tinygoop-1.1b

Model Description

A fine-tuned version of TinyLlama-1.1B-Chat with room temp iq -> quantized to 4 bits and trained on copypastas

Intended Use

• Primary Use: Not much, it barely can hold a conversation
• Secondary Uses: brainrot generation, funny responses
• Out-of-scope: Professional/business applications, factual question answering, safety-critical applications

---

Training Data

Sources:

• 334,165 copypastas
• The script from the television show "House"

Hardware used in training

• GPU: NVIDIA GeForce RTX 4090
• CUDA: 12.1
• Framework: PyTorch 2.5.1+cu121
• Transformers: Latest
• PEFT: Latest
• BitsAndBytes: 4-bit quantization

---

Basic Usage

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "S-teven/tinygoop-1.1b"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

prompt = "hey"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    do_sample=True,
    temperature=1.2,
    top_p=0.95,
    repetition_penalty=1.05
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Hardware Requirements

Precision	VRAM Required	Hardware
4-bit Quantized	~800MB	Any modern GPU
CPU (FP32)	~4GB RAM	Modern CPU (slow)

---

Limitations & Biases

Content Warning: This model was trained on copypasta data and may generate:

• Offensive or inappropriate content
• Nonsensical or chaotic responses
• Biases present in online communities

Not suitable for:

• Most things
• Professional or business use
• Educational applications
• Factual information retrieval
• Content requiring safety guarantees