bartowski commited on
Commit
bc9d6a9
·
verified ·
1 Parent(s): 358f428

measurement.json

Browse files
Files changed (2) hide show
  1. README.md +149 -0
  2. measurement.json +0 -0
README.md ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pipeline_tag: text-generation
3
+ base_model: bigcode/starcoder2-15b
4
+ datasets:
5
+ - bigcode/self-oss-instruct-sc2-exec-filter-50k
6
+ license: bigcode-openrail-m
7
+ library_name: transformers
8
+ tags:
9
+ - code
10
+ model-index:
11
+ - name: starcoder2-15b-instruct-v0.1
12
+ results:
13
+ - task:
14
+ type: text-generation
15
+ dataset:
16
+ name: LiveCodeBench (code generation)
17
+ type: livecodebench-codegeneration
18
+ metrics:
19
+ - type: pass@1
20
+ value: 20.4
21
+ - task:
22
+ type: text-generation
23
+ dataset:
24
+ name: LiveCodeBench (self repair)
25
+ type: livecodebench-selfrepair
26
+ metrics:
27
+ - type: pass@1
28
+ value: 20.9
29
+ - task:
30
+ type: text-generation
31
+ dataset:
32
+ name: LiveCodeBench (test output prediction)
33
+ type: livecodebench-testoutputprediction
34
+ metrics:
35
+ - type: pass@1
36
+ value: 29.8
37
+ - task:
38
+ type: text-generation
39
+ dataset:
40
+ name: LiveCodeBench (code execution)
41
+ type: livecodebench-codeexecution
42
+ metrics:
43
+ - type: pass@1
44
+ value: 28.1
45
+ - task:
46
+ type: text-generation
47
+ dataset:
48
+ name: HumanEval
49
+ type: humaneval
50
+ metrics:
51
+ - type: pass@1
52
+ value: 72.6
53
+ - task:
54
+ type: text-generation
55
+ dataset:
56
+ name: HumanEval+
57
+ type: humanevalplus
58
+ metrics:
59
+ - type: pass@1
60
+ value: 63.4
61
+ - task:
62
+ type: text-generation
63
+ dataset:
64
+ name: MBPP
65
+ type: mbpp
66
+ metrics:
67
+ - type: pass@1
68
+ value: 75.2
69
+ - task:
70
+ type: text-generation
71
+ dataset:
72
+ name: MBPP+
73
+ type: mbppplus
74
+ metrics:
75
+ - type: pass@1
76
+ value: 61.2
77
+ - task:
78
+ type: text-generation
79
+ dataset:
80
+ name: DS-1000
81
+ type: ds-1000
82
+ metrics:
83
+ - type: pass@1
84
+ value: 40.6
85
+ quantized_by: bartowski
86
+ ---
87
+
88
+ ## Exllama v2 Quantizations of starcoder2-15b-instruct-v0.1
89
+
90
+ Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.20">turboderp's ExLlamaV2 v0.0.20</a> for quantization.
91
+
92
+ <b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>
93
+
94
+ Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.
95
+
96
+ Conversion was done using the default calibration dataset.
97
+
98
+ Default arguments used except when the bits per weight is above 6.0, at that point the lm_head layer is quantized at 8 bits per weight instead of the default 6.
99
+
100
+ Original model: https://huggingface.co/bigcode/starcoder2-15b-instruct-v0.1
101
+
102
+
103
+ <a href="https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2/tree/8_0">8.0 bits per weight</a>
104
+
105
+ <a href="https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2/tree/6_5">6.5 bits per weight</a>
106
+
107
+ <a href="https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2/tree/5_0">5.0 bits per weight</a>
108
+
109
+ <a href="https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2/tree/4_25">4.25 bits per weight</a>
110
+
111
+ <a href="https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2/tree/3_5">3.5 bits per weight</a>
112
+
113
+
114
+ ## Download instructions
115
+
116
+ With git:
117
+
118
+ ```shell
119
+ git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/starcoder2-15b-instruct-v0.1-exl2
120
+ ```
121
+
122
+ With huggingface hub (credit to TheBloke for instructions):
123
+
124
+ ```shell
125
+ pip3 install huggingface-hub
126
+ ```
127
+
128
+ To download the `main` (only useful if you only care about measurement.json) branch to a folder called `starcoder2-15b-instruct-v0.1-exl2`:
129
+
130
+ ```shell
131
+ mkdir starcoder2-15b-instruct-v0.1-exl2
132
+ huggingface-cli download bartowski/starcoder2-15b-instruct-v0.1-exl2 --local-dir starcoder2-15b-instruct-v0.1-exl2 --local-dir-use-symlinks False
133
+ ```
134
+
135
+ To download from a different branch, add the `--revision` parameter:
136
+
137
+ Linux:
138
+
139
+ ```shell
140
+ mkdir starcoder2-15b-instruct-v0.1-exl2-6_5
141
+ huggingface-cli download bartowski/starcoder2-15b-instruct-v0.1-exl2 --revision 6_5 --local-dir starcoder2-15b-instruct-v0.1-exl2-6_5 --local-dir-use-symlinks False
142
+ ```
143
+
144
+ Windows (which apparently doesn't like _ in folders sometimes?):
145
+
146
+ ```shell
147
+ mkdir starcoder2-15b-instruct-v0.1-exl2-6.5
148
+ huggingface-cli download bartowski/starcoder2-15b-instruct-v0.1-exl2 --revision 6_5 --local-dir starcoder2-15b-instruct-v0.1-exl2-6.5 --local-dir-use-symlinks False
149
+ ```
measurement.json ADDED
The diff for this file is too large to render. See raw diff