Jim Lai

grimjim

AI & ML interests

Experimenting primarily with 7B-9B parameter text completion models. Not all models are intended for direct use, but for educational and/or merge purposes.

Organizations

Posts 5

Post

1344

Uploaded two basic SLERP merges of princeton-nlp/Llama-3-Instruct-8B-SimPO and UCLA-AGI/Llama-3-Instruct-8B-SPPO-Iter3, alternating the choice of base model, for people to test out and potentially use as merge fuel. (Personally, I am drawn to intelligent and attentive models, hence the experimentation.)

grimjim/Llama-3-Instruct-8B-SPPO-Iter3-SimPO-merge
grimjim/Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge

Post

2076

We explore extremely low-weight merger as an alternative to fine-tuning; e.g., weight 1e-4. Merge formula details here:
grimjim/kukulemon-v3-soul_mix-32k-7B

View all posts

Collections 4

models 73

Jim Lai

AI & ML interests

Organizations

Posts 5

Collections 4

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

Lewdiculous/kuno-kunoichi-v1-DPO-v2-SLERP-7B-GGUF-IQ-Imatrix

grimjim/kuno-kunoichi-v1-DPO-v2-SLERP-7B-8.0bpw_h8_exl2

grimjim/kukulemon-7B-GGUF

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

grimjim/kukulemon-7B

grimjim/kukulemon-spiked-9B

grimjim/kukulemon-32K-7B

models 73

grimjim/Llama-3-Instruct-8B-SPPO-Iter3-SimPO-merge

grimjim/Llama-3-Instruct-8B-SimPO-SPPO-Iter3-merge

grimjim/llama-3-Nephilim-v1-8B-6.5bpw_h6_exl2

grimjim/llama-3-Nephilim-v1-8B

grimjim/llama-3-Nephilim-v1-8B-GGUF

grimjim/llama-3-merge-pp-instruct-8B

grimjim/llama-3-merge-avalon-8B

grimjim/Llama-3-Luminurse-v0.1-OAS-8B

grimjim/Llama-3-Luminurse-v0.1-OAS-8B-GGUF-broken

grimjim/Llama-3-Steerpike-v1-OAS-8B

datasets 1

grimjim/adversarial-10-alpaca

Jim Lai

AI & ML interests

Organizations

Posts 5

Collections 4

models 73 Sort: Recently updated

datasets 1

models 73