CompassioninMachineLearning/Qwen3-8b-compassion-seed1-20260930-CPT-merged-epoch-2
BrandonHowe/Qwen3-8b-compassion-qwen-seed1-20260930-CPT-merged-epoch-2 is an 8 billion parameter Qwen3-based language model, fine-tuned by BrandonHowe on a specialized dataset focused on compassion. This model, with a 32768 token context length, is designed to explore and potentially generate text related to the concept of compassion, utilizing a dataset of 10,000 distinct documents. It is provided as a standalone BF16 merged model, ready for direct loading without adapters.
Loading preview...
Model Overview
This model, BrandonHowe/Qwen3-8b-compassion-qwen-seed1-20260930-CPT-merged-epoch-2, is an 8 billion parameter language model based on the Qwen3 architecture. It has been fine-tuned by BrandonHowe, specifically focusing on the concept of compassion.
Key Characteristics
- Architecture: Qwen3-based, 8 billion parameters.
- Context Length: Supports a substantial context window of 32768 tokens.
- Training Data: Fine-tuned on the
CompassioninMachineLearning/compassion_12185_cleaneddataset, comprising 10,000 distinct documents with 2,000 repeat exposures per epoch. - Format: Provided as a standalone, merged BF16 model (epoch 2.0, step 750), packaged into eight safetensors shards. It does not require an adapter for loading.
- Validation: Weights are validated BF16, ensuring lossless packaging.
Intended Use and Evaluation
This model is specifically designed for exploring and generating content related to compassion. It's important to note that the training process did not inherently establish an improvement in compassion; users are encouraged to evaluate its performance in this domain separately. Detailed training parameters, base revision, and document selection hashes are available in the run_manifest.json file.