Updating model files
Browse files
README.md
CHANGED
@@ -8,6 +8,17 @@ tags:
|
|
8 |
- medical
|
9 |
inference: false
|
10 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
# medalpaca-13B-GGML
|
13 |
|
@@ -57,35 +68,46 @@ Further instructions here: [text-generation-webui/docs/llama.cpp-models.md](http
|
|
57 |
Note: at this time text-generation-webui may not support the new May 19th llama.cpp quantisation methods for q4_0, q4_1 and q8_0 files.
|
58 |
|
59 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
# Original model card: MedAlpaca 13b
|
61 |
|
62 |
|
63 |
## Table of Contents
|
64 |
|
65 |
-
[Model Description](#model-description)
|
66 |
-
- [Architecture](#architecture)
|
67 |
-
- [Training Data](#trainig-data)
|
68 |
-
[Model Usage](#model-usage)
|
69 |
-
[Limitations](#limitations)
|
70 |
|
71 |
## Model Description
|
72 |
### Architecture
|
73 |
-
`medalpaca-13b` is a large language model specifically fine-tuned for medical domain tasks.
|
74 |
-
It is based on LLaMA (Large Language Model Meta AI) and contains 13 billion parameters.
|
75 |
The primary goal of this model is to improve question-answering and medical dialogue tasks.
|
76 |
|
77 |
### Training Data
|
78 |
-
The training data for this project was sourced from various resources.
|
79 |
-
Firstly, we used Anki flashcards to automatically generate questions,
|
80 |
-
from the front of the cards and anwers from the back of the card.
|
81 |
-
Secondly, we generated medical question-answer pairs from [Wikidoc](https://www.wikidoc.org/index.php/Main_Page).
|
82 |
-
We extracted paragraphs with relevant headings, and used Chat-GPT 3.5
|
83 |
-
to generate questions from the headings and using the corresponding paragraphs
|
84 |
-
as answers. This dataset is still under development and we believe
|
85 |
-
that approximately 70% of these question answer pairs are factual correct.
|
86 |
-
Thirdly, we used StackExchange to extract question-answer pairs, taking the
|
87 |
-
top-rated question from five categories: Academia, Bioinformatics, Biology,
|
88 |
-
Fitness, and Health. Additionally, we used a dataset from [ChatDoctor](https://arxiv.org/abs/2303.14070)
|
89 |
consisting of 200,000 question-answer pairs, available at https://github.com/Kent0n-Li/ChatDoctor.
|
90 |
|
91 |
| Source | n items |
|
@@ -119,7 +141,7 @@ print(answer)
|
|
119 |
|
120 |
## Limitations
|
121 |
The model may not perform effectively outside the scope of the medical domain.
|
122 |
-
The training data primarily targets the knowledge level of medical students,
|
123 |
which may result in limitations when addressing the needs of board-certified physicians.
|
124 |
-
The model has not been tested in real-world applications, so its efficacy and accuracy are currently unknown.
|
125 |
It should never be used as a substitute for a doctor's opinion and must be treated as a research tool only.
|
|
|
8 |
- medical
|
9 |
inference: false
|
10 |
---
|
11 |
+
<div style="width: 100%;">
|
12 |
+
<img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
|
13 |
+
</div>
|
14 |
+
<div style="display: flex; justify-content: space-between; width: 100%;">
|
15 |
+
<div style="display: flex; flex-direction: column; align-items: flex-start;">
|
16 |
+
<p><a href="https://discord.gg/UBgz4VXf">Chat & support: my new Discord server</a></p>
|
17 |
+
</div>
|
18 |
+
<div style="display: flex; flex-direction: column; align-items: flex-end;">
|
19 |
+
<p><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? Patreon coming soon!</a></p>
|
20 |
+
</div>
|
21 |
+
</div>
|
22 |
|
23 |
# medalpaca-13B-GGML
|
24 |
|
|
|
68 |
Note: at this time text-generation-webui may not support the new May 19th llama.cpp quantisation methods for q4_0, q4_1 and q8_0 files.
|
69 |
|
70 |
|
71 |
+
## Want to support my work?
|
72 |
+
|
73 |
+
I've had a lot of people ask if they can contribute. I love providing models and helping people, but it is starting to rack up pretty big cloud computing bills.
|
74 |
+
|
75 |
+
So if you're able and willing to contribute, it'd be most gratefully received and will help me to keep providing models, and work on various AI projects.
|
76 |
+
|
77 |
+
Donaters will get priority support on any and all AI/LLM/model questions, and I'll gladly quantise any model you'd like to try.
|
78 |
+
|
79 |
+
* Patreon: coming soon! (just awaiting approval)
|
80 |
+
* Ko-Fi: https://ko-fi.com/TheBlokeAI
|
81 |
+
* Discord: https://discord.gg/UBgz4VXf
|
82 |
# Original model card: MedAlpaca 13b
|
83 |
|
84 |
|
85 |
## Table of Contents
|
86 |
|
87 |
+
[Model Description](#model-description)
|
88 |
+
- [Architecture](#architecture)
|
89 |
+
- [Training Data](#trainig-data)
|
90 |
+
[Model Usage](#model-usage)
|
91 |
+
[Limitations](#limitations)
|
92 |
|
93 |
## Model Description
|
94 |
### Architecture
|
95 |
+
`medalpaca-13b` is a large language model specifically fine-tuned for medical domain tasks.
|
96 |
+
It is based on LLaMA (Large Language Model Meta AI) and contains 13 billion parameters.
|
97 |
The primary goal of this model is to improve question-answering and medical dialogue tasks.
|
98 |
|
99 |
### Training Data
|
100 |
+
The training data for this project was sourced from various resources.
|
101 |
+
Firstly, we used Anki flashcards to automatically generate questions,
|
102 |
+
from the front of the cards and anwers from the back of the card.
|
103 |
+
Secondly, we generated medical question-answer pairs from [Wikidoc](https://www.wikidoc.org/index.php/Main_Page).
|
104 |
+
We extracted paragraphs with relevant headings, and used Chat-GPT 3.5
|
105 |
+
to generate questions from the headings and using the corresponding paragraphs
|
106 |
+
as answers. This dataset is still under development and we believe
|
107 |
+
that approximately 70% of these question answer pairs are factual correct.
|
108 |
+
Thirdly, we used StackExchange to extract question-answer pairs, taking the
|
109 |
+
top-rated question from five categories: Academia, Bioinformatics, Biology,
|
110 |
+
Fitness, and Health. Additionally, we used a dataset from [ChatDoctor](https://arxiv.org/abs/2303.14070)
|
111 |
consisting of 200,000 question-answer pairs, available at https://github.com/Kent0n-Li/ChatDoctor.
|
112 |
|
113 |
| Source | n items |
|
|
|
141 |
|
142 |
## Limitations
|
143 |
The model may not perform effectively outside the scope of the medical domain.
|
144 |
+
The training data primarily targets the knowledge level of medical students,
|
145 |
which may result in limitations when addressing the needs of board-certified physicians.
|
146 |
+
The model has not been tested in real-world applications, so its efficacy and accuracy are currently unknown.
|
147 |
It should never be used as a substitute for a doctor's opinion and must be treated as a research tool only.
|