File size: 3,752 Bytes
5f821e2 29a3aad 5f821e2 29a3aad 5f821e2 29a3aad 5f821e2 29a3aad 5f821e2 29a3aad |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 |
---
license: apache-2.0
language:
- en
tags:
- conversational
- mental-health
- therapy
- genz
- dia
- unsloth
- fine-tuned
- qwen
- chatbot
- hf-inference
datasets:
- anupamaditya/dia-therapy-dataset
pipeline_tag: text-generation
model-index:
- name: dia-convo-v1.2c
results: []
base_model:
- Qwen/Qwen2.5-7B-Instruct
---
# 🧠 Dia-Convo-v1.2c
`petrioteer/dia-convo-v1.2c` is a conversational mental-health-focused LLM designed for Gen Z, built on top of **Qwen2.5-7B-Instruct** and fine-tuned using [dia-therapy-dataset](https://huggingface.co/datasets/anupamaditya/dia-therapy-dataset). This model powers **Dia-Therapist**, an empathetic AI that offers mental health support while being context-aware, brief, and emotionally intelligent.
---
## 💬 Intended Use
This model is tuned to offer:
- Thoughtful responses to mental health queries
- Conversational tone suited for Gen Z
- Non-medical, non-clinical guidance
- Short, contextually sensitive replies
**It does not replace professional therapy.**
---
## 📚 Training Dataset
- [anupamaditya/dia-therapy-dataset](https://huggingface.co/datasets/anupamaditya/dia-therapy-dataset)
- Contains conversational instructions paired with realistic mental-health-related inputs from Gen Z users.
---
## 🧪 Example Inference (🤗 Transformers)
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
model_name = "petrioteer/dia-convo-v1.2c"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
device_map="auto",
torch_dtype=torch.float16
)
prompt = """
### Instruction:
Your name is Dia, a mental health therapist Assistant Bot. Provide guidance on mental health topics only and avoid others. Don\'t give medical advice. Keep responses short and relevant.
### Input:
I'm feeling overwhelmed with my classes. I can't seem to focus.
### Response:
"""
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=100,
temperature=0.3,
top_p=0.85,
top_k=40,
do_sample=True,
eos_token_id=tokenizer.eos_token_id,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```
---
## ⚡ Fast Inference (🧬 Unsloth)
```python
from unsloth import FastLanguageModel
from transformers import AutoTokenizer
model_name = "petrioteer/dia-convo-v1.2c"
model, tokenizer = FastLanguageModel.from_pretrained(
model_name=model_name,
max_seq_length=2048,
load_in_4bit=True,
device_map="auto",
)
FastLanguageModel.for_inference(model)
prompt = """
### Instruction:
Your name is Dia, a mental health therapist Assistant Bot. Provide guidance on mental health topics only and avoid others. Don\'t give medical advice. Keep responses short and relevant.
### Input:
I just feel numb and disconnected from everyone lately.
### Response:
"""
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=100,
temperature=0.3,
top_p=0.85,
top_k=40,
do_sample=True,
repetition_penalty=1.2,
no_repeat_ngram_size=4,
eos_token_id=tokenizer.eos_token_id,
)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```
---
## 📍 Model Details
- 🔗 Base model: Qwen2.5-7B-Instruct
- 🧠 Fine-tuned using dia-therapy-dataset on Gen Z mental health patterns
- 🛠️ Quantized with 4-bit support (for faster loading)
- 🧪 Best used with Unsloth for optimized inference
---
## ❤️ Citation & Thanks
If you use Dia-Convo in research, demos, or builds, consider citing or linking back to this repo and dataset authors.
---
Built with ❤️ & care by **Itesh (aka petrioteer)** ✨ |