Downloads · 30 days
46
31% of all-time downloads
deepseek-v3-pruned-8experts-mtp
ckoh04/deepseek-v3-pruned-8experts-mtp
deepseek-v3-pruned-8experts-mtp is a text generation model from ckoh04. Use it when you need the model to write or continue text. It is set up for transformers.
Yeahh... so I pruned this like an idiot before. I totally removed the MTP module.
All-time downloads
149
Public
Parameters
40B
43.8 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors43.8 GB · 100%
Parameter types
How the weights are stored.
F8_E4M336.2B · 90%
Try a prompt
Base models
Model card
From the Hugging Face model README
DeepSeek V3 Pruned 8 Experts With MTP
Yeahh... so I pruned this like an idiot before. I totally removed the MTP module.
We now have num_nextn_predict_layers = 1 and the appended model.layers.61.* MTP module, including enorm, hnorm, eh_proj, embed_tokens, and shared_head tensors.
The routed expert banks have been pruned to experts 0..7, and layer 61 uses the same 8 routed experts as the main MoE layers.
- Task
- Text Generation
- Library
- transformers
- Type
- deepseek_v3
- Created
- Jul 23, 2026
- Updated
- Jul 23, 2026