jyohukuchan/qwen38-flash-next-bf16-logits
Qwen3.8-Flash-Next BF16 reference logits (KL evaluation set) Full-vocabulary FP32 logits of the official BF16 model Qwen/Qwen3.8-Flash-Next (revision de4b8e4d43b917e7706784d8bb445c9af86a3540), captured so quantized checkpoints (EXL3 and others) can be scored by position-level KL divergence KL(reference || candidate) without renting a 360 GB BF16 run. Contents reference/ — BF16 reference logits (exllamav3 engine, BF16 weights), 21 cases / 32359 positions. Raw FP32… See the full description on the dataset page: https://huggingface.co/datasets/jyohukuchan/qwen38-flash-next-bf16-logits.
| Author / organization | jyohukuchan |
|---|---|
| License (as declared) | other |
| Downloads | 0 (as reported by Hugging Face) |
| Likes | 1 |
| Created | 24 October 2026 |
| Last updated at source | 24 October 2026 |
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.