Downloads · 30 days
0
jajmangold/Qwen3.5-0.8B-fni8
Qwen3.5-0.8B-fni8 is a image-text-to-text model from jajmangold. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
⚠️ EXPERIMENTAL — do not rely on this checkpoint. It was converted before the hybrid-config-drop converter fix: the hybrid (Gated-DeltaNet) portion of the config was dropped during conversion, so the file may not load…
Downloads · 30 days
0
Access
Public
Updated Jul 14, 2026
Repo size
1.4 GB
Likes
0
Public
Click a slice to open those files.
.fni81.4 GB · 100%
From the Hugging Face model README
⚠️ EXPERIMENTAL — do not rely on this checkpoint.
It was converted before the hybrid-config-drop converter fix: the hybrid (Gated-DeltaNet) portion of the config was dropped during conversion, so the file may not load or run correctly. A corrected reconversion is pending. Even once reconverted, full decode additionally requires the Track-2 DeltaNet int8 kernel (currently a fp16 linear-attention fallback).
Qwen3.5/3.6-family hybrid (Gated-DeltaNet) model, quantized from Qwen/Qwen3.5-0.8B. Repackaged to the .fni8 resident format (~1.4 GB) for the fni8 W8A8/W4A8 DP4A kernels on NVIDIA Volta (sm_70) — Tesla V100 / CMP 100-210.
See the experimental warning above. Not validated end-to-end.
__dp4a CUDA-core intrinsic. On the CMP 100-210 fleet (whose fp16 tensor cores are firmware-limited) dp4a is the fast path, not a compromise.fni8-serve is the LLM runtime (load_fni8_state_dict(<file>) into an LLMEngine; the architecture is read from the file). ComfyUI-fni8 is for diffusion DiTs only and does not load this model.
Part of the fni8 stack: kernels · LLM serving · ComfyUI DiTs.