fixed size of inference

eval_categorisation_base.gin

@@ -23,7 +23,7 @@ eval_script.evaluate:
 
 utils.DatasetConfig:
   mixture_or_task_name = %MIXTURE_OR_TASK_NAME
-  task_feature_lengths = None  # Auto-computes the max feature lengths.
+  task_feature_lengths = {"inputs": 512, "targets": 2}
   split = 'validation'
   batch_size = 32
   shuffle = False

finetune_categorisation_base.gin

@@ -12,11 +12,15 @@ include "t5x/configs/runs/finetune.gin"
 
 MIXTURE_OR_TASK_NAME = "categorise"
 TASK_FEATURE_LENGTHS = {"inputs": 512, "targets": 2}
-TRAIN_STEPS = 2_005_000  # 1000000 pre-trained steps + 10000 fine-tuning steps.
+TRAIN_STEPS = 1_510_000  # 1000000 pre-trained steps + 10000 fine-tuning steps.
 USE_CACHED_TASKS = False
 DROPOUT_RATE = 0.1
 RANDOM_SEED = 0
 
+#Fixing a small error
+infer_eval/utils.DatasetConfig.task_feature_lengths = TASK_FEATURE_LENGTHS
+
+
 # Pere: Only necessary if we load a t5 model. We can start with an t5x model here
 # `LOSS_NORMALIZING_FACTOR`: When fine-tuning a model that was pre-trained
 # using Mesh Tensorflow (e.g. the public T5 / mT5 / ByT5 models), this should be
@@ -29,7 +33,10 @@ RANDOM_SEED = 0
 #INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/norwegian_t5x_base/checkpoint_1360000"
 #INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/pk_nb_t5x_base_run1_lr_1/checkpoint_1100000"
 #INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/pk_nb_t5x_base_scandinavian/checkpoint_1100000"
-INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/norwegian_t5x_base/checkpoint_2000000"
+#INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/norwegian_t5x_base/checkpoint_2000000"
+INITIAL_CHECKPOINT_PATH = "gs://nb-t5x-us-central2/norwegian_NCC_plus_English_t5x_base/checkpoint_1500000"
+
+
 
 #train_script.train:
 #  eval_period = 500

train_base.sh

@@ -1,6 +1,7 @@
 PROJECT_DIR=${HOME}"/models/t5-parliament-categorisation"
 T5X_DIR="../../t5x"  # directory where the t5x is cloned.
-MODEL_DIR="gs://nb-t5x-us-central2/eval2_norwegian_NCC_2_000_000"
+#Needs to be updated when moving to tpu-v4  it should then be in another zone
+MODEL_DIR="gs://nb-t5x-us-central2/finetuned/norwegian_NCC_pluss_english_1_500_000"
 export PYTHONPATH=${PROJECT_DIR}
 
 python3 ${T5X_DIR}/t5x/train.py \