Update README.md
Browse files
README.md
CHANGED
|
@@ -18,4 +18,21 @@ tags:
|
|
| 18 |
|
| 19 |
## reference
|
| 20 |
- base model: microsoft/[phi-4](https://huggingface.co/microsoft/phi-4)
|
| 21 |
-
- gguf-connector ([pypi](https://pypi.org/project/gguf-connector/)|[reop](https://github.com/calcuis/gguf-connector))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
## reference
|
| 20 |
- base model: microsoft/[phi-4](https://huggingface.co/microsoft/phi-4)
|
| 21 |
+
- gguf-connector ([pypi](https://pypi.org/project/gguf-connector/)|[reop](https://github.com/calcuis/gguf-connector))
|
| 22 |
+
|
| 23 |
+
## appendix
|
| 24 |
+
|
| 25 |
+
|-------------------------|-------------------------------------------------------------------------------|
|
| 26 |
+
| **Developers** | Microsoft Research |
|
| 27 |
+
| **Description** | `phi-4` is a state-of-the-art open model built upon a blend of synthetic datasets, data from filtered public domain websites, and acquired academic books and Q&A datasets. The goal of this approach was to ensure that small capable models were trained with data focused on high quality and advanced reasoning.<br><br>`phi-4` underwent a rigorous enhancement and alignment process, incorporating both supervised fine-tuning and direct preference optimization to ensure precise instruction adherence and robust safety measures |
|
| 28 |
+
| **Architecture** | 14B parameters, dense decoder-only Transformer model |
|
| 29 |
+
| **Inputs** | Text, best suited for prompts in the chat format |
|
| 30 |
+
| **Context length** | 16K tokens |
|
| 31 |
+
| **GPUs** | 1920 H100-80G |
|
| 32 |
+
| **Training time** | 21 days |
|
| 33 |
+
| **Training data** | 9.8T tokens |
|
| 34 |
+
| **Outputs** | Generated text in response to input |
|
| 35 |
+
| **Dates** | October 2024 – November 2024 |
|
| 36 |
+
| **Status** | Static model trained on an offline dataset with cutoff dates of June 2024 and earlier for publicly available data |
|
| 37 |
+
| **Release date** | December 12, 2024 |
|
| 38 |
+
| **License** | MIT |
|