rohithporandla417/llama-3.2-3b-Programming

Same weights, converted vs unsloth/Llama-3.2-3B-Instruct-bnb-4bit. Same weights stored in another dtype (U8/BF16 → U8/F16).

Tensors changed: MLP 0/420, Attention 0/560, Embeddings 0/1, Norms 0/57

12 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse