Spaces:
Runtime error
Runtime error
Add all NLP tasks
Browse files
app.py
CHANGED
@@ -2,6 +2,7 @@ import os
|
|
2 |
import uuid
|
3 |
from pathlib import Path
|
4 |
|
|
|
5 |
import streamlit as st
|
6 |
from datasets import get_dataset_config_names
|
7 |
from dotenv import load_dotenv
|
@@ -83,10 +84,7 @@ with st.expander("Advanced configuration"):
|
|
83 |
domain="https://datasets-preview.huggingface.tech",
|
84 |
params={"dataset": selected_dataset, "config": selected_config, "split": selected_split},
|
85 |
).json()
|
86 |
-
|
87 |
-
col_names = []
|
88 |
-
for c in columns:
|
89 |
-
col_names.append(c["column"]["name"])
|
90 |
# splits = metadata[0]["splits"]
|
91 |
# split_names = list(splits.values())
|
92 |
# eval_split = splits.get("eval_split", split_names[0])
|
@@ -104,28 +102,105 @@ with st.expander("Advanced configuration"):
|
|
104 |
# TODO: propagate this information to payload
|
105 |
# TODO: make it task specific
|
106 |
col_mapping = {}
|
107 |
-
|
108 |
-
|
109 |
st.markdown("`text` column")
|
110 |
st.text("")
|
111 |
st.text("")
|
112 |
st.text("")
|
113 |
st.text("")
|
114 |
st.markdown("`target` column")
|
115 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
116 |
st.markdown("`context` column")
|
117 |
st.text("")
|
118 |
st.text("")
|
119 |
st.text("")
|
120 |
st.text("")
|
121 |
st.markdown("`question` column")
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
|
130 |
with st.form(key="form"):
|
131 |
|
@@ -158,6 +233,7 @@ with st.form(key="form"):
|
|
158 |
},
|
159 |
},
|
160 |
}
|
|
|
161 |
project_json_resp = http_post(
|
162 |
path="/projects/create", payload=payload, token=HF_TOKEN, domain=AUTOTRAIN_BACKEND_API
|
163 |
).json()
|
|
|
2 |
import uuid
|
3 |
from pathlib import Path
|
4 |
|
5 |
+
import pandas as pd
|
6 |
import streamlit as st
|
7 |
from datasets import get_dataset_config_names
|
8 |
from dotenv import load_dotenv
|
|
|
84 |
domain="https://datasets-preview.huggingface.tech",
|
85 |
params={"dataset": selected_dataset, "config": selected_config, "split": selected_split},
|
86 |
).json()
|
87 |
+
col_names = list(pd.json_normalize(rows_resp["rows"][0]["row"]).columns)
|
|
|
|
|
|
|
88 |
# splits = metadata[0]["splits"]
|
89 |
# split_names = list(splits.values())
|
90 |
# eval_split = splits.get("eval_split", split_names[0])
|
|
|
102 |
# TODO: propagate this information to payload
|
103 |
# TODO: make it task specific
|
104 |
col_mapping = {}
|
105 |
+
if selected_task in ["binary_classification", "multi_class_classification"]:
|
106 |
+
with col1:
|
107 |
st.markdown("`text` column")
|
108 |
st.text("")
|
109 |
st.text("")
|
110 |
st.text("")
|
111 |
st.text("")
|
112 |
st.markdown("`target` column")
|
113 |
+
with col2:
|
114 |
+
text_col = st.selectbox("This column should contain the text you want to classify", col_names)
|
115 |
+
target_col = st.selectbox(
|
116 |
+
"This column should contain the labels you want to assign to the text", col_names
|
117 |
+
)
|
118 |
+
col_mapping[text_col] = "text"
|
119 |
+
col_mapping[target_col] = "target"
|
120 |
+
|
121 |
+
elif selected_task == "entity_extraction":
|
122 |
+
with col1:
|
123 |
+
st.markdown("`tokens` column")
|
124 |
+
st.text("")
|
125 |
+
st.text("")
|
126 |
+
st.text("")
|
127 |
+
st.text("")
|
128 |
+
st.markdown("`tags` column")
|
129 |
+
with col2:
|
130 |
+
tokens_col = st.selectbox(
|
131 |
+
"This column should contain the parts of the text (as an array of tokens) you want to assign labels to",
|
132 |
+
col_names,
|
133 |
+
)
|
134 |
+
tags_col = st.selectbox(
|
135 |
+
"This column should contain the labels to associate to each part of the text", col_names
|
136 |
+
)
|
137 |
+
col_mapping[tokens_col] = "tokens"
|
138 |
+
col_mapping[tags_col] = "tags"
|
139 |
+
|
140 |
+
elif selected_task == "translation":
|
141 |
+
with col1:
|
142 |
+
st.markdown("`source` column")
|
143 |
+
st.text("")
|
144 |
+
st.text("")
|
145 |
+
st.text("")
|
146 |
+
st.text("")
|
147 |
+
st.markdown("`target` column")
|
148 |
+
with col2:
|
149 |
+
text_col = st.selectbox("This column should contain the text you want to translate", col_names)
|
150 |
+
target_col = st.selectbox(
|
151 |
+
"This column should contain an example translation of the source text", col_names
|
152 |
+
)
|
153 |
+
col_mapping[text_col] = "source"
|
154 |
+
col_mapping[target_col] = "target"
|
155 |
+
|
156 |
+
elif selected_task == "summarization":
|
157 |
+
with col1:
|
158 |
+
st.markdown("`text` column")
|
159 |
+
st.text("")
|
160 |
+
st.text("")
|
161 |
+
st.text("")
|
162 |
+
st.text("")
|
163 |
+
st.markdown("`target` column")
|
164 |
+
with col2:
|
165 |
+
text_col = st.selectbox("This column should contain the text you want to summarize", col_names)
|
166 |
+
target_col = st.selectbox("This column should contain an example summarization of the text", col_names)
|
167 |
+
col_mapping[text_col] = "text"
|
168 |
+
col_mapping[target_col] = "target"
|
169 |
+
|
170 |
+
elif selected_task == "extractive_question_answering":
|
171 |
+
with col1:
|
172 |
st.markdown("`context` column")
|
173 |
st.text("")
|
174 |
st.text("")
|
175 |
st.text("")
|
176 |
st.text("")
|
177 |
st.markdown("`question` column")
|
178 |
+
st.text("")
|
179 |
+
st.text("")
|
180 |
+
st.text("")
|
181 |
+
st.text("")
|
182 |
+
st.markdown("`answers.text` column")
|
183 |
+
st.text("")
|
184 |
+
st.text("")
|
185 |
+
st.text("")
|
186 |
+
st.text("")
|
187 |
+
st.markdown("`answers.answer_start` column")
|
188 |
+
with col2:
|
189 |
+
context_col = st.selectbox("This column should contain the question's context", col_names)
|
190 |
+
question_col = st.selectbox(
|
191 |
+
"This column should contain the question to be answered, given the context", col_names
|
192 |
+
)
|
193 |
+
answers_text_col = st.selectbox(
|
194 |
+
"This column should contain example answers to the question, extracted from the context", col_names
|
195 |
+
)
|
196 |
+
answers_start_col = st.selectbox(
|
197 |
+
"This column should contain the indices in the context of the first character of each answers.text",
|
198 |
+
col_names,
|
199 |
+
)
|
200 |
+
col_mapping[context_col] = "context"
|
201 |
+
col_mapping[question_col] = "question"
|
202 |
+
col_mapping[answers_text_col] = "answers.text"
|
203 |
+
col_mapping[answers_start_col] = "answers.answer_start"
|
204 |
|
205 |
with st.form(key="form"):
|
206 |
|
|
|
233 |
},
|
234 |
},
|
235 |
}
|
236 |
+
print(f"Payload: {payload}")
|
237 |
project_json_resp = http_post(
|
238 |
path="/projects/create", payload=payload, token=HF_TOKEN, domain=AUTOTRAIN_BACKEND_API
|
239 |
).json()
|
utils.py
CHANGED
@@ -57,6 +57,9 @@ def get_metadata(dataset_name: str) -> Union[Dict, None]:
|
|
57 |
|
58 |
|
59 |
def get_compatible_models(task, dataset_name):
|
60 |
-
|
|
|
|
|
|
|
61 |
compatible_models = api.list_models(filter=filt)
|
62 |
return [model.modelId for model in compatible_models]
|
|
|
57 |
|
58 |
|
59 |
def get_compatible_models(task, dataset_name):
|
60 |
+
# TODO: relax filter on PyTorch models once supported in AutoTrain
|
61 |
+
filt = ModelFilter(
|
62 |
+
task=AUTOTRAIN_TASK_TO_HUB_TASK[task], trained_dataset=dataset_name, library=["transformers", "pytorch"]
|
63 |
+
)
|
64 |
compatible_models = api.list_models(filter=filt)
|
65 |
return [model.modelId for model in compatible_models]
|