init
Browse files
experiments/model_finetuning_topic.py
CHANGED
@@ -76,9 +76,6 @@ def main(
|
|
76 |
best_model_path = pj(output_dir, "best_model")
|
77 |
|
78 |
tokenizer = AutoTokenizer.from_pretrained(model)
|
79 |
-
model = AutoModelForSequenceClassification.from_pretrained(
|
80 |
-
model, id2label=ID2LABEL, label2id=LABEL2ID, num_labels=len(LABEL2ID), problem_type="multi_label_classification"
|
81 |
-
)
|
82 |
dataset = load_dataset(dataset, dataset_type)
|
83 |
tokenized_datasets = dataset.map(
|
84 |
lambda x: tokenizer(x["text"], padding="max_length", truncation=True, max_length=256), batched=True
|
@@ -150,15 +147,14 @@ def main(
|
|
150 |
else:
|
151 |
metric = {}
|
152 |
for single_test in test_split:
|
153 |
-
model = AutoModelForSequenceClassification.from_pretrained(
|
154 |
-
best_model_path,
|
155 |
-
num_labels=len(LABEL2ID),
|
156 |
-
problem_type="multi_label_classification",
|
157 |
-
id2label=ID2LABEL,
|
158 |
-
label2id=LABEL2ID
|
159 |
-
)
|
160 |
trainer = Trainer(
|
161 |
-
model=
|
|
|
|
|
|
|
|
|
|
|
|
|
162 |
args=TrainingArguments(
|
163 |
output_dir=output_dir,
|
164 |
evaluation_strategy="no",
|
@@ -176,7 +172,7 @@ def main(
|
|
176 |
if not skip_upload:
|
177 |
logging.info("uploading to huggingface")
|
178 |
model_organization = "tweettemposhift"
|
179 |
-
|
180 |
best_model_path,
|
181 |
num_labels=len(LABEL2ID),
|
182 |
problem_type="multi_label_classification",
|
@@ -184,7 +180,7 @@ def main(
|
|
184 |
label2id=LABEL2ID
|
185 |
)
|
186 |
tokenizer = AutoTokenizer.from_pretrained(best_model_path)
|
187 |
-
|
188 |
tokenizer.push_to_hub(f"{model_organization}/{model_alias}", use_auth_token=True)
|
189 |
repo = Repository(model_alias, f"{model_organization}/{model_alias}")
|
190 |
for i in glob(f"{best_model_path}/*"):
|
|
|
76 |
best_model_path = pj(output_dir, "best_model")
|
77 |
|
78 |
tokenizer = AutoTokenizer.from_pretrained(model)
|
|
|
|
|
|
|
79 |
dataset = load_dataset(dataset, dataset_type)
|
80 |
tokenized_datasets = dataset.map(
|
81 |
lambda x: tokenizer(x["text"], padding="max_length", truncation=True, max_length=256), batched=True
|
|
|
147 |
else:
|
148 |
metric = {}
|
149 |
for single_test in test_split:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
150 |
trainer = Trainer(
|
151 |
+
model=AutoModelForSequenceClassification.from_pretrained(
|
152 |
+
best_model_path,
|
153 |
+
num_labels=len(LABEL2ID),
|
154 |
+
problem_type="multi_label_classification",
|
155 |
+
id2label=ID2LABEL,
|
156 |
+
label2id=LABEL2ID
|
157 |
+
),
|
158 |
args=TrainingArguments(
|
159 |
output_dir=output_dir,
|
160 |
evaluation_strategy="no",
|
|
|
172 |
if not skip_upload:
|
173 |
logging.info("uploading to huggingface")
|
174 |
model_organization = "tweettemposhift"
|
175 |
+
model_instance = AutoModelForSequenceClassification.from_pretrained(
|
176 |
best_model_path,
|
177 |
num_labels=len(LABEL2ID),
|
178 |
problem_type="multi_label_classification",
|
|
|
180 |
label2id=LABEL2ID
|
181 |
)
|
182 |
tokenizer = AutoTokenizer.from_pretrained(best_model_path)
|
183 |
+
model_instance.push_to_hub(f"{model_organization}/{model_alias}", use_auth_token=True)
|
184 |
tokenizer.push_to_hub(f"{model_organization}/{model_alias}", use_auth_token=True)
|
185 |
repo = Repository(model_alias, f"{model_organization}/{model_alias}")
|
186 |
for i in glob(f"{best_model_path}/*"):
|