eshonindex/DiscordLM-0.5B

DiscordLM 0.5B

A fine-tuned language model specialized in Discord platform knowledge — permissions, bot development, API usage, moderation, server management, and troubleshooting.

Fine-tuned from Qwen/Qwen2.5-0.5B-Instruct using LoRA on ~1,600 curated Discord documentation examples.

Quantized Version

A WebLLM-ready MLC quantized version (q4f16_1, ~280MB) is available at: eshonindex/DiscordLM-0.5B-q4f16_1-MLC

Usage

With Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("eshonindex/DiscordLM-0.5B")
tokenizer = AutoTokenizer.from_pretrained("eshonindex/DiscordLM-0.5B")

messages = [
    {"role": "system", "content": "You are a Discord expert assistant."},
    {"role": "user", "content": "How do Discord permissions work?"}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

With WebLLM (Browser)

import * as webllm from "@mlc-ai/web-llm";

const engine = new webllm.MLCEngine();
await engine.reload("eshonindex/DiscordLM-0.5B-q4f16_1-MLC");

const reply = await engine.chat.completions.create({
  messages: [{ role: "user", content: "How do I set up a Discord bot?" }],
});
console.log(reply.choices[0].message.content);

Training Details

• Base model: Qwen/Qwen2.5-0.5B-Instruct (494M params)
• Method: LoRA (rank=16, alpha=32)
• Dataset: ~1,591 examples from Discord documentation, support articles, and curated Q&A
• Training: 2-5 epochs, lr=2e-4, cosine schedule
• Latest version: DiscordLM-aligned-v1

Example Questions

• "What are Discord permissions and how do they work?"
• "How do I create a Discord bot?"
• "Explain Discord's rate limits for the API"
• "What is the difference between server roles and channel overrides?"
• "How do I set up automod in Discord?"