v000000's picture
Create README.md
bf2faaf verified
|
raw
history blame
1.15 kB
metadata
tags:
  - llama
  - merge

in progress

The full name is:

Llama-3-15B-EtherealMaid-t0.0001-Inverted

A merge of the following models using a custom NearSwap(t0.0001) algorithm (inverted):

With ZeusLabs/L3-Aethora-15B-V2 as the base model.

#Fixed
def lerp(a, b, t):
    return a * (1 - t) + b * t

def nearswap(v0, v1, t):
    lweight = np.abs(v0 - v1)
    with np.errstate(divide='ignore', invalid='ignore'):
        lweight = np.where(lweight != 0, t / lweight, 1.0)
    lweight = np.nan_to_num(lweight, nan=1.0, posinf=1.0, neginf=1.0)
    np.clip(lweight, a_min=0.0, a_max=1.0, out=lweight)
    return lerp(v0, v1, lweight)

Credit Alchemonaut

Prompt Template

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{output}<|eot_id|>