SaylorTwift HF staff commited on
Commit
276951d
1 Parent(s): c7a98ae

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +208 -0
README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: Evaluation run of Yhyu13/llama-30B-hf-openassitant
3
+ dataset_summary: "Dataset automatically created during the evaluation run of model\
4
+ \ [Yhyu13/llama-30B-hf-openassitant](https://huggingface.co/Yhyu13/llama-30B-hf-openassitant)\
5
+ \ on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).\n\
6
+ \nThe dataset is composed of 3 configuration, each one coresponding to one of the\
7
+ \ evaluated task.\n\nThe dataset has been created from 1 run(s). Each run can be\
8
+ \ found as a specific split in each configuration, the split being named using the\
9
+ \ timestamp of the run.The \"train\" split is always pointing to the latest results.\n\
10
+ \nAn additional configuration \"results\" store all the aggregated results of the\
11
+ \ run (and is used to compute and display the agregated metrics on the [Open LLM\
12
+ \ Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)).\n\
13
+ \nTo load the details from a run, you can for instance do the following:\n```python\n\
14
+ from datasets import load_dataset\ndata = load_dataset(\"open-llm-leaderboard/details_Yhyu13__llama-30B-hf-openassitant\"\
15
+ ,\n\t\"harness_winogrande_5\",\n\tsplit=\"train\")\n```\n\n## Latest results\n\n\
16
+ These are the [latest results from run 2023-10-15T22:09:11.828298](https://huggingface.co/datasets/open-llm-leaderboard/details_Yhyu13__llama-30B-hf-openassitant/blob/main/results_2023-10-15T22-09-11.828298.json)(note\
17
+ \ that their might be results for other tasks in the repos if successive evals didn't\
18
+ \ cover the same tasks. You find each in the results and the \"latest\" split for\
19
+ \ each eval):\n\n```python\n{\n \"all\": {\n \"em\": 0.0014681208053691276,\n\
20
+ \ \"em_stderr\": 0.0003921042190298701,\n \"f1\": 0.06332634228187943,\n\
21
+ \ \"f1_stderr\": 0.0013742294190200051,\n \"acc\": 0.47445656434133393,\n\
22
+ \ \"acc_stderr\": 0.010516415781576863\n },\n \"harness|drop|3\": {\n\
23
+ \ \"em\": 0.0014681208053691276,\n \"em_stderr\": 0.0003921042190298701,\n\
24
+ \ \"f1\": 0.06332634228187943,\n \"f1_stderr\": 0.0013742294190200051\n\
25
+ \ },\n \"harness|gsm8k|5\": {\n \"acc\": 0.14859742228961334,\n \
26
+ \ \"acc_stderr\": 0.009797503180527876\n },\n \"harness|winogrande|5\"\
27
+ : {\n \"acc\": 0.8003157063930545,\n \"acc_stderr\": 0.011235328382625849\n\
28
+ \ }\n}\n```"
29
+ repo_url: https://huggingface.co/Yhyu13/llama-30B-hf-openassitant
30
+ leaderboard_url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
31
+ point_of_contact: clementine@hf.co
32
+ configs:
33
+ - config_name: harness_drop_3
34
+ data_files:
35
+ - split: 2023_10_15T22_09_11.828298
36
+ path:
37
+ - '**/details_harness|drop|3_2023-10-15T22-09-11.828298.parquet'
38
+ - split: latest
39
+ path:
40
+ - '**/details_harness|drop|3_2023-10-15T22-09-11.828298.parquet'
41
+ - config_name: harness_gsm8k_5
42
+ data_files:
43
+ - split: 2023_10_15T22_09_11.828298
44
+ path:
45
+ - '**/details_harness|gsm8k|5_2023-10-15T22-09-11.828298.parquet'
46
+ - split: latest
47
+ path:
48
+ - '**/details_harness|gsm8k|5_2023-10-15T22-09-11.828298.parquet'
49
+ - config_name: harness_winogrande_5
50
+ data_files:
51
+ - split: 2023_10_15T22_09_11.828298
52
+ path:
53
+ - '**/details_harness|winogrande|5_2023-10-15T22-09-11.828298.parquet'
54
+ - split: latest
55
+ path:
56
+ - '**/details_harness|winogrande|5_2023-10-15T22-09-11.828298.parquet'
57
+ - config_name: results
58
+ data_files:
59
+ - split: 2023_10_15T22_09_11.828298
60
+ path:
61
+ - results_2023-10-15T22-09-11.828298.parquet
62
+ - split: latest
63
+ path:
64
+ - results_2023-10-15T22-09-11.828298.parquet
65
+ ---
66
+
67
+ # Dataset Card for Evaluation run of Yhyu13/llama-30B-hf-openassitant
68
+
69
+ ## Dataset Description
70
+
71
+ - **Homepage:**
72
+ - **Repository:** https://huggingface.co/Yhyu13/llama-30B-hf-openassitant
73
+ - **Paper:**
74
+ - **Leaderboard:** https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
75
+ - **Point of Contact:** clementine@hf.co
76
+
77
+ ### Dataset Summary
78
+
79
+ Dataset automatically created during the evaluation run of model [Yhyu13/llama-30B-hf-openassitant](https://huggingface.co/Yhyu13/llama-30B-hf-openassitant) on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).
80
+
81
+ The dataset is composed of 3 configuration, each one coresponding to one of the evaluated task.
82
+
83
+ The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
84
+
85
+ An additional configuration "results" store all the aggregated results of the run (and is used to compute and display the agregated metrics on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)).
86
+
87
+ To load the details from a run, you can for instance do the following:
88
+ ```python
89
+ from datasets import load_dataset
90
+ data = load_dataset("open-llm-leaderboard/details_Yhyu13__llama-30B-hf-openassitant",
91
+ "harness_winogrande_5",
92
+ split="train")
93
+ ```
94
+
95
+ ## Latest results
96
+
97
+ These are the [latest results from run 2023-10-15T22:09:11.828298](https://huggingface.co/datasets/open-llm-leaderboard/details_Yhyu13__llama-30B-hf-openassitant/blob/main/results_2023-10-15T22-09-11.828298.json)(note that their might be results for other tasks in the repos if successive evals didn't cover the same tasks. You find each in the results and the "latest" split for each eval):
98
+
99
+ ```python
100
+ {
101
+ "all": {
102
+ "em": 0.0014681208053691276,
103
+ "em_stderr": 0.0003921042190298701,
104
+ "f1": 0.06332634228187943,
105
+ "f1_stderr": 0.0013742294190200051,
106
+ "acc": 0.47445656434133393,
107
+ "acc_stderr": 0.010516415781576863
108
+ },
109
+ "harness|drop|3": {
110
+ "em": 0.0014681208053691276,
111
+ "em_stderr": 0.0003921042190298701,
112
+ "f1": 0.06332634228187943,
113
+ "f1_stderr": 0.0013742294190200051
114
+ },
115
+ "harness|gsm8k|5": {
116
+ "acc": 0.14859742228961334,
117
+ "acc_stderr": 0.009797503180527876
118
+ },
119
+ "harness|winogrande|5": {
120
+ "acc": 0.8003157063930545,
121
+ "acc_stderr": 0.011235328382625849
122
+ }
123
+ }
124
+ ```
125
+
126
+ ### Supported Tasks and Leaderboards
127
+
128
+ [More Information Needed]
129
+
130
+ ### Languages
131
+
132
+ [More Information Needed]
133
+
134
+ ## Dataset Structure
135
+
136
+ ### Data Instances
137
+
138
+ [More Information Needed]
139
+
140
+ ### Data Fields
141
+
142
+ [More Information Needed]
143
+
144
+ ### Data Splits
145
+
146
+ [More Information Needed]
147
+
148
+ ## Dataset Creation
149
+
150
+ ### Curation Rationale
151
+
152
+ [More Information Needed]
153
+
154
+ ### Source Data
155
+
156
+ #### Initial Data Collection and Normalization
157
+
158
+ [More Information Needed]
159
+
160
+ #### Who are the source language producers?
161
+
162
+ [More Information Needed]
163
+
164
+ ### Annotations
165
+
166
+ #### Annotation process
167
+
168
+ [More Information Needed]
169
+
170
+ #### Who are the annotators?
171
+
172
+ [More Information Needed]
173
+
174
+ ### Personal and Sensitive Information
175
+
176
+ [More Information Needed]
177
+
178
+ ## Considerations for Using the Data
179
+
180
+ ### Social Impact of Dataset
181
+
182
+ [More Information Needed]
183
+
184
+ ### Discussion of Biases
185
+
186
+ [More Information Needed]
187
+
188
+ ### Other Known Limitations
189
+
190
+ [More Information Needed]
191
+
192
+ ## Additional Information
193
+
194
+ ### Dataset Curators
195
+
196
+ [More Information Needed]
197
+
198
+ ### Licensing Information
199
+
200
+ [More Information Needed]
201
+
202
+ ### Citation Information
203
+
204
+ [More Information Needed]
205
+
206
+ ### Contributions
207
+
208
+ [More Information Needed]