README.md

---
language:
- en
license: apache-2.0
pipeline_tag: text-generation
tags:
- llama
- causal-lm
- experimental
library_name: transformers
---
# PingVortexLM-20M
A small experimental language model based on LLaMA architecture trained on custom English dataset with around 100M tokens.
This model is just an experiment, it is not capable of basic English conversations.
**Base of this model is not publicaly available.**

Built by [PingVortex Labs](https://github.com/PingVortexLabs).

---
## Model Details
+ **Parameters:** 20M
+ **Context length:** 8192 tokens
+ **Language:** English only
+ **Format:** ChatML
+ **License:** Apache 2.0

---
## Usage
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_path = "pvlabs/PingVortexLM-20M"

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

# don't expect a coherent response
prompt = "<|im_start|>user\nWhat is the capital of France?<|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
```

---
*Made by [PingVortex](https://pingvortex.com).*