phi-3 Technical Report](https://aka.ms/phi3-tech-report).

45
assets/models/system/phi-3-small-8k-instruct/evaluation.md Normal file
Просмотреть файл

@ -0,0 +1,45 @@
We report the results for Phi-3-Small-8K-Instruct on standard open-source benchmarks measuring the model's reasoning ability (both common sense reasoning and logical reasoning). We compare to Mixtral-8x7b, Gemini-Pro, Gemma 7B, Llama-3-8B-Instruct, GPT-3.5-Turbo-1106, and GPT-4-Turbo-1106.
All the reported numbers are produced with the exact same pipeline to ensure that the numbers are comparable. These numbers might differ from other published numbers due to slightly different choices in the evaluation.
As is now standard, we use few-shot prompts to evaluate the models, at temperature 0.
The prompts and number of shots are part of a Microsoft internal tool to evaluate language models, and in particular we did no optimization to the pipeline for Phi-3.
More specifically, we do not change prompts, pick different few-shot examples, change prompt format, or do any other form of optimization for the model.
The number of k–shot examples is listed per-benchmark.
|Benchmark|Phi-3-Small-8K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|---------|-----------------------|--------|-------------|-------------------|-----------------|----------|------------------------|
|AGI Eval<br>5-shot|45.1|42.1|45.2|42.0|48.4|49.0|59.6|
|MMLU<br>5-shot|75.7|63.6|70.5|66.5|71.4|66.7|84.0|
|BigBench Hard<br>3-shot|79.1|59.6|69.7|51.5|68.3|75.6|87.7|
|ANLI<br>7-shot|58.1|48.7|55.2|57.3|58.1|64.2|71.7|
|HellaSwag<br>5-shot|77.0|49.8|70.4|71.1|78.8|76.2|88.3|
|ARC Challenge<br>10-shot|90.7|78.3|87.3|82.8|87.4|88.3|95.6|
|ARC Easy<br>10-shot|97.0|91.4|95.6|93.4|96.3|96.1|98.8|
|BoolQ<br>2-shot|84.8|66.0|76.6|80.9|79.1|86.4|91.3|
|CommonsenseQA<br>10-shot|80.0|76.2|78.1|79.0|79.6|81.8|86.7|
|MedQA<br>2-shot|65.4|49.6|62.2|60.5|63.4|58.2|83.7|
|OpenBookQA<br>10-shot|88.0|78.6|85.8|82.6|86.0|86.4|93.4|
|PIQA<br>5-shot|86.9|78.1|86.0|75.7|86.6|86.2|90.1|
|Social IQA<br>5-shot|79.2|65.5|75.9|73.9|68.3|75.4|81.7|
|TruthfulQA (MC2)<br>10-shot|70.2|52.1|60.1|63.2|67.7|72.6|85.2|
|WinoGrande<br>5-shot|81.5|55.6|62.0|65.0|68.8|72.2|86.7|
|TriviaQA<br>5-shot|58.1|72.3|82.2|67.7|85.8|80.2|73.3|
|GSM8K Chain of Thought<br>8-shot|89.6|59.8|64.7|77.4|78.1|80.4|94.2|
|HumanEval<br>0-shot|61.0|34.1|37.8|60.4|62.2|64.4|79.9|
|MBPP<br>3-shot|71.7|51.5|60.2|67.7|77.8|73.2|86.7|
|Average|75.7|61.8|69.8|69.4|74.3|75.4|85.2|
We take a closer look at different categories across 80 public benchmark datasets at the table below:
|Benchmark|Phi-3-Small-8K-Instruct<br>7b|Gemma<br>7B|Mixtral<br>8x7B|Llama-3-Instruct<br>8b|GPT-3.5-Turbo<br>version 1106|Gemini<br>Pro|GPT-4-Turbo<br>version 1106 (Chat)|
|--------|------------------------|--------|-------------|-------------------|-------------------|----------|------------------------|
|Popular aggregated benchmark|71.1|59.4|66.2|59.9|67.0|67.5|80.5|
|Reasoning|82.4|69.1|77.0|75.7|78.3|80.4|89.3|
|Language understanding|70.6|58.4|64.9|65.4|70.4|75.3|81.6|
|Code generation|60.7|45.6|52.7|56.4|70.4|66.7|76.1|
|Math|51.6|35.8|40.3|41.1|52.8|50.9|67.1|
|Factual knowledge|38.6|46.7|58.6|43.1|63.4|54.6|45.9|
|Multilingual|62.5|63.2|63.4|65.0|69.1|76.5|82.0|
|Robustness|72.9|38.4|51.0|64.5|69.3|69.7|84.6|

49
assets/models/system/phi-3-small-8k-instruct/notes.md Normal file
Просмотреть файл

@ -0,0 +1,49 @@
## License
The model is licensed under the MIT license.
## Intended Uses
**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 which require:
1) Memory/compute constrained environments
2) Latency bound scenarios
3) Strong reasoning (especially code, math and logic)
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 fairness 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.
**Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the license the model is released under.**
## 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.
## Responsible AI Considerations
Like other language models, the Phi series 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.
## Content Filtering
Prompts and completions are passed through a default configuration of Azure AI Content Safety classification models to detect and prevent the output of harmful content. Learn more about [Azure AI Content Safety](https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview). Configuration options for content filtering vary when you deploy a model for production in Azure AI; [learn more](https://learn.microsoft.com/en-us/azure/ai-studio/how-to/model-catalog-overview).

2
assets/models/system/phi-3-small-8k-instruct/spec.yaml
Просмотреть файл

@ -10,6 +10,8 @@ properties:
finetune-min-sku-spec: 96|4|880|256
finetune-recommended-sku: Standard_NC96ads_A100_v4, Standard_ND96amsr_A100_v4
tags:
notes: "notes.md"
evaluation: "evaluation.md"
freePlayground: "true"
displayName: "Phi-3 small instruct (8k)"
summary: "A 7B parameters model, proves better quality than Phi-3-mini, with a focus on high-quality, reasoning-dense data."