You can deploy this model and start earning money today!

AfkaraLP/
gemma-3-270M-coralmc-1800it-8batch-v12

Name: gemma-3-270M-coralmc-1800it-8batch-v12
Availability: InStock

Views

No views yet

gemma-3-270M-coralmc-1800it-8batch-v12 : GGUF

This model was finetuned and converted to GGUF format using Unsloth.

Example usage:

For text only LLMs: ./llama.cpp/llama-cli -hf AfkaraLP/gemma-3-270M-coralmc-1800it-8batch-v12 --jinja
For multimodal models: ./llama.cpp/llama-mtmd-cli -hf AfkaraLP/gemma-3-270M-coralmc-1800it-8batch-v12 --jinja

Prompting Guide for CoralMC-Gemma-270M

This model was fine-tuned on chat logs from the CoralMC Minecraft server. It specializes in simulating server chatter, handling "griefer/clan" slang, and roleplaying specific users.

Because the training data followed a strict format, you must structure your prompts specifically to get high-quality results. The model relies heavily on the System Prompt to hold the context history.

1. Unconditional Chat Generation (Creative Mode)

Use this mode to generate random, realistic server banter from scratch without providing any background context.

System Prompt: (Leave Empty)
User Prompt: Generate a CoralMC chat snippet

Example Output:

text

[xX_Dragon_Xx]: anyone for bedwars?
[Steve123]: no
[Admin_Bob]: Please do not spam in the lobby.
[xX_Dragon_Xx]: 1v1 me then

2. Conversation Continuation

Use this mode when you have a snippet of chat and want the model to predict what happens next.

User Prompt: Continue the conversation
System Prompt: You must format the system prompt using the Context: and ### headers exactly as shown below:

text

Context:
[User_A]: hop on
[User_B]: wait im launching
[User_A]: hurry up the event starts in 5
### Next Message: [User_A]: hurry up the event starts in 5
### Continue Chat:

Note: The ### Next Message: line should repeat the very last message from the context.

3. Targeted User Response (Roleplay)

Use this mode to force the model to respond as a specific user (or a specific username you invent) based on the chat history.

User Prompt: Generate response
System Prompt: Use the ### Next Speaker: header to define who is talking.

text

Context:
[FactionLeader]: who raided the base?
[NewGuy]: i think it was team red
### Next Message: [NewGuy]: i think it was team red
### Next Speaker: [FactionLeader]

Expected Output:

text

[FactionLeader]: are you sure? coords?

Available Model files:

gemma-3-270m-it.Q8_0.gguf

Ollama

An Ollama Modelfile is included for easy deployment.

Note

The model's BOS token behavior was adjusted for GGUF compatibility. This was trained 2x faster with Unsloth

AfkaraLP/gemma-3-270M-coralmc-1800it-8batch-v12

gemma-3-270M-coralmc-1800it-8batch-v12 : GGUF

Prompting Guide for CoralMC-Gemma-270M

1. Unconditional Chat Generation (Creative Mode)

2. Conversation Continuation

3. Targeted User Response (Roleplay)

Available Model files:

Ollama

Note

AfkaraLP/
gemma-3-270M-coralmc-1800it-8batch-v12