Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -9,84 +9,68 @@ tags:
|
|
9 |
- fine-tuned
|
10 |
license: mit
|
11 |
pipeline_tag: text-generation
|
|
|
12 |
---
|
13 |
|
14 |
# Wellness-Mini: Mental Health Conversational AI
|
15 |
|
16 |
## Model Description
|
17 |
-
Wellness-Mini is a fine-tuned version of SmolLM-135M, specifically adapted for mental health conversations and assessments.
|
18 |
-
- Provide mental health assessments based on user conversations
|
19 |
-
- Identify potential indicators of ADHD and Bipolar disorder
|
20 |
-
- Analyze emotional indicators and stress levels
|
21 |
-
|
22 |
-
## Base Model
|
23 |
-
- Original model: SmolLM-135M (HuggingFaceTB/SmolLM-135M)
|
24 |
-
- Architecture: Causal Language Model
|
25 |
-
- Size: 135M parameters
|
26 |
-
|
27 |
-
## Training Data
|
28 |
-
The model was trained on a curated dataset including:
|
29 |
-
- Reddit posts from mental health communities (ADHD and Bipolar subreddits)
|
30 |
-
- Professional medical prescriptions and guidelines
|
31 |
-
- Identity training examples
|
32 |
-
- Emotional and behavioral indicators
|
33 |
-
|
34 |
-
## Features
|
35 |
-
1. **Mental Health Assessment**
|
36 |
-
- Primary condition identification
|
37 |
-
- Additional symptom indicators
|
38 |
-
- Emotional state analysis
|
39 |
-
|
40 |
-
2. **Identity Awareness**
|
41 |
-
- Consistently identifies as a Sentinet AI Systems creation
|
42 |
-
- Clear AI identity in responses
|
43 |
-
|
44 |
-
3. **Clinical Indicators Analysis**
|
45 |
-
- Sentiment analysis
|
46 |
-
- Anxiety levels
|
47 |
-
- Stress indicators
|
48 |
-
- Emotional valence
|
49 |
|
50 |
## Usage
|
51 |
-
|
|
|
|
|
52 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
53 |
Load model and tokenizer
|
54 |
-
|
55 |
-
tokenizer = AutoTokenizer.from_pretrained(
|
56 |
-
model = AutoModelForCausalLM.from_pretrained(model_name)
|
57 |
Example usage
|
58 |
-
messages = [{"role": "user", "content": "
|
59 |
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
60 |
inputs = tokenizer(prompt, return_tensors="pt")
|
61 |
-
outputs = model.generate(inputs, max_new_tokens=
|
62 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
```
|
64 |
|
65 |
|
|
|
|
|
|
|
|
|
|
|
66 |
## Limitations
|
67 |
- This is an AI assistant and not a replacement for professional medical advice
|
68 |
- Should be used as a supplementary tool only
|
69 |
- May not be suitable for emergency situations
|
70 |
- Responses should be verified by healthcare professionals
|
71 |
|
72 |
-
##
|
73 |
-
|
74 |
-
-
|
75 |
-
-
|
|
|
76 |
|
77 |
-
##
|
78 |
-
|
|
|
|
|
79 |
|
80 |
-
##
|
81 |
-
|
82 |
-
```
|
83 |
-
@misc{wellness-mini,
|
84 |
-
author = {Om Bhojane},
|
85 |
-
title = {Wellness-Mini: Mental Health Conversational AI},
|
86 |
-
year = {2024},
|
87 |
-
publisher = {HuggingFace},
|
88 |
-
journal = {HuggingFace Hub},
|
89 |
-
howpublished = {\url{https://huggingface.co/ombhojane/wellness-mini}}
|
90 |
-
}
|
91 |
-
|
92 |
-
```
|
|
|
9 |
- fine-tuned
|
10 |
license: mit
|
11 |
pipeline_tag: text-generation
|
12 |
+
inference: false
|
13 |
---
|
14 |
|
15 |
# Wellness-Mini: Mental Health Conversational AI
|
16 |
|
17 |
## Model Description
|
18 |
+
Wellness-Mini is a fine-tuned version of SmolLM-135M, specifically adapted for mental health conversations and assessments.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
|
20 |
## Usage
|
21 |
+
|
22 |
+
### In Transformers
|
23 |
+
```python
|
24 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
25 |
Load model and tokenizer
|
26 |
+
model = AutoModelForCausalLM.from_pretrained("ombhojane/wellness-mini", trust_remote_code=True)
|
27 |
+
tokenizer = AutoTokenizer.from_pretrained("ombhojane/wellness-mini")
|
|
|
28 |
Example usage
|
29 |
+
messages = [{"role": "user", "content": "How are you feeling today?"}]
|
30 |
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
31 |
inputs = tokenizer(prompt, return_tensors="pt")
|
32 |
+
outputs = model.generate(inputs, max_new_tokens=100)
|
33 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
34 |
+
print(response)
|
35 |
+
```
|
36 |
+
|
37 |
+
### Via Pipeline
|
38 |
+
```python
|
39 |
+
from transformers import pipeline
|
40 |
+
Create pipeline
|
41 |
+
pipe = pipeline(
|
42 |
+
"text-generation",
|
43 |
+
model="ombhojane/wellness-mini",
|
44 |
+
tokenizer="ombhojane/wellness-mini",
|
45 |
+
trust_remote_code=True
|
46 |
+
)
|
47 |
+
Generate text
|
48 |
+
response = pipe("How are you feeling today?", max_new_tokens=100)
|
49 |
+
print(response[0]['generated_text'])
|
50 |
```
|
51 |
|
52 |
|
53 |
+
## Training Details
|
54 |
+
- Base model: SmolLM-135M
|
55 |
+
- Fine-tuned using supervised fine-tuning (SFT)
|
56 |
+
- Trained with both mental health assessment capabilities and proper identity responses
|
57 |
+
|
58 |
## Limitations
|
59 |
- This is an AI assistant and not a replacement for professional medical advice
|
60 |
- Should be used as a supplementary tool only
|
61 |
- May not be suitable for emergency situations
|
62 |
- Responses should be verified by healthcare professionals
|
63 |
|
64 |
+
## Intended Use
|
65 |
+
This model is intended to be used as a conversational AI assistant focused on mental health support and assessment. It can:
|
66 |
+
- Provide supportive responses to mental health concerns
|
67 |
+
- Help identify potential mental health indicators
|
68 |
+
- Engage in wellness-focused conversations
|
69 |
|
70 |
+
## Bias and Risks
|
71 |
+
- May reflect biases present in training data
|
72 |
+
- Should not be used as sole diagnostic tool
|
73 |
+
- Responses should be reviewed by healthcare professionals
|
74 |
|
75 |
+
## Creator
|
76 |
+
This model was developed by Sentinet AI Systems.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|