Weyaxi/Einstein-v6.1-Llama3-8B
Full fine-tune vs meta-llama/Meta-Llama-3-8B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 24/65, MLP 96/96, Attention 128/128
6 derivatives checked
- mradermacher/Einstein-v6.1-Llama3-8B-GGUF: Same weights, converted
- PrunaAI/Einstein-v6.1-Llama3-8B-bnb-4bit-smashed: Quantization
- MaziyarPanahi/Einstein-v6.1-Llama3-8B-GGUF: Same weights, converted
- TitleOS/EinsteinBagel-8B: Full fine-tune
- Weyaxi/Einstein-v6.1-LLama3-8B-Instruct-Ties: Full fine-tune
- saucam/Skyro-4X8B: MoE built from dense models
1 other repos hold byte-for-byte identical weight files.