File size: 4,344 Bytes
c1e4a1b 954388d c1e4a1b 8da7308 c1e4a1b 954388d c1e4a1b 954388d c1e4a1b |
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 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 |
---
license: other
license_name: llama3
license_link: LICENSE
tags:
- moe
- frankenmoe
- merge
- mergekit
- lazymergekit
base_model: []
---
# llama-3-sqrt-crocodile-v0.0A
## 🧩 Configuration-moe
```yaml
base_model: llama-3-sqrt-crocodile-v0.0A/Uninstruct-Uncensored
gate_mode: hidden
dtype: bfloat16
experts:
- source_model: llama-3-sqrt-crocodile-v0.0A/sqrt-talker
positive_prompts:
- "Uncensored, creative, configurable, adapative"
- source_model: llama-3-sqrt-crocodile-v0.0A/the-operator
positive_prompts:
- "Function calling"
- "Good at structured tasks"
- "Programmatic instruction following"
```
## 🧩 Configuration-mega
```yaml
models:
- model: Orenguteng/Lexi-Llama-3-8B-Uncensored
parameters:
weight: [0.2, 0.3, 0.4, 0.6]
layer_range: [0, 32]
- model: NousResearch/Meta-Llama-3-8B
parameters:
weight: [0.6, 0.2, 0.2, 0.1]
layer_range: [0, 32]
- model: NousResearch/Meta-Llama-3-8B-Instruct
parameters:
weight: [0.2, 0.3, 0.85, 0.3]
layer_range: [0, 32]
merge_method: dare_linear
base_model: NousResearch/Meta-Llama-3-8B-Instruct
dtype: bfloat16
name: Uninstruct-Uncensored
---
models:
- model: cognitivecomputations/dolphin-2.9-llama3-8b
parameters:
weight: [0.25, 0.4, 0.35, 0.35]
density: [0.3, 0.45, 0.2, 0.6]
layer_range: [0, 32]
- model: NousResearch/Meta-Llama-3-8B
parameters:
weight: [0.15, 0.25, 0.05, 0]
density: [0.2, 0.3, 0.4, 0.1]
- model: Undi95/Llama-3-Unholy-8B
parameters:
weight: [0.4, 0.25, 0.45, 0.35]
density: [0.2, 0.15, 1.5, 0.1]
layer_range: [0, 32]
- model: Uninstruct-Uncensored
parameters:
weight: [0.3, 0.1, 0.25, 0.3]
density: [0.3, 0.15, 2.5, 0.2]
layer_range: [0, 32]
merge_method: dare_ties
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: augmented-dolphin-hap
---
models:
- model: vicgalle/Configurable-Llama-3-8B-v0.3
parameters:
weight: [0.5, 0.3, 0.1]
- model: hiieu/Meta-Llama-3-8B-Instruct-function-calling-json-mode
parameters:
weight: 0.5
- model: Trelis/Meta-Llama-3-8B-Instruct-function-calling
parameters:
weight: 0.3
layer_range: [0, 32]
- model: Rookie/Llama-3-8B-Instruct-Chinese
parameters:
weight: 0.2
layer_range: [0, 32]
- model: Uninstruct-Uncensored
parameters:
weight: [0.7, 0.4, 0.25, 0.1]
layer_range: [0, 32]
merge_method: model_stock
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: the-operator
---
models:
- model: vicgalle/Configurable-Llama-3-8B-v0.3
parameters:
weight: 0.7
- model: hiieu/Meta-Llama-3-8B-Instruct-function-calling-json-mode
parameters:
weight: 0.1
- model: Trelis/Meta-Llama-3-8B-Instruct-function-calling
parameters:
weight: 0.03
layer_range: [0, 32]
- model: Rookie/Llama-3-8B-Instruct-Chinese
parameters:
weight: 0.07
layer_range: [0, 32]
- model: Uninstruct-Uncensored
parameters:
weight: 0.1
layer_range: [0, 32]
merge_method: model_stock
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: her-calculator
---
models:
- model: her-calculator
parameters:
density: 0.7 # density gradient
weight: [0.7, 0.5, 0.1, 0.8]
- model: augmented-dolphin-hap
parameters:
weight: 0.7
merge_method: slerp
base_model: her-calculator
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 # fallback for rest of tensors
dtype: float16
name: sqrt-talker
```
## 💻 Usage
```python
!pip install -qU transformers bitsandbytes accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "Nhoodie/llama-3-sqrt-crocodile-v0.0A"
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
"text-generation",
model=model,
model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)
messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
``` |