ModelVerificadoExterno

Kimi K2.7 Code DFlash

DFlash speculative-decoding draft model for moonshotai/Kimi-K2.7-Code, trained with SpecForge (PR #593) on NVIDIA Nemotron-Post-Training-Dataset-v2 (stem+chat+math+code).

Fuentehuggingface.co/cm00cm/Kimi-K2.7-Code-DFlash/tree/07290209664305eae6c119a3c0ab5c1c8743a40b

TipoOtherModelo basemoonshotai/Kimi-K2.7-CodeVersiónRevision 0729020Publicador@cm00cmCcalidad

Fuente del modelo

Tipo: Other
Tamaño del archivo: 6,48 GB
Modelo base: moonshotai/Kimi-K2.7-Code
Versión: vRevision 0729020
Actualizado: 2 oct 2026
Publicador: @cm00cm
Descargas: 28
Fuente: Hugging Face

Descripción de la fuente

Kimi-K2.7-Code DFlash draft

DFlash speculative-decoding draft model for moonshotai/Kimi-K2.7-Code, trained with SpecForge (PR #593) on NVIDIA Nemotron-Post-Training-Dataset-v2 (stem+chat+math+code).

6-layer Qwen3-style draft (hidden 7168); consumes target hidden states at layers [1,12,24,35,47,58]; block_size 8.
Target vocab/tokenizer: Kimi-K2.7-Code (vocab 163840, mask_token_id 163838).
Checkpoint: epoch_4_step_334000 — Work-in-progress snapshot (epoch_4_step_334000) — training still running.

Load with trust_remote_code=True (model code in dflash.py). Intended as the draft in SGLang DFlash speculative decoding paired with the Kimi-K2.7-Code target.

Fuentes

1 fuente

Hugging FacePrincipalVersión 07290209664305eae6c119a3c0ab5c1c8743a40b

huggingface.co/cm00cm/Kimi-K2.7-Code-DFlash/tree/07290209664305eae6c119a3c0ab5c1c8743a40b

Verificado 19 ago

Artefactos del modelo

1 artefacto

model.safetensors

safetensors · 6,48 GB · SHA-256 a2602716a72c…b50c · Hugging Face

Descargar

Extractos de fuentes

3 extractos

Readme11 julhuggingface.co/cm00cm/Kimi-K2.7-Code-DFlash

--- license: other base_model: moonshotai/Kimi-K2.7-Code library_name: transformers tags: - speculative-decoding - dflash - eagle - draft-model - kimi-k2 - specforge --- # Kimi-K2.7-Code DFlash draft DFlash speculative-decoding **draft model** for [moonshotai/Kimi-K2.7-Code](https://huggingface.co/moonshotai/Kimi-K2.7-Code), trained with [SpecForge](https://github.com/sgl-project/SpecForge) (PR #593) on NVIDIA Nemotron-Post-Training-Dataset-v2 (stem+chat+math+code). - 6-layer Qwen3-style draft (hidden 7168); consumes target hidden states at layers [1,12,24,35,47,58]; block_size 8. - Target vocab/tokenizer: Kimi-K2.7-Code (vocab 163840, mask_token_id 163838). - Checkpoint: epoch_4_step_334000 — **Work-in-progress snapshot (epoch_4_step_334000)** — training still running. Load with `trust_remote_code=True` (model code in `dflash.py`). Intended as the draft in SGLang DFlash speculative decoding paired with the Kimi-K2.7-Code target.

Note19 agohuggingface.co/cm00cm/Kimi-K2.7-Code-DFlash

Source context: 28 downloads · 1 likes · Pipeline feature-extraction · Library transformers · Repo cm00cm/Kimi-K2.7-Code-DFlash

Note18 julhuggingface.co/cm00cm/Kimi-K2.7-Code-DFlash

Source context: 2326 downloads · 1 likes · Pipeline feature-extraction · Library transformers · Repo cm00cm/Kimi-K2.7-Code-DFlash