Commit
•
526c97b
1
Parent(s):
916f9cd
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- chargoddard/Open-Platypus-Chat
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
tags:
|
| 7 |
+
- llama
|
| 8 |
+
---
|
| 9 |
+
Experimental ReLoRA-trained model using the OpenPlatypus dataset. Ran for one epoch, with three lora restarts.
|
| 10 |
+
|
| 11 |
+
Not recommended for use yet. Mostly tossing this up for testing.
|
| 12 |
+
|
| 13 |
+
Base model was [llama2-22b-blocktriangular](https://huggingface.co/chargoddard/llama2-22b-blocktriangular).
|
| 14 |
+
|
| 15 |
+
Relevant training parameters:
|
| 16 |
+
```
|
| 17 |
+
adapter: qlora
|
| 18 |
+
load_in_4bit: true
|
| 19 |
+
lora_r: 32
|
| 20 |
+
lora_alpha: 16
|
| 21 |
+
lora_dropout: 0.001
|
| 22 |
+
lora_target_linear: true
|
| 23 |
+
relora_steps: 150
|
| 24 |
+
relora_warmup_steps: 10
|
| 25 |
+
gradient_accumulation_steps: 2
|
| 26 |
+
micro_batch_size: 3
|
| 27 |
+
```
|
| 28 |
+
|
| 29 |
+
Uses the same prompt format as [Ypotryll-22b](https://huggingface.co/chargoddard/ypotryll-22b-epoch2-qlora).
|
| 30 |
+
Prefix messages with `" ***System:"`, `" ***Query:"`, or `" ***Response:"`, paying attention to whitespace.
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
|