Dominicvdb/pokerqwen

Model Card for Model ID

PokerBench Qwen3-8B LoRA Adapter

This repository contains a LoRA adapter for Qwen3-8B, fine-tuned on the PokerBench dataset for poker-related text generation and reasoning.

Model Details

Model Description

This is a poker-domain PEFT / LoRA adapter trained on top of Qwen3-8B using supervised fine-tuning (SFT).
It is intended to improve poker-related responses, including strategy discussion, hand analysis, poker terminology, and decision reasoning.

• Model type: LoRA adapter for a causal language model
• Base model: Qwen/Qwen3-8B
• Adapter format: PEFT adapter weights only
• Fine-tuning method: LoRA / SFT
• Libraries: Unsloth, PEFT, TRL, Transformers

Uses

Direct Use

This adapter is intended for poker-related generation tasks such as:

• Poker strategy explanation
• Hand and range analysis
• Pot odds and action reasoning
• Poker-focused assistant behavior

Downstream Use

This adapter can be used for:

• A poker-specialized assistant
• Domain adaptation experiments
• Further merging or adapter research workflows

Out-of-Scope Use

This adapter is not intended for:

• Legal, medical, or financial advice
• High-stakes decision-making
• Guaranteed optimal poker play
• General-purpose QA outside the poker domain

Bias, Risks, and Limitations

Because this is a domain-specific adapter, it may:

• Overemphasize poker interpretations
• Produce overconfident but incorrect advice
• Reflect limitations or biases in PokerBench
• Underperform on non-poker tasks relative to the base model

Users should treat outputs as experimental and verify important conclusions independently.

How to Get Started with the Model

This repository contains adapter weights only.
Load the base model first, then attach the adapter.

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model_name = "Qwen/Qwen3-8B"
adapter_name = "your-username/your-adapter-repo"

tokenizer = AutoTokenizer.from_pretrained(base_model_name)

base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.float16,
    device_map="auto",
)

model = PeftModel.from_pretrained(base_model, adapter_name)

prompt = "You are on the button with AKo facing an open raise. What factors matter most?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=128)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))