Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
datasets:
|
3 |
+
- jondurbin/airoboros-gpt4-1.4.1
|
4 |
+
- ehartford/dolphin
|
5 |
+
---
|
6 |
+
|
7 |
+
|
8 |
+
# Airophin: A NTK-by-Parts RoPE Scaled QLoRA Fine-tune of Llama-2-13b (LoRA weights)
|
9 |
+
|
10 |
+
|
11 |
+
<!-- LoRA Weights can be found here: https://huggingface.co/bhenrym14/airophin-13b-pntk-16k-LoRA -->
|
12 |
+
GPTQ weights can be found here: https://huggingface.co/bhenrym14/airophin-13b-pntk-16k-GPTQ
|
13 |
+
fp16 weights can be found here: https://huggingface.co/bhenrym14/airophin-13b-pntk-16k-fp16
|
14 |
+
|
15 |
+
## Overview
|
16 |
+
|
17 |
+
This is a finetune of Llama-2-13b, intended to extend the useful context window to 16384 tokens. There are two training phases:
|
18 |
+
1. It is first trained on a long-context (7000-8192 tokens) subset of [dolphin](https://huggingface.co/datasets/ehartford/dolphin), an orca-like dataset (GPT4 split only). This amounts to roughly 110mm tokens. Airoboros-like training prompt was used instead of the dolphin system prompt. Training was done with partial NTK scaling applied (scale factor of 4). This took ~20 hours.
|
19 |
+
2. The model was then finetuned on [Jon Durbin's Airoboros GPT4 1.4.1](https://huggingface.co/datasets/jondurbin/airoboros-gpt4-1.4.1), with same scaling approach, for 2 epochs. This took ~15 hours.
|
20 |
+
|
21 |
+
**This is a QLoRA fine-tune (rank 64)**.
|
22 |
+
|
23 |
+
All training was performed with 1x RTX 6000 Ada.
|
24 |
+
|
25 |
+
For full model card, including how to use PNTK, see any of the two merged models linked above.
|