language, license, pipeline_tag, tags, library_name, datasets, base_model

language	license	pipeline_tag	tags	library_name	datasets	base_model
en	apache-2.0	text-generation	llama causal-lm finetuned chytrej instruct tiny chatml	transformers	HuggingFaceTB/everyday-conversations-llama3.1-2k	pvlabs/Chytrej2-Mini

Chytrej2-Mini-It

A fine-tuned version of Chytrej2-Mini (20M params, LLaMA architecture) trained on conversational data. Don't expect great answers.

Built by PingVortex Labs.

---

Model Details

• Parameters: 20M
• Context length: 1024 tokens
• Language: English only
• Format: ChatML
• Base model: pvlabs/Chytrej2-Mini
• Architecture: LLaMA
• License: Apache 2.0

---

Training

Fine-tuned on HuggingFaceTB/everyday-conversations-llama3.1-2k dataset.

---

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_path = "pvlabs/Chytrej2-Mini-It"

tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path, dtype=torch.float16)
model.eval()

prompt = "<|im_start|>user\nHello<|im_end|>\n<|im_start|>assistant\n"
inputs = tokenizer(prompt, return_tensors="pt")

with torch.no_grad():
    output = model.generate(
        **inputs,
        max_new_tokens=200,
        do_sample=True,
        temperature=0.7,
        top_p=0.9,
        eos_token_id=tokenizer.convert_tokens_to_ids("<|im_end|>"),
        pad_token_id=tokenizer.eos_token_id,
    )

generated = tokenizer.decode(output[0][inputs["input_ids"].shape[1]:], skip_special_tokens=False)
print(generated)

---

Prompt Format (ChatML)

The model uses the standard ChatML format:

<|im_start|>user
Your message here<|im_end|>
<|im_start|>assistant

For multi-turn, chain turns:

<|im_start|>user
Hi!<|im_end|>
<|im_start|>assistant
Hello! How can I help you today?<|im_end|>
<|im_start|>user
What's 2+2?<|im_end|>
<|im_start|>assistant

---

Made by PingVortex.