Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: llama2
|
3 |
+
library_name: transformers
|
4 |
+
tags:
|
5 |
+
- llama
|
6 |
+
- merge
|
7 |
+
datasets:
|
8 |
+
- GBaker/MedQA-USMLE-4-options
|
9 |
+
- cognitivecomputations/samantha-data
|
10 |
+
- shibing624/medical
|
11 |
+
---
|
12 |
+
|
13 |
+
# Dr. Samantha
|
14 |
+
|
15 |
+
<p align="center">
|
16 |
+
<img src="https://huggingface.co/sethuiyer/Dr_Samantha-7b/resolve/main/dr_samantha_anime_style_reduced_quality.webp" height="256px" alt="SynthIQ">
|
17 |
+
</p>
|
18 |
+
|
19 |
+
## Overview
|
20 |
+
|
21 |
+
Dr. Samantha is a language model made by merging `Severus27/BeingWell_llama2_7b` and `ParthasarathyShanmugam/llama-2-7b-samantha` using [mergekit](https://github.com/cg123/mergekit).
|
22 |
+
|
23 |
+
Has capabilities of a medical knowledge-focused model (trained on USMLE databases and doctor-patient interactions) with the philosophical, psychological, and relational understanding of the Samantha-7b model.
|
24 |
+
|
25 |
+
As both a medical consultant and personal counselor, Dr.Samantha could effectively support both physical and mental wellbeing - important for whole-person care.
|
26 |
+
|
27 |
+
|
28 |
+
# Yaml Config
|
29 |
+
|
30 |
+
```yaml
|
31 |
+
|
32 |
+
slices:
|
33 |
+
- sources:
|
34 |
+
- model: Severus27/BeingWell_llama2_7b
|
35 |
+
layer_range: [0, 32]
|
36 |
+
- model: ParthasarathyShanmugam/llama-2-7b-samantha
|
37 |
+
layer_range: [0, 32]
|
38 |
+
|
39 |
+
merge_method: slerp
|
40 |
+
base_model: TinyPixel/Llama-2-7B-bf16-sharded
|
41 |
+
|
42 |
+
parameters:
|
43 |
+
t:
|
44 |
+
- filter: self_attn
|
45 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
46 |
+
- filter: mlp
|
47 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
48 |
+
- value: 0.5 # fallback for rest of tensors
|
49 |
+
tokenizer_source: union
|
50 |
+
|
51 |
+
dtype: bfloat16
|
52 |
+
|
53 |
+
```
|