UnstableLlama commited on
Commit
9e064fd
·
verified ·
1 Parent(s): 8a21934

Unify Love-Aligned cards, label SC quants, and link EXL3 variants

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -23,7 +23,7 @@ tags:
23
  <p style="margin:12px 0; line-height:1.7;">A Gemma 4 31B finetune built on <a href="https://huggingface.co/ReadyArt/gemma-4-31B-it-scotoma-2" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">ReadyArt / gemma-4-31B-it-scotoma-2</a>, with an additional round of preference training. This release uses the <b>v4, epoch 2</b> checkpoint.</p>
24
  <p style="margin:12px 0; line-height:1.7;">This is my attempt to explore the direction Jakub Pachocki describes in <a href="https://openai.com/index/an-alien-mind/" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;"><i>An Alien Mind</i></a>, particularly the section “Teaching machines to love.” He frames value alignment around principles that generalize to unfamiliar situations, including honesty, integrity, and love for humanity. Love-Aligned takes that idea as a starting point for a small, independent finetuning experiment: explicitly training for love of the user and humanity, then examining how that orientation appears in conversation.</p>
25
  <div style="border-top:1px solid #543044; margin-top:24px; padding-top:18px; font-family:monospace; font-size:13px; line-height:1.9;">
26
- <a href="https://huggingface.co/UnstableLlama/Love-Aligned-31B-exl3-6.00bpw" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">EXL3 · 6.00 bpw</a> &nbsp; / &nbsp; <a href="https://huggingface.co/UnstableLlama/Love-Aligned-31B-LoRA" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">LoRA adapter</a> &nbsp; / &nbsp; <a href="https://huggingface.co/mradermacher/Love-Aligned-31B-i1-GGUF" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">GGUF imatrix</a> &nbsp; / &nbsp; <a href="https://huggingface.co/mradermacher/Love-Aligned-31B-GGUF" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">GGUF</a>
27
  </div>
28
  </div>
29
 
 
23
  <p style="margin:12px 0; line-height:1.7;">A Gemma 4 31B finetune built on <a href="https://huggingface.co/ReadyArt/gemma-4-31B-it-scotoma-2" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">ReadyArt / gemma-4-31B-it-scotoma-2</a>, with an additional round of preference training. This release uses the <b>v4, epoch 2</b> checkpoint.</p>
24
  <p style="margin:12px 0; line-height:1.7;">This is my attempt to explore the direction Jakub Pachocki describes in <a href="https://openai.com/index/an-alien-mind/" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;"><i>An Alien Mind</i></a>, particularly the section “Teaching machines to love.” He frames value alignment around principles that generalize to unfamiliar situations, including honesty, integrity, and love for humanity. Love-Aligned takes that idea as a starting point for a small, independent finetuning experiment: explicitly training for love of the user and humanity, then examining how that orientation appears in conversation.</p>
25
  <div style="border-top:1px solid #543044; margin-top:24px; padding-top:18px; font-family:monospace; font-size:13px; line-height:1.9;">
26
+ <a href="https://huggingface.co/models?other=base_model:quantized:UnstableLlama/Love-Aligned-31B" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">EXL3s</a> &nbsp; / &nbsp; <a href="https://huggingface.co/UnstableLlama/Love-Aligned-31B-LoRA" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">LoRA adapter</a> &nbsp; / &nbsp; <a href="https://huggingface.co/mradermacher/Love-Aligned-31B-i1-GGUF" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">GGUF imatrix</a> &nbsp; / &nbsp; <a href="https://huggingface.co/mradermacher/Love-Aligned-31B-GGUF" style="color:#f2a0c4; text-decoration:underline; text-underline-offset:3px;">GGUF</a>
27
  </div>
28
  </div>
29