Update README.md
Browse files
README.md
CHANGED
@@ -4,4 +4,69 @@ language:
|
|
4 |
- en
|
5 |
tags:
|
6 |
- medical
|
7 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
- en
|
5 |
tags:
|
6 |
- medical
|
7 |
+
---
|
8 |
+
# Palmyra-med-20b
|
9 |
+
|
10 |
+
## Model description
|
11 |
+
**Palmyra-med-20b** is a 20 billion parameter Large Language Model that has been uptrained on
|
12 |
+
Palmyra-Large with a specialized custom curated medical dataset.
|
13 |
+
The main objective of this model is to enhance performance in tasks related to medical dialogue
|
14 |
+
and question-answering.
|
15 |
+
|
16 |
+
## Usage
|
17 |
+
The model is compatible with the huggingface `AutoModelForCausalLM` and can be easily run on a single 40GB A100.
|
18 |
+
|
19 |
+
```py
|
20 |
+
import torch
|
21 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
22 |
+
|
23 |
+
model_name = "Writer/palmyra-med-20b"
|
24 |
+
|
25 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False)
|
26 |
+
|
27 |
+
model = AutoModelForCausalLM.from_pretrained(
|
28 |
+
model_name,
|
29 |
+
device_map="auto"
|
30 |
+
torch_dtype=torch.float16,
|
31 |
+
)
|
32 |
+
|
33 |
+
prompt = "Can you explain in simple terms how vaccines help our body fight diseases?"
|
34 |
+
|
35 |
+
input_text = (
|
36 |
+
"A chat between a curious user and an artificial intelligence assistant. "
|
37 |
+
"The assistant gives helpful, detailed, and polite answers to the user's questions. "
|
38 |
+
"USER: {prompt} "
|
39 |
+
"ASSISTANT:"
|
40 |
+
)
|
41 |
+
|
42 |
+
model_inputs = tokenizer(input_text.format(prompt=prompt), return_tensors="pt").to(
|
43 |
+
"cuda"
|
44 |
+
)
|
45 |
+
|
46 |
+
gen_conf = {
|
47 |
+
"temperature": 0.7,
|
48 |
+
"repetition_penalty": 1.0,
|
49 |
+
"max_new_tokens": 100,
|
50 |
+
"do_sample": True,
|
51 |
+
}
|
52 |
+
|
53 |
+
out_tokens = model.generate(**model_inputs, **gen_conf)
|
54 |
+
|
55 |
+
response_ids = out_tokens[0][len(model_inputs.input_ids[0]) :]
|
56 |
+
output = tokenizer.decode(response_ids, skip_special_tokens=True)
|
57 |
+
|
58 |
+
print(output)
|
59 |
+
## output ##
|
60 |
+
# Vaccines stimulate the production of antibodies by the body's immune system.
|
61 |
+
# Antibodies are proteins produced by B lymphocytes in response to foreign substances,such as viruses and bacteria.
|
62 |
+
# The antibodies produced by the immune system can bind to and neutralize the pathogens, preventing them from invading and damaging the host cells.
|
63 |
+
# Vaccines work by introducing antigens, which are components of the pathogen, into the body.
|
64 |
+
# The immune system then produces antibodies against the antigens, which can recognize and neutralize the pathogen if it enters the body in the future.
|
65 |
+
# The use of vaccines has led to a significant reduction in the incidence and severity of many diseases, including measles, mumps, rubella, and polio.
|
66 |
+
```
|
67 |
+
|
68 |
+
## Limitation
|
69 |
+
The model may not operate efficiently beyond the confines of the healthcare field.
|
70 |
+
Since it has not been subjected to practical scenarios, its real-time efficacy and precision remain undetermined.
|
71 |
+
Under no circumstances should it replace the advice of a medical professional, and it must be regarded solely as a tool for research purposes.
|
72 |
+
|