Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: other
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
base_model:
|
| 6 |
+
- tiiuae/Falcon3-7B-Instruct
|
| 7 |
+
pipeline_tag: text-generation
|
| 8 |
+
tags:
|
| 9 |
+
- phi4
|
| 10 |
+
- phi3
|
| 11 |
+
- phi
|
| 12 |
+
- phi-moe
|
| 13 |
+
- moe
|
| 14 |
+
- llama
|
| 15 |
+
- 4bit
|
| 16 |
+
---
|
| 17 |
+
# Phi4 MoE 2x14B Instruct
|
| 18 |
+
|
| 19 |
+
Mixture of Experts of Phi4 14B-IT & 14B-IT.
|
| 20 |
+
|
| 21 |
+
- 14.2B parameters (4bit quant with bitsandbytes)
|
| 22 |
+
- BF16-U8 (Dynamic Quants by Unsloth using bnb-4bit)
|
| 23 |
+
- Phi4 (Phi3, Llama)
|
| 24 |
+
- Instruct
|
| 25 |
+
|
| 26 |
+
## Model Summary
|
| 27 |
+
|
| 28 |
+
| | |
|
| 29 |
+
|-------------------------|-------------------------------------------------------------------------------|
|
| 30 |
+
| **Developers** | Microsoft Research |
|
| 31 |
+
| **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 |
|
| 32 |
+
| **Architecture** | 14B parameters, dense decoder-only Transformer model |
|
| 33 |
+
| **Inputs** | Text, best suited for prompts in the chat format |
|
| 34 |
+
| **Context length** | 16K tokens |
|
| 35 |
+
| **GPUs** | 1920 H100-80G |
|
| 36 |
+
| **Training time** | 21 days |
|
| 37 |
+
| **Training data** | 9.8T tokens |
|
| 38 |
+
| **Outputs** | Generated text in response to input |
|
| 39 |
+
| **Dates** | October 2024 – November 2024 |
|
| 40 |
+
| **Status** | Static model trained on an offline dataset with cutoff dates of June 2024 and earlier for publicly available data |
|
| 41 |
+
| **Release date** | December 12, 2024 |
|
| 42 |
+
| **License** | MIT |
|