Jim Lai

grimjim

regisss's profile picture

Retreatcost's profile picture

rocca's profile picture

AI & ML interests

Experimenting primarily with 7B-12B parameter text completion models. Not all models are intended for direct end use, but aim for research and/or educational purposes. Recent Contributions: stabilized refusal direction ablation via Gram-Schmidt orthonormalization and norm-preserving interventions; confirmed reasoning transfer via model merger.

Recent Activity

new activity 4 days ago

grimjim/gemma-3-12b-it-biprojected-abliterated:What's the difference between projected and biprojected?

new activity 4 days ago

grimjim/gemma-3-12b-it-MPOAdd-v1:Idea

updated a model 8 days ago

grimjim/gemma-3-12b-it-MPOAdd-v1

View all activity

Organizations

grimjim 's collections 5

Highlighted work

My "greatest hits", sort of

grimjim/gemma-3-12b-it-norm-preserved-biprojected-abliterated

Image-Text-to-Text • 12B • Updated 17 days ago • 3.75k • 18
grimjim/gemma-3-12b-it-projection-abliterated

Image-Text-to-Text • 12B • Updated Oct 30 • 35 • 9
grimjim/gemma-3-12b-it-biprojected-abliterated

Image-Text-to-Text • 12B • Updated Nov 1 • 20 • 6
grimjim/DeepSauerHuatuoSkywork-R1-o1-Llama-3.1-8B

Text Generation • 8B • Updated Feb 7 • 12 • 6

Quantized models

Select models helpfully quantized by others as well as myself

grimjim/meta-llama-Llama-3.2-1B-Instruct-exl2

Text Generation • Updated Nov 29, 2024 • 14
grimjim/kuno-kunoichi-v1-DPO-v2-SLERP-7B-GGUF

Text Generation • 7B • Updated Aug 26, 2024 • 51 • 2
Lewdiculous/kuno-kunoichi-v1-DPO-v2-SLERP-7B-GGUF-IQ-Imatrix

7B • Updated Mar 14, 2024 • 193 • 10
grimjim/kuno-kunoichi-v1-DPO-v2-SLERP-7B-8.0bpw_h8_exl2

Text Generation • Updated Aug 26, 2024 • 9

Experimental and negative results

Models that didn't always quite work out, but may still be of interest.

grimjim/franken-kunoichi-IDUS-11B

Text Generation • 11B • Updated Apr 15, 2024 • 12 • 1
grimjim/kunoichi-squared-model_stock-7B

Text Generation • 7B • Updated Apr 15, 2024 • 12 • 1
grimjim/wizard-elem-to-32k-7B

Text Generation • 7B • Updated May 27, 2024 • 17
grimjim/Llama-3-Perky-Pat-Instruct-8B

Text Generation • 8B • Updated Jul 4, 2024 • 11 • 4

Full weight models

Select models uploaded in safetensors format. Currently all are merges. Annotations here.

grimjim/kuno-kunoichi-v1-DPO-v2-SLERP-7B

Text Generation • 7B • Updated May 10, 2024 • 74 • 4
grimjim/kukulemon-7B

Text Generation • 7B • Updated Mar 21, 2024 • 81 • 11
grimjim/kukulemon-spiked-9B

Text Generation • 9B • Updated Apr 27, 2024 • 17 • 3
grimjim/kukulemon-32K-7B

Text Generation • 7B • Updated Apr 27, 2024 • 13 • 1

Mirrored mergekit-ready models

Mirrored models tweaked to be more friendly for mergekit. No pickles allowed.

grimjim/llama-3-nvidia-ChatQA-1.5-8B

Text Generation • 8B • Updated May 2, 2024 • 22 • 3
grimjim/mistralai-Mistral-Nemo-Instruct-2407

Text Generation • 12B • Updated Nov 17 • 79
grimjim/mistralai-Mistral-Nemo-Base-2407

Text Generation • 12B • Updated Nov 16 • 51
grimjim/llama-3-aaditya-OpenBioLLM-8B

Text Generation • 8B • Updated May 29, 2024 • 28