Odysseus 2 Chat

A chat-tuned version of Odysseus 2, fine-tuned with LoRA on the Alpaca-cleaned dataset (51,760 examples).

Model Details

• Base model: dogman189/odysseus-2
• Fine-tuning: LoRA (rank 8, alpha 16) on Alpaca-cleaned
• Architecture: GPT-2 (GPT2LMHeadModel)
• Parameters: ~320M
• Context length: 512 tokens
• Vocabulary: 16,000 BPE tokens

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "dogman189/odysseus-2-chat"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

prompt = "User: What is 2+2?\n\nAssistant:"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Chat Format

The model was fine-tuned on the format:

User: {instruction}

Assistant: {response}

Training Details

• LoRA fine-tune of the base model on alpaca-cleaned (51,760 examples)
• LoRA rank 8, alpha 16, dropout 0.1, targeting c_proj, c_fc, c_attn
• 6 epochs (default)

Limitations

Small model fine-tuned on a limited chat dataset. Responses may be inaccurate or nonsensical. Not intended for production use.