binary

batch_finetune_eu_jav_large_binary.sh

@@ -0,0 +1,11 @@
+PROJECT_DIR=${HOME}"/models/eu-jav-categorisation"
+export PYTHONPATH=${PROJECT_DIR}
+INITIAL_CHECKPOINT_PATH=\"gs://t5-data/pretrained_models/t5x/mt5_large/checkpoint_1000000\"
+TRAIN_STEPS=1002000
+
+python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_large.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets_binary\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_binary_v1\" &&
+python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_large.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets_binary\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_largebinary__v2\" &&
+python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_large.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets_binary\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_binary_v3\" &&
+python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_large.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets_binary\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_binary_v4\" &&
+python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_large.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets_binary\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_binary_v5\"
+

tasks.py

@@ -115,11 +115,11 @@ def categorise_binary_preprocessor(ds):
 
   def fulltext(t):
     if t=="0":
-        t="1"
+        t="0"
     elif t=="1":
-        t="1"
+        t="0"
     elif t=="2":
-        t="2"
+        t="1"
     return t
 
   def to_inputs_and_targets(ex):
@@ -137,6 +137,24 @@ def categorise_binary_preprocessor(ds):
 
 
 
+seqio.TaskRegistry.add(
+    "classify_tweets_binary",
+    source=seqio.TextLineDataSource(
+        split_to_filepattern=tsv_path,
+        #num_input_examples=num_nq_examples
+        ),
+    preprocessors=[
+      functools.partial(
+          t5.data.preprocessors.parse_tsv,
+          field_names=["annotator1","annotator2","annotator3","target","source","id"]),
+      categorise_binary_preprocessor,
+      seqio.preprocessors.tokenize_and_append_eos,
+    ],
+    metric_fns=[metrics.accuracy,my_metrics.f1_macro],
+    output_features=DEFAULT_OUTPUT_FEATURES,
+)
+
+
 seqio.TaskRegistry.add(
     "classify_tweets",
     source=seqio.TextLineDataSource(