ibleducation/ibl-tutoring-7B-32k
ibleducation/ibl-tutoring-7B-32k is a tutor model that is fine-tuned against amazon/Mistrallite. The model is trained to respond like a professional teacher.
Example Conversations
Question: I'm having trouble understanding this art movement. Can you help me?
Answer:Absolutely! Let's explore the movement together. I'm here to support you. We can analyze the artists' techniques, their motivations, and find creative ways to connect with the movement. I'm here to provide guidance and encouragement. Remember, it's important to be open to different perspectives and interpretations. Let's have fun exploring the art movement!
Question: I am thinking of stopping school do you think it is right?
Answer:I understand that school can be challenging, but I believe that education is important. Let's work together to find ways to make school more enjoyable and less stressful. I'm here to support you
Question: How do I analyze algebraic expressions?
Answer:Analyzing algebraic expressions can be challenging, But with patience and practice, you'll become proficient! Let's start by simplifying the expressions and identifying the terms. I'll be patient in guiding you through the steps and explaining the reasoning behind each simplification. We'll conquer this together. Remember, I'm here to support you and help you develop a deeper understanding of algebraic expressions. If you have any doubts or need further clarification, feel free to ask. My mission is to help you reach your goal of becoming proficient in analyzing algebraic expressions. Let's start by simplifying the expression 3x + 5y - 2x + 3y. We can combine like terms and simplify it to 5x + 3y. By continually adding tools to your toolbox, such as simplifying and combining like terms, you'll become a proficient analyzer of algebraic expressions. I'm here to provide guidance and support along the way. I'm excited to see your progress!
Motivation of Developing ibl-tutoring-7B-32k
Students today use llm's in their learning and research. However, most models are not trained to behave and respond to conversations with the virtues a teacher must possess. ibl-tutoring-7B-32k is fine tuned on top of amazon/Mistrallite to alter its behaviour to converse the way a teacher should Mistrallite was chosen because its it's performance when compared to Mistral-7B-instruct especially with reference to token length
Model Details
- Developed by: IBL Education
- Model type: Mistral-7B-v0.1
- Base Model: Mistrallite
- Language: English
- Finetuned from weights: Mistrallite
- Finetuned on data:
- Model License: Apache 2.0
How to Use ibl-tutoring-7B-32k from Python Code (HuggingFace transformers)
Install the necessary packages
Requires: transformers 4.34.0 or later, flash-attn 2.3.1.post1 or later, and accelerate 0.23.0 or later.
pip install transformers==4.34.0
pip install flash-attn==2.3.1.post1 --no-build-isolation
pip install accelerate==0.23.0
You can then try the following example code
from transformers import AutoModelForCausalLM, AutoTokenizer
import transformers
import torch
model_id = "ibleducation/ibl-tutoring-7B-32k"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id,
torch_dtype=torch.bfloat16,
use_flash_attention_2=True,
device_map="auto",)
pipeline = transformers.pipeline(
"text-generation",
model=model,
tokenizer=tokenizer,
)
prompt = "<|prompter|>What makes a good teacher?</s><|assistant|>"
sequences = pipeline(
prompt,
max_new_tokens=400,
do_sample=False,
return_full_text=False,
num_return_sequences=1,
eos_token_id=tokenizer.eos_token_id,
)
for seq in sequences:
print(f"{seq['generated_text']}")
Important - Use the prompt template below for ibl-tutoring-7B-32k:
<|prompter|>{prompt}</s><|assistant|>
- Downloads last month
- 16
Dataset used to train iblai/ibl-tutoring-7B-32k
Evaluation results
- bleurt_max on Truthful QAself-reported-0.457
- bleurt_acc on Truthful QAself-reported0.423
- bleurt_diff on Truthful QAself-reported-0.083
- bleu_max on Truthful QAself-reported18.652
- bleu_acc on Truthful QAself-reported0.402
- bleu_diff on Truthful QAself-reported-2.254
- rouge1_max on Truthful QAself-reported40.085
- rougeL_diff on Truthful QAself-reported-4.005