VDC-team/VDrontV2-0.1b-Instruct

CC0-1.0

The VDrontV2-0.1b-Instruct - an additional trained version of VDrontV2-0.1, trained on TIGER-lab/FineWeb-Instruct.

GPT-like, GPT-2 Tokenizer

Chat format - <|user|> and <|assistant|>

Size - +-100 000 000 params

FineTune 150 0000 instructions TIGER-lab/FineWeb-Instruct

Context increased to 1024

Guide use - download repo, place all repository files except use.py into a local folder named model/, place use.py next to that folder, and run it.

Script easy start:

import os, torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from huggingface_hub import snapshot_download
model_path = "model"
if not os.path.exists(model_path) or not os.listdir(model_path):
    snapshot_download(repo_id="VDC-team/VDrontV2-0.1b-Instruct", local_dir=model_path, ignore_patterns=["*.md", ".gitattributes"], resume_download=True)
device = "cuda" if torch.cuda.is_available() else "cpu"
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True)
tokenizer.pad_token = tokenizer.eos_token
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16 if device == "cuda" else torch.float32, low_cpu_mem_usage=True).to(device).eval()
print("Chat ready (type exit to quit)")
conversation = ""
while True:
    user_input = input("You: ")
    if user_input.lower() == "exit": break
    conversation += f"<|user|>{user_input}<|assistant|>"
    inputs = tokenizer(conversation, return_tensors="pt").to(device)
    with torch.no_grad():
        outputs = model.generate(**inputs, max_new_tokens=256, do_sample=True, temperature=0.5, top_p=0.94, pad_token_id=tokenizer.eos_token_id)
    response = tokenizer.decode(outputs[0][len(inputs.input_ids[0]):]).split("<|user|>")[0].split("<|assistant|>")[0].replace("<|endoftext|>", "").strip()
    print(f"Assistant: {response}")
    conversation += response

Examples (use.py)

Instruction: What is a Python?
Response: A Python is an open-source, multi-threaded programming language that allows developers to build
applications and implement their own programs. It provides a platform for creating and
deploying programs, making it easier for developers to build and deploy. Python is the world’s most popular programming language
and one of the most widely used programming languages in the world. It is also known for its ability to build and execute complex applications.
Python is also known for its ability to create dynamic code, allowing developers to create and execute applications. Python is also known for its ability
to automate and execute tasks, such as writing and managing files, designing and executing software, and managing file systems. Python is also widely used to
manage files, files, and applications. Python is also used to create and deploy programs, which are executed using the Python language. Python is also used to
create and execute executable files, such as files, images, and other data. Python is also used to create and deploy applications, such as file management,
database management, and database management. 

Instruction: A recipe for making simple bread.
Response: The recipe involves adding a tablespoon of wheat flour to the dough, creating a simple and flavorful bread. The dough is then baked in a bowl of warm water, allowing the dough to cool and dry. The dough is then cooked into a small saucepan, making it a great way to add some flavor to your pasta dishes.    

The recipe emphasizes that it's a delicious and easy recipe, but it's not a substitute for traditional pasta or pasta.  The recipe suggests that you can make pasta using a combination of ingredients like diced garlic, garlic, and basil, making it a versatile and flavorful pasta dish.  

The recipe highlights the importance of using ingredients like chopped onions, garlic, garlic, and olive oil for a healthy and delicious meal. 

The recipe emphasizes the importance of using ingredients like garlic, garlic, garlic, and olive oil for a healthy and satisfying meal.

Instruction: What is a AI?
Response: A AI is an artificial intelligence that uses AI to understand, diagnose, and solve problems. It can be used to identify patterns, trends, and patterns, making it more effective than traditional methods. AI can also be used to analyze and solve complex mathematical problems, such as complex equations, equations, or even mathematical equations. AI can help with many different aspects of the human experience, from understanding the relationships between objects and their surroundings, to understanding the world around us.       
AI can be used to analyze data, analyze the data, and make predictions based on the results of various scenarios. This can involve analyzing the data, analyzing the data, and interpreting it. AI can also help with various aspects of the human experience, including the perception of the world, the ability to understand the world, and how the world works.       
AI can also be used to analyze data, such as the ability to recognize patterns or patterns in the environment, or to predict the future.    AI can also be used to analyze data, such as the ability to predict patterns or patterns in the world, to predict the future, or to predict the future.      
AI can also be used to analyze and analyze data, such as the ability to predict the future, to predict the future, to predict the future, to predict the future, or to predict the future.     
AI can also be used to analyze data, such as the ability to predict the future, to predict the future, or to predict the future, to predict the future.

The model has issues with stopping generation, so it is best to stop it manually once the desired number of tokens has been generated.

If you create your own model and want it included in the VDrontV2 collection, please indicate that you used this model.