AlekseiPravdin's picture
Update README.md
c5711a8 verified
|
raw
history blame
2.66 kB
metadata
license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - AlekseiPravdin/KSI-RP-NSK-128k-7B
  - grimjim/fireblossom-32K-7B
  - gguf
  - Q2_K
  - Q3_K_L
  - Q3_K_M
  - Q3_K_S
  - Q4_0
  - Q4_1
  - Q4_K_S
  - Q4_k_m
  - Q5_0
  - Q5_1
  - Q6_K
  - Q5_K_S
  - Q5_k_m
  - Q8_0
  - 128k
language:
  - en
  - ru
  - th

KSIF-RPG-128k-slerp ⭐️⭐️

KSIF-RPG-128k-slerp is a merge of the following models using mergekit:

🧩 Configuration

slices:
  - sources:
      - model: AlekseiPravdin/KSI-RP-NSK-128k-7B
        layer_range: [0, 32]
      - model: grimjim/fireblossom-32K-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: AlekseiPravdin/KSI-RP-NSK-128k-7B
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

Eval embedding benchmark (with 70 specific quesions):

inf.jpg md28g.jpg SK.jpg ks-inf.jpg command-r.jpg NSK.jpg NSMv2.jpg aura.jpg ivanDrogo.jpg KSI.jpg KSI-RPG.jpg llama3.jpg KSIF.jpg