Spaces:
Sleeping
Sleeping
add slider for page size
Browse files- app.py +3 -1
- arxiv.ipynb +119 -0
app.py
CHANGED
@@ -38,6 +38,7 @@ def handle_dataset_view(page: int = 1, page_size: int = 10) -> Dict[str, Any]:
|
|
38 |
result = {
|
39 |
"total_records": total_records,
|
40 |
"current_page": page,
|
|
|
41 |
"records": records
|
42 |
}
|
43 |
logging.info(f"Returning result: {result}")
|
@@ -70,12 +71,13 @@ with gr.Blocks() as demo:
|
|
70 |
|
71 |
with gr.Tab("View Dataset"):
|
72 |
page_number = gr.Number(value=1, label="Page Number", precision=0)
|
|
|
73 |
refresh_button = gr.Button("Refresh Dataset View")
|
74 |
dataset_info = gr.JSON(label="Dataset Info")
|
75 |
|
76 |
refresh_button.click(
|
77 |
fn=handle_dataset_view,
|
78 |
-
inputs=[page_number],
|
79 |
outputs=dataset_info
|
80 |
)
|
81 |
|
|
|
38 |
result = {
|
39 |
"total_records": total_records,
|
40 |
"current_page": page,
|
41 |
+
"page_size": page_size,
|
42 |
"records": records
|
43 |
}
|
44 |
logging.info(f"Returning result: {result}")
|
|
|
71 |
|
72 |
with gr.Tab("View Dataset"):
|
73 |
page_number = gr.Number(value=1, label="Page Number", precision=0)
|
74 |
+
page_size = gr.Slider(minimum=5, maximum=50, value=10, step=5, label="Page Size")
|
75 |
refresh_button = gr.Button("Refresh Dataset View")
|
76 |
dataset_info = gr.JSON(label="Dataset Info")
|
77 |
|
78 |
refresh_button.click(
|
79 |
fn=handle_dataset_view,
|
80 |
+
inputs=[page_number, page_size],
|
81 |
outputs=dataset_info
|
82 |
)
|
83 |
|
arxiv.ipynb
ADDED
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": 1,
|
6 |
+
"metadata": {},
|
7 |
+
"outputs": [],
|
8 |
+
"source": [
|
9 |
+
"import arxiv\n",
|
10 |
+
"\n",
|
11 |
+
"client = arxiv.Client(delay_seconds=3, num_retries=3)\n",
|
12 |
+
"\n",
|
13 |
+
"\n",
|
14 |
+
"max_results: int = 10\n",
|
15 |
+
"\n",
|
16 |
+
"search = arxiv.Search(\n",
|
17 |
+
" query=\"2304.08485\", \n",
|
18 |
+
" max_results=max_results, \n",
|
19 |
+
" sort_by=arxiv.SortCriterion.SubmittedDate\n",
|
20 |
+
" )"
|
21 |
+
]
|
22 |
+
},
|
23 |
+
{
|
24 |
+
"cell_type": "code",
|
25 |
+
"execution_count": 2,
|
26 |
+
"metadata": {},
|
27 |
+
"outputs": [
|
28 |
+
{
|
29 |
+
"name": "stdout",
|
30 |
+
"output_type": "stream",
|
31 |
+
"text": [
|
32 |
+
"arxiv.Search(query='2304.08485', id_list=[], max_results=10, sort_by=<SortCriterion.SubmittedDate: 'submittedDate'>, sort_order=<SortOrder.Descending: 'descending'>)\n"
|
33 |
+
]
|
34 |
+
}
|
35 |
+
],
|
36 |
+
"source": [
|
37 |
+
"print(search)"
|
38 |
+
]
|
39 |
+
},
|
40 |
+
{
|
41 |
+
"cell_type": "code",
|
42 |
+
"execution_count": 24,
|
43 |
+
"metadata": {},
|
44 |
+
"outputs": [
|
45 |
+
{
|
46 |
+
"name": "stdout",
|
47 |
+
"output_type": "stream",
|
48 |
+
"text": [
|
49 |
+
"entry_id: http://arxiv.org/abs/2304.08485v2\n",
|
50 |
+
"updated: 2023-12-11 17:46:14+00:00\n",
|
51 |
+
"published: 2023-04-17 17:59:25+00:00\n",
|
52 |
+
"title: Visual Instruction Tuning\n",
|
53 |
+
"authors: [arxiv.Result.Author('Haotian Liu'), arxiv.Result.Author('Chunyuan Li'), arxiv.Result.Author('Qingyang Wu'), arxiv.Result.Author('Yong Jae Lee')]\n",
|
54 |
+
"summary: Instruction tuning large language models (LLMs) using machine-generated\n",
|
55 |
+
"instruction-following data has improved zero-shot capabilities on new tasks,\n",
|
56 |
+
"but the idea is less explored in the multimodal field. In this paper, we\n",
|
57 |
+
"present the first attempt to use language-only GPT-4 to generate multimodal\n",
|
58 |
+
"language-image instruction-following data. By instruction tuning on such\n",
|
59 |
+
"generated data, we introduce LLaVA: Large Language and Vision Assistant, an\n",
|
60 |
+
"end-to-end trained large multimodal model that connects a vision encoder and\n",
|
61 |
+
"LLM for general-purpose visual and language understanding.Our early experiments\n",
|
62 |
+
"show that LLaVA demonstrates impressive multimodel chat abilities, sometimes\n",
|
63 |
+
"exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and\n",
|
64 |
+
"yields a 85.1% relative score compared with GPT-4 on a synthetic multimodal\n",
|
65 |
+
"instruction-following dataset. When fine-tuned on Science QA, the synergy of\n",
|
66 |
+
"LLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make\n",
|
67 |
+
"GPT-4 generated visual instruction tuning data, our model and code base\n",
|
68 |
+
"publicly available.\n",
|
69 |
+
"comment: NeurIPS 2023 Oral; project page: https://llava-vl.github.io/\n",
|
70 |
+
"journal_ref: None\n",
|
71 |
+
"doi: None\n",
|
72 |
+
"primary_category: cs.CV\n",
|
73 |
+
"categories: ['cs.CV', 'cs.AI', 'cs.CL', 'cs.LG']\n",
|
74 |
+
"links: [arxiv.Result.Link('http://arxiv.org/abs/2304.08485v2', title=None, rel='alternate', content_type=None), arxiv.Result.Link('http://arxiv.org/pdf/2304.08485v2', title='pdf', rel='related', content_type=None)]\n",
|
75 |
+
"pdf_url: http://arxiv.org/pdf/2304.08485v2\n",
|
76 |
+
"_raw: {'id': 'http://arxiv.org/abs/2304.08485v2', 'guidislink': True, 'link': 'http://arxiv.org/abs/2304.08485v2', 'updated': '2023-12-11T17:46:14Z', 'updated_parsed': time.struct_time(tm_year=2023, tm_mon=12, tm_mday=11, tm_hour=17, tm_min=46, tm_sec=14, tm_wday=0, tm_yday=345, tm_isdst=0), 'published': '2023-04-17T17:59:25Z', 'published_parsed': time.struct_time(tm_year=2023, tm_mon=4, tm_mday=17, tm_hour=17, tm_min=59, tm_sec=25, tm_wday=0, tm_yday=107, tm_isdst=0), 'title': 'Visual Instruction Tuning', 'title_detail': {'type': 'text/plain', 'language': None, 'base': '', 'value': 'Visual Instruction Tuning'}, 'summary': 'Instruction tuning large language models (LLMs) using machine-generated\\ninstruction-following data has improved zero-shot capabilities on new tasks,\\nbut the idea is less explored in the multimodal field. In this paper, we\\npresent the first attempt to use language-only GPT-4 to generate multimodal\\nlanguage-image instruction-following data. By instruction tuning on such\\ngenerated data, we introduce LLaVA: Large Language and Vision Assistant, an\\nend-to-end trained large multimodal model that connects a vision encoder and\\nLLM for general-purpose visual and language understanding.Our early experiments\\nshow that LLaVA demonstrates impressive multimodel chat abilities, sometimes\\nexhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and\\nyields a 85.1% relative score compared with GPT-4 on a synthetic multimodal\\ninstruction-following dataset. When fine-tuned on Science QA, the synergy of\\nLLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make\\nGPT-4 generated visual instruction tuning data, our model and code base\\npublicly available.', 'summary_detail': {'type': 'text/plain', 'language': None, 'base': '', 'value': 'Instruction tuning large language models (LLMs) using machine-generated\\ninstruction-following data has improved zero-shot capabilities on new tasks,\\nbut the idea is less explored in the multimodal field. In this paper, we\\npresent the first attempt to use language-only GPT-4 to generate multimodal\\nlanguage-image instruction-following data. By instruction tuning on such\\ngenerated data, we introduce LLaVA: Large Language and Vision Assistant, an\\nend-to-end trained large multimodal model that connects a vision encoder and\\nLLM for general-purpose visual and language understanding.Our early experiments\\nshow that LLaVA demonstrates impressive multimodel chat abilities, sometimes\\nexhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and\\nyields a 85.1% relative score compared with GPT-4 on a synthetic multimodal\\ninstruction-following dataset. When fine-tuned on Science QA, the synergy of\\nLLaVA and GPT-4 achieves a new state-of-the-art accuracy of 92.53%. We make\\nGPT-4 generated visual instruction tuning data, our model and code base\\npublicly available.'}, 'authors': [{'name': 'Haotian Liu'}, {'name': 'Chunyuan Li'}, {'name': 'Qingyang Wu'}, {'name': 'Yong Jae Lee'}], 'author_detail': {'name': 'Yong Jae Lee'}, 'author': 'Yong Jae Lee', 'arxiv_comment': 'NeurIPS 2023 Oral; project page: https://llava-vl.github.io/', 'links': [{'href': 'http://arxiv.org/abs/2304.08485v2', 'rel': 'alternate', 'type': 'text/html'}, {'title': 'pdf', 'href': 'http://arxiv.org/pdf/2304.08485v2', 'rel': 'related', 'type': 'application/pdf'}], 'arxiv_primary_category': {'term': 'cs.CV', 'scheme': 'http://arxiv.org/schemas/atom'}, 'tags': [{'term': 'cs.CV', 'scheme': 'http://arxiv.org/schemas/atom', 'label': None}, {'term': 'cs.AI', 'scheme': 'http://arxiv.org/schemas/atom', 'label': None}, {'term': 'cs.CL', 'scheme': 'http://arxiv.org/schemas/atom', 'label': None}, {'term': 'cs.LG', 'scheme': 'http://arxiv.org/schemas/atom', 'label': None}]}\n"
|
77 |
+
]
|
78 |
+
}
|
79 |
+
],
|
80 |
+
"source": [
|
81 |
+
"results = []\n",
|
82 |
+
"for result in client.results(search):\n",
|
83 |
+
" results.append(result)\n",
|
84 |
+
" # print all key value pairs in \"key: value\" format\n",
|
85 |
+
" for key, value in vars(result).items():\n",
|
86 |
+
" print(f\"{key}: {value}\")\n",
|
87 |
+
"\n"
|
88 |
+
]
|
89 |
+
},
|
90 |
+
{
|
91 |
+
"cell_type": "code",
|
92 |
+
"execution_count": null,
|
93 |
+
"metadata": {},
|
94 |
+
"outputs": [],
|
95 |
+
"source": []
|
96 |
+
}
|
97 |
+
],
|
98 |
+
"metadata": {
|
99 |
+
"kernelspec": {
|
100 |
+
"display_name": ".venv",
|
101 |
+
"language": "python",
|
102 |
+
"name": "python3"
|
103 |
+
},
|
104 |
+
"language_info": {
|
105 |
+
"codemirror_mode": {
|
106 |
+
"name": "ipython",
|
107 |
+
"version": 3
|
108 |
+
},
|
109 |
+
"file_extension": ".py",
|
110 |
+
"mimetype": "text/x-python",
|
111 |
+
"name": "python",
|
112 |
+
"nbconvert_exporter": "python",
|
113 |
+
"pygments_lexer": "ipython3",
|
114 |
+
"version": "3.10.13"
|
115 |
+
}
|
116 |
+
},
|
117 |
+
"nbformat": 4,
|
118 |
+
"nbformat_minor": 2
|
119 |
+
}
|