Hugging Face's logo

CultriX
/
Qwen2.5-14B-Brocav7

Text Generation
Transformers
Safetensors
qwen2
mergekit
Merge
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
Qwen2.5-14B-Brocav7
File size: 2,408 Bytes 
06acee7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
 
merge_method: della_linear # Consider if the 'della_linear' merge method is optimal based on the latest benchmarks. Would a different method better balance task priorities?
base_model: CultriX/Qwen2.5-14B-Wernickev3
dtype: bfloat16
parameters:
  epsilon: 0.01  # Epsilon is set for fine-grain precision. Consider validating this epsilon value against current benchmark results to ensure it effectively fine-tunes model behaviors and aligns with the intended benchmarks.
  lambda: 1.5    # Optimized for significant model contributions. Double-check if this value prioritizes the right model strengths effectively.
  normalize: true
  smoothing_factor: 0.08 # Double-check if this smoothing factor sufficiently supports task-specific performance across benchmarks, especially with diverse model contributions. Balanced blending to preserve model diversity.

gradient_clipping:
  CultriX/Qwen2.5-14B-Wernickev3: 0.85
  CultriX/Qwenfinity-2.5-14B: 0.82
  djuna/Q2.5-Veltha-14B-0.5: 0.92
  CultriX/Qwen2.5-14B-Broca: 0.86
  qingy2019/Qwen2.5-Math-14B-Instruct: 0.94
  CultriX/SeQwence-14Bv1: 0.87
  sometimesanotion/Qwen2.5-14B-Vimarckoso: 0.90
  allknowingroger/QwenSlerp6-14B: 0.86

models:
  - model: CultriX/Qwen2.5-14B-Wernickev3
    parameters:
      weight: 0.25
      density: 0.72
  - model: CultriX/Qwenfinity-2.5-14B
    parameters:
      weight: 0.22
      density: 0.68
  - model: djuna/Q2.5-Veltha-14B-0.5
    parameters:
      weight: 0.20
      density: 0.75
  - model: CultriX/Qwen2.5-14B-Broca
    parameters:
      weight: 0.16
      density: 0.68
  - model: qingy2019/Qwen2.5-Math-14B-Instruct
    parameters:
      weight: 0.19
      density: 0.75
  - model: CultriX/SeQwence-14Bv1
    parameters:
      weight: 0.13
      density: 0.65
  - model: sometimesanotion/Qwen2.5-14B-Vimarckoso
    parameters:
      weight: 0.11
      density: 0.62
  - model: allknowingroger/QwenSlerp6-14B
    parameters:
      weight: 0.09
      density: 0.65

adaptive_merge_parameters:
  task_weights:
    tinyArc: 1.65
    tinyHellaswag: 1.55
    tinyMMLU: 1.7
    tinyTruthfulQA: 1.95
    tinyTruthfulQA_mc1: 1.75
    tinyWinogrande: 1.8
    IFEval: 2.0
    BBH: 1.75
    MATH: 2.2
    GPQA: 1.85
    MUSR: 1.95
    MMLU-PRO: 1.85  # Consider reviewing task_weights to confirm alignment with latest performance data and benchmarks for optimal configuration.

tokenizer_source: CultriX/Qwen2.5-14B-Wernickev3