bartowski commited on
Commit
495577f
1 Parent(s): 3520272

measurement.json

Browse files
Files changed (2) hide show
  1. README.md +174 -0
  2. measurement.json +0 -0
README.md ADDED
@@ -0,0 +1,174 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model:
4
+ - cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser
5
+ - Locutusque/Hyperion-1.5-Mistral-7B
6
+ - ibm/merlinite-7b
7
+ library_name: transformers
8
+ tags:
9
+ - mergekit
10
+ - merge
11
+ - code
12
+ model-index:
13
+ - name: Magic-Dolphin-7b
14
+ results:
15
+ - task:
16
+ type: text-generation
17
+ name: Text Generation
18
+ dataset:
19
+ name: AI2 Reasoning Challenge (25-Shot)
20
+ type: ai2_arc
21
+ config: ARC-Challenge
22
+ split: test
23
+ args:
24
+ num_few_shot: 25
25
+ metrics:
26
+ - type: acc_norm
27
+ value: 65.78
28
+ name: normalized accuracy
29
+ source:
30
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
31
+ name: Open LLM Leaderboard
32
+ - task:
33
+ type: text-generation
34
+ name: Text Generation
35
+ dataset:
36
+ name: HellaSwag (10-Shot)
37
+ type: hellaswag
38
+ split: validation
39
+ args:
40
+ num_few_shot: 10
41
+ metrics:
42
+ - type: acc_norm
43
+ value: 85.61
44
+ name: normalized accuracy
45
+ source:
46
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
47
+ name: Open LLM Leaderboard
48
+ - task:
49
+ type: text-generation
50
+ name: Text Generation
51
+ dataset:
52
+ name: MMLU (5-Shot)
53
+ type: cais/mmlu
54
+ config: all
55
+ split: test
56
+ args:
57
+ num_few_shot: 5
58
+ metrics:
59
+ - type: acc
60
+ value: 64.64
61
+ name: accuracy
62
+ source:
63
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
64
+ name: Open LLM Leaderboard
65
+ - task:
66
+ type: text-generation
67
+ name: Text Generation
68
+ dataset:
69
+ name: TruthfulQA (0-shot)
70
+ type: truthful_qa
71
+ config: multiple_choice
72
+ split: validation
73
+ args:
74
+ num_few_shot: 0
75
+ metrics:
76
+ - type: mc2
77
+ value: 58.01
78
+ source:
79
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
80
+ name: Open LLM Leaderboard
81
+ - task:
82
+ type: text-generation
83
+ name: Text Generation
84
+ dataset:
85
+ name: Winogrande (5-shot)
86
+ type: winogrande
87
+ config: winogrande_xl
88
+ split: validation
89
+ args:
90
+ num_few_shot: 5
91
+ metrics:
92
+ - type: acc
93
+ value: 79.64
94
+ name: accuracy
95
+ source:
96
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
97
+ name: Open LLM Leaderboard
98
+ - task:
99
+ type: text-generation
100
+ name: Text Generation
101
+ dataset:
102
+ name: GSM8k (5-shot)
103
+ type: gsm8k
104
+ config: main
105
+ split: test
106
+ args:
107
+ num_few_shot: 5
108
+ metrics:
109
+ - type: acc
110
+ value: 51.18
111
+ name: accuracy
112
+ source:
113
+ url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=InferenceIllusionist/Magic-Dolphin-7b
114
+ name: Open LLM Leaderboard
115
+ quantized_by: bartowski
116
+ pipeline_tag: text-generation
117
+ ---
118
+
119
+ ## Exllama v2 Quantizations of Magic-Dolphin-7b
120
+
121
+ Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.14">turboderp's ExLlamaV2 v0.0.14</a> for quantization.
122
+
123
+ <b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
124
+
125
+ Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
126
+
127
+ Original model: https://huggingface.co/InferenceIllusionist/Magic-Dolphin-7b
128
+
129
+ | Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
130
+ | ----- | ---- | ------- | ------ | ------ | ------ | ------------ |
131
+ | [8_0](https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2/tree/8_0) | 8.0 | 8.0 | 8.4 GB | 9.8 GB | 11.8 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
132
+ | [6_5](https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2/tree/6_5) | 6.5 | 8.0 | 7.2 GB | 8.6 GB | 10.6 GB | Very similar to 8.0, good tradeoff of size vs performance, **recommended**. |
133
+ | [5_0](https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2/tree/5_0) | 5.0 | 6.0 | 6.0 GB | 7.4 GB | 9.4 GB | Slightly lower quality vs 6.5, but usable on 8GB cards. |
134
+ | [4_25](https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2/tree/4_25) | 4.25 | 6.0 | 5.3 GB | 6.7 GB | 8.7 GB | GPTQ equivalent bits per weight, slightly higher quality. |
135
+ | [3_5](https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2/tree/3_5) | 3.5 | 6.0 | 4.7 GB | 6.1 GB | 8.1 GB | Lower quality, only use if you have to. |
136
+
137
+ ## Download instructions
138
+
139
+ With git:
140
+
141
+ ```shell
142
+ git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/Magic-Dolphin-7b-exl2 Magic-Dolphin-7b-exl2-6_5
143
+ ```
144
+
145
+ With huggingface hub (credit to TheBloke for instructions):
146
+
147
+ ```shell
148
+ pip3 install huggingface-hub
149
+ ```
150
+
151
+ To download the `main` (only useful if you only care about measurement.json) branch to a folder called `Magic-Dolphin-7b-exl2`:
152
+
153
+ ```shell
154
+ mkdir Magic-Dolphin-7b-exl2
155
+ huggingface-cli download bartowski/Magic-Dolphin-7b-exl2 --local-dir Magic-Dolphin-7b-exl2 --local-dir-use-symlinks False
156
+ ```
157
+
158
+ To download from a different branch, add the `--revision` parameter:
159
+
160
+ Linux:
161
+
162
+ ```shell
163
+ mkdir Magic-Dolphin-7b-exl2-6_5
164
+ huggingface-cli download bartowski/Magic-Dolphin-7b-exl2 --revision 6_5 --local-dir Magic-Dolphin-7b-exl2-6_5 --local-dir-use-symlinks False
165
+ ```
166
+
167
+ Windows (which apparently doesn't like _ in folders sometimes?):
168
+
169
+ ```shell
170
+ mkdir Magic-Dolphin-7b-exl2-6.5
171
+ huggingface-cli download bartowski/Magic-Dolphin-7b-exl2 --revision 6_5 --local-dir Magic-Dolphin-7b-exl2-6.5 --local-dir-use-symlinks False
172
+ ```
173
+
174
+ Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
measurement.json ADDED
The diff for this file is too large to render. See raw diff