Hugging Face's logo

Reverb
/
MedLLaMA-3

Text Generation
Transformers
Safetensors
English
Arabic
llama
llama-3-8b
sft
medical
text-generation-inference
Inference Endpoints
Model card Files Community
Reverb commited on
Commit
b546a32
•
1 Parent(s): e42c484

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +66 -65
README.md CHANGED
@@ -1,66 +1,67 @@
1
- ---
2
- tags:
3
- - llama-3-8b
4
- - sft
5
- - medical
6
- base_model:
7
- - meta-llama/Meta-Llama-3-8B
8
- license: cc-by-nc-nd-4.0
9
- ---
10
-
11
- # JSL-MedLlama-3-8B-v2.0
12
- [<img src="https://repository-images.githubusercontent.com/104670986/2e728700-ace4-11ea-9cfc-f3e060b25ddf">](http://www.johnsnowlabs.com)
13
-
14
-
15
- This model is developed by [John Snow Labs](https://www.johnsnowlabs.com/).
16
-
17
- This model is available under a [CC-BY-NC-ND](https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en) license and must also conform to this [Acceptable Use Policy](https://huggingface.co/johnsnowlabs). If you need to license this model for commercial use, please contact us at [email protected].
18
-
19
-
20
-
21
- ## 💻 Usage
22
-
23
- ```python
24
- !pip install -qU transformers accelerate
25
-
26
- from transformers import AutoTokenizer
27
- import transformers
28
- import torch
29
-
30
- model = "johnsnowlabs/JSL-MedLlama-3-8B-v2.0"
31
- messages = [{"role": "user", "content": "What is a large language model?"}]
32
-
33
- tokenizer = AutoTokenizer.from_pretrained(model)
34
- prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
35
- pipeline = transformers.pipeline(
36
- "text-generation",
37
- model=model,
38
- torch_dtype=torch.float16,
39
- device_map="auto",
40
- )
41
-
42
- outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
43
- print(outputs[0]["generated_text"])
44
- ```
45
- ## 🏆 Evaluation
46
-
47
- | Tasks |Version|Filter|n-shot| Metric |Value | |Stderr|
48
- |-------------------------------|-------|------|-----:|--------|-----:|---|-----:|
49
- |stem |N/A |none | 0|acc |0.6466|± |0.0056|
50
- | | |none | 0|acc_norm|0.6124|± |0.0066|
51
- | - medmcqa |Yaml |none | 0|acc |0.6118|± |0.0075|
52
- | | |none | 0|acc_norm|0.6118|± |0.0075|
53
- | - medqa_4options |Yaml |none | 0|acc |0.6143|± |0.0136|
54
- | | |none | 0|acc_norm|0.6143|± |0.0136|
55
- | - anatomy (mmlu) | 0|none | 0|acc |0.7185|± |0.0389|
56
- | - clinical_knowledge (mmlu) | 0|none | 0|acc |0.7811|± |0.0254|
57
- | - college_biology (mmlu) | 0|none | 0|acc |0.8264|± |0.0317|
58
- | - college_medicine (mmlu) | 0|none | 0|acc |0.7110|± |0.0346|
59
- | - medical_genetics (mmlu) | 0|none | 0|acc |0.8300|± |0.0378|
60
- | - professional_medicine (mmlu)| 0|none | 0|acc |0.7868|± |0.0249|
61
- | - pubmedqa | 1|none | 0|acc |0.7420|± |0.0196|
62
-
63
- |Groups|Version|Filter|n-shot| Metric |Value | |Stderr|
64
- |------|-------|------|-----:|--------|-----:|---|-----:|
65
- |stem |N/A |none | 0|acc |0.6466|± |0.0056|
 
66
  | | |none | 0|acc_norm|0.6124|± |0.0066|
 
1
+ ---
2
+ tags:
3
+ - llama-3-8b
4
+ - sft
5
+ - medical
6
+ base_model:
7
+ - meta-llama/Meta-Llama-3-8B
8
+ license: cc-by-nc-nd-4.0
9
+ datasets:
10
+ - lighteval/med_mcqa
11
+ - qiaojin/PubMedQA
12
+ - bigbio/med_qa
13
+ ---
14
+
15
+ # MedLLaMA-3
16
+ [<img src="https://repository-images.githubusercontent.com/104670986/2e728700-ace4-11ea-9cfc-f3e060b25ddf">](http://www.johnsnowlabs.com)
17
+
18
+
19
+ This model is developed by Basel Anaya.
20
+
21
+
22
+ ## 💻 Usage
23
+
24
+ ```python
25
+ !pip install -qU transformers accelerate
26
+
27
+ from transformers import AutoTokenizer
28
+ import transformers
29
+ import torch
30
+
31
+ model = "Reverb/MedLLaMA-3"
32
+ messages = [{"role": "user", "content": "What is a large language model?"}]
33
+
34
+ tokenizer = AutoTokenizer.from_pretrained(model)
35
+ prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
36
+ pipeline = transformers.pipeline(
37
+ "text-generation",
38
+ model=model,
39
+ torch_dtype=torch.float16,
40
+ device_map="auto",
41
+ )
42
+
43
+ outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
44
+ print(outputs[0]["generated_text"])
45
+ ```
46
+ ## 🏆 Evaluation
47
+
48
+ | Tasks |Version|Filter|n-shot| Metric |Value | |Stderr|
49
+ |-------------------------------|-------|------|-----:|--------|-----:|---|-----:|
50
+ |stem |N/A |none | 0|acc |0.6466|± |0.0056|
51
+ | | |none | 0|acc_norm|0.6124|± |0.0066|
52
+ | - medmcqa |Yaml |none | 0|acc |0.6118|± |0.0075|
53
+ | | |none | 0|acc_norm|0.6118|± |0.0075|
54
+ | - medqa_4options |Yaml |none | 0|acc |0.6143|± |0.0136|
55
+ | | |none | 0|acc_norm|0.6143|± |0.0136|
56
+ | - anatomy (mmlu) | 0|none | 0|acc |0.7185|± |0.0389|
57
+ | - clinical_knowledge (mmlu) | 0|none | 0|acc |0.7811|± |0.0254|
58
+ | - college_biology (mmlu) | 0|none | 0|acc |0.8264|± |0.0317|
59
+ | - college_medicine (mmlu) | 0|none | 0|acc |0.7110|± |0.0346|
60
+ | - medical_genetics (mmlu) | 0|none | 0|acc |0.8300|± |0.0378|
61
+ | - professional_medicine (mmlu)| 0|none | 0|acc |0.7868|± |0.0249|
62
+ | - pubmedqa | 1|none | 0|acc |0.7420|± |0.0196|
63
+
64
+ |Groups|Version|Filter|n-shot| Metric |Value | |Stderr|
65
+ |------|-------|------|-----:|--------|-----:|---|-----:|
66
+ |stem |N/A |none | 0|acc |0.6466|± |0.0056|
67
  | | |none | 0|acc_norm|0.6124|± |0.0066|