Update README.md
Browse files
README.md
CHANGED
@@ -33,10 +33,6 @@ Shows similar emergent language nuance abilities compared to 8B.
|
|
33 |
|
34 |
Unaligned and somewhat lazy. Use rep_pen 1.1
|
35 |
|
36 |
-
# Quants
|
37 |
-
* [Q8_0 static, imatrix](https://huggingface.co/v000000/L3-11.5B-DuS-MoonRoot-Q8_0-GGUF)
|
38 |
-
* [Q6_K static, imatrix](https://huggingface.co/v000000/L3-11.5B-DuS-MoonRoot-Q6_K-GGUF)
|
39 |
-
|
40 |
# merge
|
41 |
|
42 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
@@ -56,6 +52,7 @@ The following models were included in the merge:
|
|
56 |
|
57 |
The following YAML configuration was used to produce this model:
|
58 |
|
|
|
59 |
```yaml
|
60 |
slices:
|
61 |
- sources:
|
@@ -66,6 +63,62 @@ slices:
|
|
66 |
layer_range: [8, 32]
|
67 |
merge_method: passthrough
|
68 |
dtype: bfloat16
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
```
|
70 |
|
71 |
---
|
|
|
33 |
|
34 |
Unaligned and somewhat lazy. Use rep_pen 1.1
|
35 |
|
|
|
|
|
|
|
|
|
36 |
# merge
|
37 |
|
38 |
This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).
|
|
|
52 |
|
53 |
The following YAML configuration was used to produce this model:
|
54 |
|
55 |
+
---Step 3
|
56 |
```yaml
|
57 |
slices:
|
58 |
- sources:
|
|
|
63 |
layer_range: [8, 32]
|
64 |
merge_method: passthrough
|
65 |
dtype: bfloat16
|
66 |
+
|
67 |
+
```
|
68 |
+
---Step 2
|
69 |
+
```yaml
|
70 |
+
slices:
|
71 |
+
- sources:
|
72 |
+
- model: v000000/L3-8B-Poppy-Sunspice-experiment-c+Blackroot/Llama-3-8B-Abomination-LORA
|
73 |
+
layer_range: [0, 32]
|
74 |
+
- model: v000000/L3-8B-Poppy-Sunspice-experiment-c+ResplendentAI/BlueMoon_Llama3
|
75 |
+
layer_range: [0, 32]
|
76 |
+
merge_method: slerp
|
77 |
+
base_model: v000000/L3-8B-Poppy-Sunspice-experiment-c+Blackroot/Llama-3-8B-Abomination-LORA
|
78 |
+
parameters:
|
79 |
+
t:
|
80 |
+
- filter: self_attn
|
81 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
82 |
+
- filter: mlp
|
83 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
84 |
+
- value: 0.5
|
85 |
+
dtype: bfloat16
|
86 |
+
random_seed: 0
|
87 |
+
|
88 |
+
```
|
89 |
+
---Step 1
|
90 |
+
```yaml
|
91 |
+
models:
|
92 |
+
- model: crestf411/L3-8B-sunfall-abliterated-v0.2
|
93 |
+
parameters:
|
94 |
+
weight: 0.1
|
95 |
+
density: 0.18
|
96 |
+
- model: Hastagaras/HALU-8B-LLAMA3-BRSLURP
|
97 |
+
parameters:
|
98 |
+
weight: 0.1
|
99 |
+
density: 0.3
|
100 |
+
- model: Nitral-Archive/Poppy_Porpoise-Biomix
|
101 |
+
parameters:
|
102 |
+
weight: 0.1
|
103 |
+
density: 0.42
|
104 |
+
- model: cgato/L3-TheSpice-8b-v0.8.3
|
105 |
+
parameters:
|
106 |
+
weight: 0.2
|
107 |
+
density: 0.54
|
108 |
+
- model: Sao10K/L3-8B-Stheno-v3.2
|
109 |
+
parameters:
|
110 |
+
weight: 0.2
|
111 |
+
density: 0.66
|
112 |
+
- model: Nitral-AI/Poppy_Porpoise-0.72-L3-8B
|
113 |
+
parameters:
|
114 |
+
weight: 0.3
|
115 |
+
density: 0.78
|
116 |
+
merge_method: dare_ties
|
117 |
+
base_model: NousResearch/Meta-Llama-3-8B-Instruct
|
118 |
+
parameters:
|
119 |
+
int8_mask: true
|
120 |
+
dtype: bfloat16
|
121 |
+
|
122 |
```
|
123 |
|
124 |
---
|