your-username/affine-custom-model-v1
A custom-trained reasoning model optimized for the Affine subnet on Bittensor.
Model Details
- Architecture: GPT-OSS based transformer
- Training: Fine-tuned on multi-task reasoning datasets
- Optimization: Custom RL training for Affine environments
- Modified: 2025-11-13
Environments
Optimized for:
- SAT solving
- Abductive reasoning (ABD)
- Deductive reasoning (DED)
- ALFWorld navigation
- WebShop interaction
- BabyAI tasks
- SciWorld experiments
- TextCraft games
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained(
"your-username/affine-custom-model-v1",
torch_dtype="auto",
device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("your-username/affine-custom-model-v1")
messages = [{"role": "user", "content": "Solve this SAT problem..."}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt")
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0]))
Performance
This model achieves competitive performance across all Affine evaluation environments.
License
Apache 2.0
- Downloads last month
- 16