ztiganj/qwen25-14b-gbs-phase2-simnum-step80-bf16-compact
BF16 compact inference copy generated from an existing Hugging Face checkpoint repository.
Source
- Source repository:
ztiganj/qwen25-14b-gbs-phase2-simnum-step80
- Conversion: floating-point tensors in
.safetensors shards were recast to bfloat16.
- Purpose: reduce private storage use while preserving an inference-ready checkpoint.
Do not make this repository public without a separate release/safety review.
