fraserprice/DeepSeek-V4-Flash-DSpark

Partial fine-tune vs deepseek-ai/DeepSeek-V4-Flash. Partial change: 1% of the weights changed (attention, mlp, norms), in 0 of 43 layers.

Tensors changed: Embeddings 0/1, MLP 27/1759, Attention 14/865, Norms 4/239, other 0/4

On the Hub · Downloads on Model Pulse