---
language:
- vi
license: afl-3.0
library_name: transformers
tags:
- NLP
- Vietnamese
base_model: Viet-Mistral/Vistral-7B-Chat
datasets:
- Tamnemtf/hcmue_qa
pipeline_tag: text-generation
---

# Model Card for Model ID

Chatbots can be programmed with a large knowledge base on answer users' questions on a variety of topics. They can provide facts, data, explanations, definitions, etc.
Complete tasks. Chatbots can be integrated with other systems and APIs to actually do things for users. Based on a user's preferences and past interactions, chatbots can suggest products, services, content and more that might be relevant and useful to the user.
Provide customer service. Chatbots can handle many simple customer service interactions to answer questions, handle complaints, process returns, etc. This allows human agents to focus on more complex issues.
Generate conversational responses - Using NLP and machine learning, chatbots can understand natural language and generate conversational responses, creating fluent interactions.



## Model Details

### Model Description

<!-- Provide a longer summary of what this model is. -->
- **Model type:** Mistral
- **Language(s) (NLP):** Vietnamese
- **Finetuned from model :** [Viet-Mistral/Vistral-7B-Chat](https://huggingface.co/Viet-Mistral/Vistral-7B-Chat)

### Purpose
This model is fine-tuned in oder to serve our scientific research topics which is created a chatbot to serve student to know about univesity information. This chatbot is a virtual assistant for students to help answer questions and resolve students' concerns.

### Training Data

Our dataset was make base on our university sudent notebook. It includes majors, university regulations and other information about our university.  
[hcmue_qa](https://huggingface.co/datasets/Tamnemtf/hcmue_qa)

### Training Procedure 

```python
  # LoRA attention dimension
lora_r = 64

# Alpha parameter for LoRA scaling
lora_alpha = 16

# Dropout probability for LoRA layers
lora_dropout = 0.1

################################################################################
# bitsandbytes parameters
################################################################################

# Activate 4-bit precision base model loading
use_4bit = True

# Compute dtype for 4-bit base models
bnb_4bit_compute_dtype = "float16"

# Quantization type (fp4 or nf4)
bnb_4bit_quant_type = "nf4"

# Activate nested quantization for 4-bit base models (double quantization)
use_nested_quant = False

################################################################################
# TrainingArguments parameters
################################################################################

# Output directory where the model predictions and checkpoints will be stored
output_dir = "./results"

# Number of training epochs
num_train_epochs = 1

# Enable fp16/bf16 training (set bf16 to True with an A100)
fp16 = False
bf16 = True

# Batch size per GPU for training
per_device_train_batch_size = 2

# Batch size per GPU for evaluation
per_device_eval_batch_size = 2

# Number of update steps to accumulate the gradients for
gradient_accumulation_steps = 1

# Enable gradient checkpointing
gradient_checkpointing = True

# Maximum gradient normal (gradient clipping)
max_grad_norm = 0.3

# Initial learning rate (AdamW optimizer)
learning_rate = 2e-4

# Weight decay to apply to all layers except bias/LayerNorm weights
weight_decay = 0.001

# Optimizer to use
optim = "paged_adamw_32bit"

# Learning rate schedule (constant a bit better than cosine)
lr_scheduler_type = "constant"

# Number of training steps (overrides num_train_epochs)
max_steps = -1

# Ratio of steps for a linear warmup (from 0 to learning rate)
warmup_ratio = 0.03

# Group sequences into batches with same length
# Saves memory and speeds up training considerably
group_by_length = True

# Save checkpoint every X updates steps
save_steps = 25

# Log every X updates steps
logging_steps = 25

################################################################################
# SFT parameters
################################################################################

# Maximum sequence length to use
max_seq_length = None

# Pack multiple short examples in the same input sequence to increase efficiency
packing = False

# Load the entire model on the GPU 0
device_map = {"": 0}
```

## Contact

nguyndantdm6@gmail.com