Update README.md
Browse files
README.md
CHANGED
@@ -3,11 +3,11 @@ tags:
|
|
3 |
- merge
|
4 |
- mergekit
|
5 |
- lazymergekit
|
6 |
-
- Trappu/Nemo-Picaro-fixed
|
7 |
- anthracite-org/magnum-v2-12b
|
|
|
8 |
base_model:
|
9 |
-
- Trappu/Nemo-Picaro-fixed
|
10 |
- anthracite-org/magnum-v2-12b
|
|
|
11 |
model-index:
|
12 |
- name: Magnum-Picaro-0.7-v2-12b
|
13 |
results:
|
@@ -24,7 +24,8 @@ model-index:
|
|
24 |
value: 30.03
|
25 |
name: strict accuracy
|
26 |
source:
|
27 |
-
url:
|
|
|
28 |
name: Open LLM Leaderboard
|
29 |
- task:
|
30 |
type: text-generation
|
@@ -39,7 +40,8 @@ model-index:
|
|
39 |
value: 35.75
|
40 |
name: normalized accuracy
|
41 |
source:
|
42 |
-
url:
|
|
|
43 |
name: Open LLM Leaderboard
|
44 |
- task:
|
45 |
type: text-generation
|
@@ -54,7 +56,8 @@ model-index:
|
|
54 |
value: 4.76
|
55 |
name: exact match
|
56 |
source:
|
57 |
-
url:
|
|
|
58 |
name: Open LLM Leaderboard
|
59 |
- task:
|
60 |
type: text-generation
|
@@ -69,7 +72,8 @@ model-index:
|
|
69 |
value: 9.73
|
70 |
name: acc_norm
|
71 |
source:
|
72 |
-
url:
|
|
|
73 |
name: Open LLM Leaderboard
|
74 |
- task:
|
75 |
type: text-generation
|
@@ -84,7 +88,8 @@ model-index:
|
|
84 |
value: 19.56
|
85 |
name: acc_norm
|
86 |
source:
|
87 |
-
url:
|
|
|
88 |
name: Open LLM Leaderboard
|
89 |
- task:
|
90 |
type: text-generation
|
@@ -101,7 +106,8 @@ model-index:
|
|
101 |
value: 28.67
|
102 |
name: accuracy
|
103 |
source:
|
104 |
-
url:
|
|
|
105 |
name: Open LLM Leaderboard
|
106 |
license: apache-2.0
|
107 |
pipeline_tag: text-generation
|
@@ -110,7 +116,7 @@ library_name: transformers
|
|
110 |
|
111 |
# Details
|
112 |
|
113 |
-
This model is a merge between [Trappu/Nemo-Picaro-
|
114 |
|
115 |
The reason why I decided to merge it with Magnum (and don't recommend Picaro alone) is because that model, aside from its obvious flaws (rampant impersonation, stupid, etc...), is a one-trick pony and will be really rough for the average LLM user to handle. The idea was to have Magnum work as some sort of stabilizer to fix the issues that emerge from the lack of multiturn/smart data in Picaro's dataset. It worked, I think. I enjoy the outputs and it's smart enough to work with.
|
116 |
|
@@ -159,14 +165,14 @@ Static: https://huggingface.co/mradermacher/Magnum-Picaro-0.7-v2-12b-GGUF
|
|
159 |
# Magnum-Picaro-0.7-v2-12b
|
160 |
|
161 |
Magnum-Picaro-0.7-v2-12b is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
|
162 |
-
* [Trappu/Nemo-Picaro-
|
163 |
* [anthracite-org/magnum-v2-12b](https://huggingface.co/anthracite-org/magnum-v2-12b)
|
164 |
|
165 |
## 🧩 Configuration
|
166 |
|
167 |
```yaml
|
168 |
models:
|
169 |
-
- model: Trappu/Nemo-Picaro-
|
170 |
parameters:
|
171 |
density: 0.7
|
172 |
weight: 0.5
|
@@ -176,7 +182,7 @@ models:
|
|
176 |
weight: 0.5
|
177 |
|
178 |
merge_method: ties
|
179 |
-
base_model: Trappu/Nemo-Picaro-
|
180 |
parameters:
|
181 |
normalize: true
|
182 |
int8_mask: true
|
@@ -219,5 +225,4 @@ Detailed results can be found [here](https://huggingface.co/datasets/open-llm-le
|
|
219 |
|MATH Lvl 5 (4-Shot)| 4.76|
|
220 |
|GPQA (0-shot) | 9.73|
|
221 |
|MuSR (0-shot) |19.56|
|
222 |
-
|MMLU-PRO (5-shot) |28.67|
|
223 |
-
|
|
|
3 |
- merge
|
4 |
- mergekit
|
5 |
- lazymergekit
|
|
|
6 |
- anthracite-org/magnum-v2-12b
|
7 |
+
- Trappu/Nemo-Picaro-12B
|
8 |
base_model:
|
|
|
9 |
- anthracite-org/magnum-v2-12b
|
10 |
+
- Trappu/Nemo-Picaro-12B
|
11 |
model-index:
|
12 |
- name: Magnum-Picaro-0.7-v2-12b
|
13 |
results:
|
|
|
24 |
value: 30.03
|
25 |
name: strict accuracy
|
26 |
source:
|
27 |
+
url: >-
|
28 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
29 |
name: Open LLM Leaderboard
|
30 |
- task:
|
31 |
type: text-generation
|
|
|
40 |
value: 35.75
|
41 |
name: normalized accuracy
|
42 |
source:
|
43 |
+
url: >-
|
44 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
45 |
name: Open LLM Leaderboard
|
46 |
- task:
|
47 |
type: text-generation
|
|
|
56 |
value: 4.76
|
57 |
name: exact match
|
58 |
source:
|
59 |
+
url: >-
|
60 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
61 |
name: Open LLM Leaderboard
|
62 |
- task:
|
63 |
type: text-generation
|
|
|
72 |
value: 9.73
|
73 |
name: acc_norm
|
74 |
source:
|
75 |
+
url: >-
|
76 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
77 |
name: Open LLM Leaderboard
|
78 |
- task:
|
79 |
type: text-generation
|
|
|
88 |
value: 19.56
|
89 |
name: acc_norm
|
90 |
source:
|
91 |
+
url: >-
|
92 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
93 |
name: Open LLM Leaderboard
|
94 |
- task:
|
95 |
type: text-generation
|
|
|
106 |
value: 28.67
|
107 |
name: accuracy
|
108 |
source:
|
109 |
+
url: >-
|
110 |
+
https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=Trappu/Magnum-Picaro-0.7-v2-12b
|
111 |
name: Open LLM Leaderboard
|
112 |
license: apache-2.0
|
113 |
pipeline_tag: text-generation
|
|
|
116 |
|
117 |
# Details
|
118 |
|
119 |
+
This model is a merge between [Trappu/Nemo-Picaro-12B](https://huggingface.co/Trappu/Nemo-Picaro-12B), a model trained on my own little dataset free of synthetic data, which focuses solely on storywriting and scenrio prompting (Example: `[ Scenario: bla bla bla; Tags: bla bla bla ]`), and [anthracite-org/magnum-v2-12b](https://huggingface.co/anthracite-org/magnum-v2-12b).
|
120 |
|
121 |
The reason why I decided to merge it with Magnum (and don't recommend Picaro alone) is because that model, aside from its obvious flaws (rampant impersonation, stupid, etc...), is a one-trick pony and will be really rough for the average LLM user to handle. The idea was to have Magnum work as some sort of stabilizer to fix the issues that emerge from the lack of multiturn/smart data in Picaro's dataset. It worked, I think. I enjoy the outputs and it's smart enough to work with.
|
122 |
|
|
|
165 |
# Magnum-Picaro-0.7-v2-12b
|
166 |
|
167 |
Magnum-Picaro-0.7-v2-12b is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
|
168 |
+
* [Trappu/Nemo-Picaro-12B](https://huggingface.co/Trappu/Nemo-Picaro-12B)
|
169 |
* [anthracite-org/magnum-v2-12b](https://huggingface.co/anthracite-org/magnum-v2-12b)
|
170 |
|
171 |
## 🧩 Configuration
|
172 |
|
173 |
```yaml
|
174 |
models:
|
175 |
+
- model: Trappu/Nemo-Picaro-12B
|
176 |
parameters:
|
177 |
density: 0.7
|
178 |
weight: 0.5
|
|
|
182 |
weight: 0.5
|
183 |
|
184 |
merge_method: ties
|
185 |
+
base_model: Trappu/Nemo-Picaro-12B
|
186 |
parameters:
|
187 |
normalize: true
|
188 |
int8_mask: true
|
|
|
225 |
|MATH Lvl 5 (4-Shot)| 4.76|
|
226 |
|GPQA (0-shot) | 9.73|
|
227 |
|MuSR (0-shot) |19.56|
|
228 |
+
|MMLU-PRO (5-shot) |28.67|
|
|