models(gallery): add arliai-llama-3-8b-formax-v1.0 (#2783)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2024-07-12 12:22:17 +02:00 committed by GitHub
parent 96127e9967
commit dcccfc2cce
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -1768,6 +1768,29 @@
- filename: Replete-Coder-Instruct-8b-Merged-Q4_K_M.gguf
sha256: 5374a38023b3d8617d266f94e4eff4c5d996b3197e6c42ae27315110bcc75d33
uri: huggingface://bartowski/Replete-Coder-Instruct-8b-Merged-GGUF/Replete-Coder-Instruct-8b-Merged-Q4_K_M.gguf
- !!merge <<: *llama3
name: "arliai-llama-3-8b-formax-v1.0"
description: |
Formax is a model that specializes in following response format instructions. Tell it the format of it's response and it will follow it perfectly. Great for data processing and dataset creation tasks.
Base model: https://huggingface.co/failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
Training:
4096 sequence length
Training duration is around 2 days on 2x3090Ti
1 epoch training with a massive dataset for minimized repetition sickness.
LORA with 64-rank 128-alpha resulting in ~2% trainable weights.
urls:
- https://huggingface.co/OwenArli/ArliAI-Llama-3-8B-Formax-v1.0
- https://huggingface.co/bartowski/ArliAI-Llama-3-8B-Formax-v1.0-GGUF
overrides:
context_size: 4096
parameters:
model: ArliAI-Llama-3-8B-Formax-v1.0-Q4_K_M.gguf
files:
- filename: ArliAI-Llama-3-8B-Formax-v1.0-Q4_K_M.gguf
sha256: e6a47a11eb67c1d4cd92e3512d3288a5d937c41a3319e95c3b8b2332428af239
uri: huggingface://bartowski/ArliAI-Llama-3-8B-Formax-v1.0-GGUF/ArliAI-Llama-3-8B-Formax-v1.0-Q4_K_M.gguf
- name: "llama-3-sec-chat"
url: "github:mudler/LocalAI/gallery/chatml.yaml@master"
urls: