nafisabdkhan/DeepSeek-Coder-V2-Lite-Instruct-mlx-fp16

Same weights, converted vs deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct. Same weights stored in another dtype (BF16 → F16), plus 78 new tensors.

Tensors changed: Embeddings 0/1, Norms 0/82, lm_head 0/1, Attention 0/108, MLP 0/107

On the Hub · Downloads on Model Pulse