ExternalVendorCode/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2025-01-17 18:30:07 +00:00
Code Issues Actions 16 Packages Projects Releases Wiki Activity

models(gallery): add humanish-roleplay-llama-3.1-8b-i1 (#3126)

Browse Source 
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2024-08-05 10:11:00 +02:00 committed by GitHub
parent 6e1ec08f46
commit f15a93b19b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 17 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
gallery/index.yaml
View File

@ -307,6 +307,23 @@
- filename: L3.1-70b-glitz-v0.2.i1-Q4_K_M.gguf
sha256: 585efc83e7f6893043be2487fc09c914a381fb463ce97942ef2f25ae85103bcd
uri: huggingface://mradermacher/L3.1-70b-glitz-v0.2-i1-GGUF/L3.1-70b-glitz-v0.2.i1-Q4_K_M.gguf
- !!merge <<: *llama31
name: "humanish-roleplay-llama-3.1-8b-i1"
icon: https://cdn-uploads.huggingface.co/production/uploads/5fad8602b8423e1d80b8a965/VPwtjS3BtjEEEq7ck4kAQ.webp
urls:
- https://huggingface.co/mradermacher/Humanish-Roleplay-Llama-3.1-8B-i1-GGUF
description: |
A DPO-tuned Llama-3.1 to behave more "humanish", i.e., avoiding all the AI assistant slop. It also works for role-play (RP). To achieve this, the model was fine-tuned over a series of datasets:
General conversations from Claude Opus, from Undi95/Meta-Llama-3.1-8B-Claude
Undi95/Weyaxi-humanish-dpo-project-noemoji, to make the model react as a human, rejecting assistant-like or too neutral responses.
ResplendentAI/NSFW_RP_Format_DPO, to steer the model towards using the *action* format in RP settings. Works best if in the first message you also use this format naturally (see example)
overrides:
parameters:
model: Humanish-Roleplay-Llama-3.1-8B.i1-Q4_K_M.gguf
files:
- filename: Humanish-Roleplay-Llama-3.1-8B.i1-Q4_K_M.gguf
sha256: 18cf753684e5226b51f3defc708852ca4924f50dc8bc31c9a7d0a036a477b7a7
uri: huggingface://mradermacher/Humanish-Roleplay-Llama-3.1-8B-i1-GGUF/Humanish-Roleplay-Llama-3.1-8B.i1-Q4_K_M.gguf
## Uncensored models
- !!merge <<: *llama31
name: "darkidol-llama-3.1-8b-instruct-1.0-uncensored-i1"