allenai/Llama-3.1-Tulu-3-8B-SFT
Full fine-tune vs meta-llama/Llama-3.1-8B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: Norms 65/65, MLP 96/96, Attention 128/128
12 derivatives checked
- allenai/Llama-3.1-Tulu-3-8B-DPO: Full fine-tune
- bartowski/Llama-3.1-Tulu-3-8B-SFT-GGUF: Same weights, converted
- mradermacher/Llama-3.1-Tulu-3-8B-SFT-i1-GGUF: Quantization
- QuantFactory/Llama-3.1-Tulu-3-8B-DPO-GGUF: Full fine-tune
- allenai/Llama-3.1-Tulu-3-8B-RM: Full fine-tune
- mradermacher/Llama-3.1-Tulu-3-8B-SFT-GGUF: Same weights, converted
- DevQuasar-3/allenai.Llama-3.1-Tulu-3-8B-SFT-GGUF: Quantization
- THU-KEG/TULU3-VerIF: Full fine-tune
- PardisSzah/FaithDPO-v1: Partial fine-tune
- PardisSzah/FaithDPO-v3: Partial fine-tune
- PardisSzah/FaithDPO-v2: Partial fine-tune
- winglian/tulu3-dpo-8b-qat-v1-fixed: Full fine-tune
2 other repos hold byte-for-byte identical weight files.