scandinavian 3k xl
Browse files
scandinavian3k_solo_large.gin
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
include 't5x/examples/t5/mt5/large.gin'
|
| 2 |
+
include 'pretrain_cont.gin'
|
| 3 |
+
#include 't5x/configs/runs/pretrain.gin'
|
| 4 |
+
#iinclude 't5x/configs/runs/finetune.gin'
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
# Register necessary SeqIO Tasks/Mixtures.
|
| 8 |
+
import t5.data.mixtures
|
| 9 |
+
import tasks
|
| 10 |
+
|
| 11 |
+
MIXTURE_OR_TASK_NAME = "ncc_scandinavian_span_corruption_stream"
|
| 12 |
+
TASK_FEATURE_LENGTHS = {"inputs": 512, "targets": 512}
|
| 13 |
+
TRAIN_STEPS = 3_000_000
|
| 14 |
+
DROPOUT_RATE = 0.0 # Changed from the default since T5-1.1 recomments this.
|
| 15 |
+
INITIAL_CHECKPOINT_PATH = "gs://t5-data/pretrained_models/t5x/mt5_large/checkpoint_1000000"
|
| 16 |
+
PjitPartitioner.num_partitions = 2
|
| 17 |
+
|
| 18 |
+
|
scandinavian3k_solo_xl.gin
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
include 't5x/examples/t5/mt5/xl.gin'
|
| 2 |
+
include 'pretrain_cont.gin'
|
| 3 |
+
#include 't5x/configs/runs/pretrain.gin'
|
| 4 |
+
#iinclude 't5x/configs/runs/finetune.gin'
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
# Register necessary SeqIO Tasks/Mixtures.
|
| 8 |
+
import t5.data.mixtures
|
| 9 |
+
import tasks
|
| 10 |
+
|
| 11 |
+
MIXTURE_OR_TASK_NAME = "ncc_scandinavian_span_corruption_stream"
|
| 12 |
+
TASK_FEATURE_LENGTHS = {"inputs": 512, "targets": 512}
|
| 13 |
+
TRAIN_STEPS = 3_000_000
|
| 14 |
+
DROPOUT_RATE = 0.0 # Changed from the default since T5-1.1 recomments this.
|
| 15 |
+
INITIAL_CHECKPOINT_PATH = "gs://t5-data/pretrained_models/t5x/mt5_large/checkpoint_1000000"
|
| 16 |
+
PjitPartitioner.num_partitions = 2
|
| 17 |
+
|
| 18 |
+
|
train_large_scandinavian3k_solo.sh
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
PROJECT_DIR=${HOME}"/models/pk-nb-t5x"
|
| 2 |
+
T5X_DIR="../../t5x" # directory where the t5x is cloned.
|
| 3 |
+
MODEL_DIR="gs://nb-t5x-us-central2/scandinavian3k_t5x_large"
|
| 4 |
+
export PYTHONPATH=${PROJECT_DIR}
|
| 5 |
+
|
| 6 |
+
python3 ${T5X_DIR}/t5x/train.py \
|
| 7 |
+
--gin_search_paths=${PROJECT_DIR} \
|
| 8 |
+
--gin_file="scandinavian3k_solo_large.gin" \
|
| 9 |
+
--gin.MODEL_DIR="'${MODEL_DIR}'" \
|
train_xl_scandinavian3k_solo.sh
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
PROJECT_DIR=${HOME}"/models/pk-nb-t5x"
|
| 2 |
+
T5X_DIR="../../t5x" # directory where the t5x is cloned.
|
| 3 |
+
MODEL_DIR="gs://nb-t5x-us-central2/scandinavian3k_t5x_xl"
|
| 4 |
+
export PYTHONPATH=${PROJECT_DIR}
|
| 5 |
+
|
| 6 |
+
python3 ${T5X_DIR}/t5x/train.py \
|
| 7 |
+
--gin_search_paths=${PROJECT_DIR} \
|
| 8 |
+
--gin_file="scandinavian3k_solo_xl.gin" \
|
| 9 |
+
--gin.MODEL_DIR="'${MODEL_DIR}'" \
|