MCPcopy Create free account
hub / github.com/zai-org/CodeGeeX / _add_data_args

Function _add_data_args

codegeex/megatron/arguments.py:1152–1295  ·  view source on GitHub ↗
(parser)

Source from the content-addressed store, hash-verified

1150
1151
1152def _add_data_args(parser):
1153 group = parser.add_argument_group(title="data and dataloader")
1154
1155 group.add_argument(
1156 "--data-path",
1157 nargs="*",
1158 default=None,
1159 help="Path to the training dataset. Accepted format:"
1160 "1) a single data path, 2) multiple datasets in the"
1161 "form: dataset1-weight dataset1-path dataset2-weight "
1162 "dataset2-path ...",
1163 )
1164 group.add_argument(
1165 "--valid-data-path",
1166 nargs="*",
1167 default=None,
1168 help="Path to the validation dataset. Accepted format:"
1169 "1) a single data path, 2) multiple datasets in the"
1170 "form: dataset1-weight dataset1-path dataset2-weight "
1171 "dataset2-path ...;"
1172 "when co-evaluation is enabled, the form will be dataset1-tag dataset1-path ...",
1173 )
1174 group.add_argument("--index-cache-dir", type=str, default=None, help="Path to the index cache")
1175 group.add_argument(
1176 "--test-data-path",
1177 nargs="*",
1178 default=None,
1179 help="Path to the test dataset. Accepted format:"
1180 "1) a single data path, 2) multiple datasets in the"
1181 "form: dataset1-tag dataset1-path dataset2-tag "
1182 "dataset2-path ...",
1183 )
1184 group.add_argument(
1185 "--split",
1186 type=str,
1187 default="969, 30, 1",
1188 help="Comma-separated list of proportions for training,"
1189 " validation, and test split. For example the split "
1190 "`90,5,5` will use 90%% of data for training, 5%% for "
1191 "validation and 5%% for test.",
1192 )
1193 group.add_argument(
1194 "--vocab-file",
1195 type=str,
1196 default=None,
1197 help="Path to the vocab file.",
1198 )
1199 group.add_argument(
1200 "--merge-file",
1201 type=str,
1202 default=None,
1203 help="Path to the BPE merge file.",
1204 )
1205 group.add_argument(
1206 "--tokenizer-path",
1207 type=str,
1208 default=None,
1209 help="Path to the tokenizer dir.",

Callers 1

parse_argsFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected