save more often...

batch_finetune_eu_jav_base.sh

@@ -1,7 +1,7 @@
 PROJECT_DIR=${HOME}"/models/eu-jav-categorisation"
 export PYTHONPATH=${PROJECT_DIR}
 INITIAL_CHECKPOINT_PATH=\"gs://t5-data/pretrained_models/t5x/mt5_base/checkpoint_1000000\"
-TRAIN_STEPS=1005000
+TRAIN_STEPS=1001000
 
 python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_base.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_base_v1\" &&
 python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_base.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_base_v2\" &&

batch_finetune_eu_jav_large.sh

@@ -1,7 +1,7 @@
 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=1005000
+TRAIN_STEPS=1001000
 
 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\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_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\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_large_v2\" &&

batch_finetune_eu_jav_small.sh

@@ -1,7 +1,7 @@
 PROJECT_DIR=${HOME}"/models/eu-jav-categorisation"
 export PYTHONPATH=${PROJECT_DIR}
 INITIAL_CHECKPOINT_PATH=\"gs://t5-data/pretrained_models/t5x/mt5_small/checkpoint_1000000\"
-TRAIN_STEPS=1005000
+TRAIN_STEPS=1001000
 
 python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_small.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_small_v1\" &&
 python3 ../../t5x/t5x/train.py --gin_search_paths="./" --gin.TRAIN_STEPS=${TRAIN_STEPS} --gin_file="finetune_classification_small.gin" --gin.INITIAL_CHECKPOINT_PATH=${INITIAL_CHECKPOINT_PATH}  --gin.MIXTURE_OR_TASK_NAME=\"classify_tweets\" --gin.MODEL_DIR=\"gs://eu-jav-t5x/finetuned/italian_tweets/classify_tweets_small_v2\" &&

finetune_classification_base.gin

@@ -24,7 +24,7 @@ infer_eval/utils.DatasetConfig:
 
 #Saving every 1000 steps
 utils.SaveCheckpointConfig:
-  period = 1000
+  period = 100
 
 
 # Pere: Only necessary if we load a t5 model. We can start with an t5x model here

finetune_classification_large.gin

@@ -25,7 +25,7 @@ infer_eval/utils.DatasetConfig:
 
 #Saving every 1000 steps
 utils.SaveCheckpointConfig:
-  period = 1000
+  period = 50
 
 
 # Pere: Only necessary if we load a t5 model. We can start with an t5x model here

finetune_classification_small.gin

@@ -24,7 +24,7 @@ infer_eval/utils.DatasetConfig:
 
 #Saving every 1000 steps
 utils.SaveCheckpointConfig:
-  period = 1000
+  period = 50
 
 
 # Pere: Only necessary if we load a t5 model. We can start with an t5x model here

tasks.py

@@ -60,7 +60,7 @@ seqio.TaskRegistry.add(
       categorise_preprocessor,
       seqio.preprocessors.tokenize_and_append_eos,
     ],
-    metric_fns=[metrics.accuracy,my_metrics.f1_macro,metrics.bleu],
+    metric_fns=[metrics.accuracy,my_metrics.f1_macro],
     output_features=DEFAULT_OUTPUT_FEATURES,
 )