sn92-df3
Demand-forecasting specialist for the REAX subnet (SN92). It is a linear weight merge of two Apache-2.0 models on the same Qwen3.5-9B text layout, followed by a letter-logit scale.
| model A (declared base) | whynotBeTop/sn92-df-dm9b @ 362dfaa469b0dec69f08e75806eb2c27baef1c97 |
| model B | danyloooah/df-specialist-9b-a3 @ cadca463919e71a772b2cc5c9a3b4bd2d2238a64 |
| merge | 0.7 * A + 0.3 * B, tensor by tensor (fp32 arithmetic, bf16 output); config and tokenizer from A |
| readout | the 26 option-letter rows (A..Z) of lm_head multiplied by 2.0 (exact in bf16) |
All credit for the underlying fine-tunes goes to the authors of the two models above; this repository only combines them. Both are released under Apache-2.0, and so is this merge (see LICENSE).
- Downloads last month
- 25
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support