diff --git a/combo/data/dataset_readers/universal_dependencies_dataset_reader.py b/combo/data/dataset_readers/universal_dependencies_dataset_reader.py
index 0fa4827987e270b0f986a7df0f158c69c124594d..c094b3c85daadce0cba9a764f20c4ddfb7d710f9 100644
--- a/combo/data/dataset_readers/universal_dependencies_dataset_reader.py
+++ b/combo/data/dataset_readers/universal_dependencies_dataset_reader.py
@@ -27,6 +27,15 @@ from combo.data.vocabulary import get_slices_if_not_provided
from combo.utils import checks, pad_sequence_to_length
+def parser_field_to_dataset_reader_field(field: str):
+ if field == 'upos':
+ return 'upostag'
+ elif field == 'xpos':
+ return 'xpostag'
+ else:
+ return field
+
+
@Registry.register(DatasetReader, 'conllu')
class UniversalDependenciesDatasetReader(DatasetReader, ABC):
def __init__(
@@ -70,8 +79,9 @@ class UniversalDependenciesDatasetReader(DatasetReader, ABC):
fields = list(fields)
fields.append("semrel")
field_parsers["semrel"] = lambda line, i: line[i]
+
self.field_parsers = field_parsers
- self.fields = tuple(fields)
+ self.fields = tuple([parser_field_to_dataset_reader_field(f) for f in fields])
self.__lemma_indexers = lemma_indexers
self.__targets = targets
diff --git a/combo/example.ipynb b/combo/example.ipynb
index 1269fefdb7031762ceba5b7938592f06260d0501..f1174e33adf44f4a3b9b5f808eb39485f8819047 100644
--- a/combo/example.ipynb
+++ b/combo/example.ipynb
@@ -7,8 +7,8 @@
"metadata": {
"collapsed": true,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:43.374412Z",
- "start_time": "2023-09-29T11:59:39.164713Z"
+ "end_time": "2023-09-30T06:30:32.968386Z",
+ "start_time": "2023-09-30T06:30:28.670080Z"
}
},
"outputs": [],
@@ -48,8 +48,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:43.751419Z",
- "start_time": "2023-09-29T11:59:43.356220Z"
+ "end_time": "2023-09-30T06:30:33.433633Z",
+ "start_time": "2023-09-30T06:30:32.965078Z"
}
},
"id": "7302b7d49ac2fc38"
@@ -68,8 +68,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:43.754710Z",
- "start_time": "2023-09-29T11:59:43.743252Z"
+ "end_time": "2023-09-30T06:30:33.502604Z",
+ "start_time": "2023-09-30T06:30:33.457202Z"
}
},
"id": "e0ac599d12cc33df"
@@ -93,8 +93,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:43.787672Z",
- "start_time": "2023-09-29T11:59:43.755206Z"
+ "end_time": "2023-09-30T06:30:33.516731Z",
+ "start_time": "2023-09-30T06:30:33.469116Z"
}
},
"id": "a7f687419bddd9f8"
@@ -119,7 +119,7 @@
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
- "model_id": "55a97b4050364e31a0d13ab013262d15"
+ "model_id": "8cce881f28a94e879336221b2fab2ac3"
}
},
"metadata": {},
@@ -181,8 +181,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:51.813061Z",
- "start_time": "2023-09-29T11:59:43.766557Z"
+ "end_time": "2023-09-30T06:30:47.592831Z",
+ "start_time": "2023-09-30T06:30:33.490375Z"
}
},
"id": "d74957f422f0b05b"
@@ -204,8 +204,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:52.631225Z",
- "start_time": "2023-09-29T11:59:51.818391Z"
+ "end_time": "2023-09-30T06:30:50.225703Z",
+ "start_time": "2023-09-30T06:30:47.479441Z"
}
},
"id": "fa724d362fd6bd23"
@@ -223,7 +223,7 @@
},
{
"data": {
- "text/plain": "<generator object SimpleDataLoader.iter_instances at 0x7fd958d502e0>"
+ "text/plain": "<generator object SimpleDataLoader.iter_instances at 0x7f8db1e44430>"
},
"execution_count": 7,
"metadata": {},
@@ -251,8 +251,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:52.665717Z",
- "start_time": "2023-09-29T11:59:52.630665Z"
+ "end_time": "2023-09-30T06:30:50.244596Z",
+ "start_time": "2023-09-30T06:30:49.859096Z"
}
},
"id": "f8a10f9892005fca"
@@ -265,7 +265,7 @@
"name": "stderr",
"output_type": "stream",
"text": [
- "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertModel: ['cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transform.LayerNorm.weight']\n",
+ "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertModel: ['cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.bias', 'cls.predictions.transform.LayerNorm.weight']\n",
"- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
"- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
]
@@ -396,8 +396,8 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:56.308564Z",
- "start_time": "2023-09-29T11:59:52.662376Z"
+ "end_time": "2023-09-30T06:30:54.651620Z",
+ "start_time": "2023-09-30T06:30:49.865341Z"
}
},
"id": "437d12054baaffa1"
@@ -412,15 +412,15 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:56.323194Z",
- "start_time": "2023-09-29T11:59:56.275602Z"
+ "end_time": "2023-09-30T06:30:54.680188Z",
+ "start_time": "2023-09-30T06:30:54.660096Z"
}
},
"id": "16ae311c44073668"
},
{
"cell_type": "code",
- "execution_count": 10,
+ "execution_count": 16,
"outputs": [],
"source": [
"for i in data_loader.iter_instances():\n",
@@ -430,15 +430,15 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T11:59:56.829311Z",
- "start_time": "2023-09-29T11:59:56.317467Z"
+ "end_time": "2023-09-30T06:31:13.385199Z",
+ "start_time": "2023-09-30T06:31:12.864300Z"
}
},
"id": "e131e0ec75dc6927"
},
{
"cell_type": "code",
- "execution_count": 13,
+ "execution_count": 14,
"outputs": [],
"source": [
"a = nlp(\"__START__ Hello! My name is Dog. __END__\")"
@@ -446,41 +446,78 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T12:00:44.999132Z",
- "start_time": "2023-09-29T12:00:44.612375Z"
+ "end_time": "2023-09-30T06:31:05.928119Z",
+ "start_time": "2023-09-30T06:31:05.519479Z"
}
},
"id": "3e23413c86063183"
},
{
"cell_type": "code",
- "execution_count": 24,
+ "execution_count": 12,
"outputs": [
{
"data": {
- "text/plain": "__START__"
+ "text/plain": "'JJR'"
},
- "execution_count": 24,
+ "execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
- "a.tokens[0]"
+ "a.tokens[-2].xpostag"
],
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "end_time": "2023-09-29T12:01:53.250349Z",
- "start_time": "2023-09-29T12:01:53.215738Z"
+ "end_time": "2023-09-30T06:30:56.069386Z",
+ "start_time": "2023-09-30T06:30:56.021433Z"
}
},
"id": "d555d7f0223a624b"
},
{
"cell_type": "code",
- "execution_count": null,
- "outputs": [],
+ "execution_count": 17,
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "ID TOKEN LEMMA UPOS HEAD DEPREL \n",
+ "Al 6?KK>g3K>KKKKKKKKKKKKKKKKKKKKK PROPN 16 nmod:tmod \n",
+ "- ?33>K>>KKKKKKKKKKKKKKKKKKKKKKK PROPN 16 nmod:tmod \n",
+ "Zaman E3633?3>KKKKKKKKKKKKKKKKKKKKKK PROPN 28 nmod:tmod \n",
+ ": 3333K>>>KKKKKKKKKKKKKKKKKKKKKK PROPN 28 nmod:tmod \n",
+ "American {{{6K63g3uif3KKKKKKKKKKKKKKKKK PROPN 28 nmod:tmod \n",
+ "forces lK3K333gu>3>>KKKKKKKKKKKKKKKiK PROPN 29 nmod:tmod \n",
+ "killed uEE6K>>uuKK>KKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "Shaikh >EE3336u3K>KKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "Abdullah 6K>l63K3>ui>3>KKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "al E?KK>33>>KKKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "- ?33>K>>KKKKKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "Ani u6?6u3>3KKKKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ ", ?333K3>>KKKKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "the 3K?u36>3K>KKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "preacher 3>K3363u3>u>3>>KKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "at 3?K6>>3>KKKKKKKKKKKKKKKKKKKKiK ADJ 29 nmod:tmod \n",
+ "the 3K?u36>3K>KKKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "mosque |3u>K3336>3K>KKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "in 3?K3>33KKKKKKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "the 3K?u36>3K>KKKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "town Ku3>3|>f3KKKKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "of 3?Ku>>>KKKKKKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "Qaim u3333u63KKKKKKKKKKKKKKKKKKKKiK SYM 29 expl \n",
+ ", ?333K3>>KKKKKKKKKKKKKKKKKKKKiK SYM 29 expl \n",
+ "near u6KuKu33K>>KKKKKKKKKKKKKKKKKiK SYM 29 expl \n",
+ "the 3K?u36>3K>KKKKKKKKKKKKKKKKKKiK SYM 29 expl \n",
+ "Syrian ui3633Ku>f3KKKKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ "border i33>33u33>u>>KKKKKKKKKKKKKKKiK ADJ 29 expl \n",
+ ". ??3>>>KKKKKKKKKKKKKKKKKKKKKKiK ADJ 0 root \n"
+ ]
+ }
+ ],
"source": [
"print(\"{:5} {:15} {:15} {:10} {:10} {:10}\".format('ID', 'TOKEN', 'LEMMA', 'UPOS', 'HEAD', 'DEPREL'))\n",
"for token in a.tokens:\n",
@@ -489,20 +526,22 @@
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "start_time": "2023-09-29T11:59:57.890328Z"
+ "end_time": "2023-09-30T06:31:15.058040Z",
+ "start_time": "2023-09-30T06:31:14.723267Z"
}
},
"id": "a68cd3861e1ceb67"
},
{
"cell_type": "code",
- "execution_count": null,
+ "execution_count": 13,
"outputs": [],
"source": [],
"metadata": {
"collapsed": false,
"ExecuteTime": {
- "start_time": "2023-09-29T11:59:57.890649Z"
+ "end_time": "2023-09-30T06:30:56.070718Z",
+ "start_time": "2023-09-30T06:30:56.038704Z"
}
},
"id": "7e1a6f47f3aa54b8"
diff --git a/combo/predict.py b/combo/predict.py
index 70b8bd36e09401cee989f401e08bf28446ee11ad..98afe53055db045775e2433400c0371a402065dc 100644
--- a/combo/predict.py
+++ b/combo/predict.py
@@ -13,14 +13,14 @@ from combo.data import tokenizers, Instance, conllu2sentence, tokens2conllu, sen
from combo.data.dataset_readers.dataset_reader import DatasetReader
from combo.data.instance import JsonDict
-from combo.predictors import PredictorModel
+from combo.predictors import PredictorModule
from combo.utils import download, graph
logger = logging.getLogger(__name__)
-@Registry.register(PredictorModel, 'combo')
-class COMBO(PredictorModel):
+@Registry.register(PredictorModule, 'combo')
+class COMBO(PredictorModule):
def __init__(self,
model: models.Model,
@@ -183,7 +183,6 @@ class COMBO(PredictorModel):
field_value = "|".join(np.array(features)[arg_indices].tolist())
setattr(token, field_name, field_value)
- #token[field_name] = field_value
embeddings[token["idx"]][field_name] = predictions[f"{field_name}_token_embedding"][idx]
elif field_name == "lemma":
prediction = field_predictions[idx]
@@ -200,7 +199,6 @@ class COMBO(PredictorModel):
word_chars.append(pred_char)
setattr(token, field_name, "".join(word_chars))
- #token[field_name] = "".join(word_chars)
else:
raise NotImplementedError(f"Unknown field name {field_name}!")
diff --git a/combo/predictors/__init__.py b/combo/predictors/__init__.py
index a49f08e501b26571e664f564e8248437354b0b78..967701a342d41c61f9162881de6ebf9c224a09f6 100644
--- a/combo/predictors/__init__.py
+++ b/combo/predictors/__init__.py
@@ -1,2 +1,2 @@
from .predictor import Predictor
-from .predictor_model import PredictorModel
+from .predictor_model import PredictorModule
diff --git a/combo/predictors/predictor_model.py b/combo/predictors/predictor_model.py
index 2247ecac3e9769c79755eab080d4d8d1b56d38c5..09604208fb7bb2b480216554f0733f8e8fd68324 100644
--- a/combo/predictors/predictor_model.py
+++ b/combo/predictors/predictor_model.py
@@ -15,6 +15,8 @@ from torch.utils.hooks import RemovableHandle
from torch import Tensor
from torch import backends
+import pytorch_lightning as pl
+
from combo.common.util import sanitize
from combo.config import FromParameters
from combo.data.batch import Batch
@@ -26,13 +28,14 @@ from combo.nn import util
logger = logging.getLogger(__name__)
-class PredictorModel(FromParameters):
+class PredictorModule(pl.LightningModule, FromParameters):
"""
a `Predictor` is a thin wrapper around an AllenNLP model that handles JSON -> JSON predictions
that can be used for serving models through the web API or making predictions in bulk.
"""
def __init__(self, model: Model, dataset_reader: DatasetReader, frozen: bool = True) -> None:
+ super().__init__()
if frozen:
model.eval()
self._model = model