TheBloke/30B-Epsilon-GPTQ

Full fine-tune vs CalderaAI/30B-Epsilon. Fine-tune: 95% of the weights changed, across 60 of 60 layers.

Tensors changed: MLP 540/720, Attention 720/1020, lm_head 0/1, Embeddings 0/1, Norms 0/121

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

On the Hub · Downloads on Model Pulse