---
license: apache-2.0
language:
- en
pipeline_tag: text-generation
library_name: transformers
---



# Pico Mini V1

Pico v1 is a work in progress model. Based off Qwen 2.5 .5b model, it has been fine tuned for automatic COT and self reflection. 

When making a output, Pico will create three sections, a reasoning section, a self-reflection section and a output section.

Pico Mini v1 struggles with non-question related tasks (Small talk, roleplay, etc).

Pico Mini v1 can struggle with staying on topic at times.

Here is a example of how you can use it:

```from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the model and tokenizer from the Hugging Face Model Hub (test/test repository)
output_dir = "test/test"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

print("Loading the model and tokenizer from the Hugging Face Hub...")
model = AutoModelForCausalLM.from_pretrained(output_dir).to(device)  # Ensure model is on the same device
tokenizer = AutoTokenizer.from_pretrained(output_dir)

# Define the testing prompt
prompt = "What color is the sky?"
print(f"Testing prompt: {prompt}")

# Tokenize input and move to the same device as the model
inputs = tokenizer(prompt, return_tensors="pt").to(device)  # Ensure inputs are on the same device

# Generate response
print("Generating response...")
outputs = model.generate(
    **inputs,
    max_new_tokens=1550,  # Adjust the max tokens if needed
    temperature=0.5,      # Adjust for response randomness
    top_k=50,             # Adjust for top-k sampling
    top_p=0.9             # Adjust for nucleus sampling
)

# Decode and print the response
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print("Generated response:")
print(response)

```