rizoner/rizoner-email-writer

Rizoner Email Writer (SFT)

Fine-tuned LoRA adapter for email writing, based on Qwen 2.5-14B-Instruct.

Training Details

• Base Model: unsloth/Qwen2.5-14B-Instruct
• Training Method: Supervised Fine-Tuning (SFT) with LoRA
• Training Data: 1067 email pairs from personal email history
• Training Framework: Unsloth + TRL
• Adapter Size: ~275MB

Usage

from unsloth import FastLanguageModel

# Load model with adapter
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="rizoner/rizoner-email-writer",  # Your HuggingFace repo
    max_seq_length=2048,
    dtype=None,
    load_in_4bit=True,
)

FastLanguageModel.for_inference(model)

# Generate email
prompt = """<|im_start|>system
You are an expert email writer. Write professional, clear, and contextually appropriate emails.<|im_end|>
<|im_start|>user
Compose email to John at Company (re: Project Update)<|im_end|>
<|im_start|>assistant
"""

inputs = tokenizer([prompt], return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7)
email = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(email)

API Usage (OpenAI-compatible)

from openai import OpenAI

client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1/",
    api_key="hf_..."  # Your HuggingFace token
)

response = client.chat.completions.create(
    model="rizoner/rizoner-email-writer",
    messages=[
        {"role": "system", "content": "You are an expert email writer."},
        {"role": "user", "content": "Compose email to John at Company (re: Meeting)"}
    ]
)

print(response.choices[0].message.content)

License

Apache 2.0