Commit from model create scripts

.gitattributes

@@ -25,3 +25,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

config.gin

@@ -0,0 +1,151 @@
+from __gin__ import dynamic_registration
+import __main__ as train_script
+import seqio
+import t5.data.mixtures
+from t5x import adafactor
+from t5x.examples.t5 import network
+from t5x import gin_utils
+from t5x import models
+from t5x import partitioning
+from t5x import trainer
+from t5x import utils
+import tasks
+
+# Macros:
+# ==============================================================================
+BATCH_SIZE = 128
+DROPOUT_RATE = 0.0
+INITIAL_CHECKPOINT_PATH = \
+    'gs://t5-data/pretrained_models/t5x/mt5_large/checkpoint_1000000'
+LABEL_SMOOTHING = 0.0
+LOSS_NORMALIZING_FACTOR = None
+MIXTURE_OR_TASK_MODULE = None
+MIXTURE_OR_TASK_NAME = 'ncc_english_span_corruption_stream'
+MODEL = @models.EncoderDecoderModel()
+MODEL_DIR = 'gs://nb-t5x-us-central2/norwegian_NCC_plus_English_t5x_large'
+OPTIMIZER = @adafactor.Adafactor()
+RANDOM_SEED = None
+SHUFFLE_TRAIN_EXAMPLES = True
+TASK_FEATURE_LENGTHS = {'inputs': 512, 'targets': 512}
+TRAIN_STEPS = 1500000
+USE_CACHED_TASKS = True
+USE_HARDWARE_RNG = False
+VOCABULARY = @seqio.SentencePieceVocabulary()
+Z_LOSS = 0.0001
+
+# Parameters for adafactor.Adafactor:
+# ==============================================================================
+adafactor.Adafactor.decay_rate = 0.8
+adafactor.Adafactor.logical_factor_rules = \
+    @adafactor.standard_logical_factor_rules()
+adafactor.Adafactor.step_offset = 0
+
+# Parameters for utils.CheckpointConfig:
+# ==============================================================================
+utils.CheckpointConfig.restore = @utils.RestoreCheckpointConfig()
+utils.CheckpointConfig.save = @utils.SaveCheckpointConfig()
+
+# Parameters for utils.create_learning_rate_scheduler:
+# ==============================================================================
+utils.create_learning_rate_scheduler.base_learning_rate = 0.5
+utils.create_learning_rate_scheduler.factors = 'constant * rsqrt_decay'
+utils.create_learning_rate_scheduler.warmup_steps = 10000
+
+# Parameters for train/utils.DatasetConfig:
+# ==============================================================================
+train/utils.DatasetConfig.batch_size = %BATCH_SIZE
+train/utils.DatasetConfig.mixture_or_task_name = %MIXTURE_OR_TASK_NAME
+train/utils.DatasetConfig.module = %MIXTURE_OR_TASK_MODULE
+train/utils.DatasetConfig.pack = True
+train/utils.DatasetConfig.seed = None
+train/utils.DatasetConfig.shuffle = %SHUFFLE_TRAIN_EXAMPLES
+train/utils.DatasetConfig.split = 'train'
+train/utils.DatasetConfig.task_feature_lengths = %TASK_FEATURE_LENGTHS
+train/utils.DatasetConfig.use_cached = %USE_CACHED_TASKS
+
+# Parameters for train_eval/utils.DatasetConfig:
+# ==============================================================================
+train_eval/utils.DatasetConfig.batch_size = %BATCH_SIZE
+train_eval/utils.DatasetConfig.mixture_or_task_name = %MIXTURE_OR_TASK_NAME
+train_eval/utils.DatasetConfig.module = %MIXTURE_OR_TASK_MODULE
+train_eval/utils.DatasetConfig.pack = True
+train_eval/utils.DatasetConfig.seed = 42
+train_eval/utils.DatasetConfig.shuffle = False
+train_eval/utils.DatasetConfig.split = 'validation'
+train_eval/utils.DatasetConfig.task_feature_lengths = %TASK_FEATURE_LENGTHS
+train_eval/utils.DatasetConfig.use_cached = %USE_CACHED_TASKS
+
+# Parameters for models.EncoderDecoderModel:
+# ==============================================================================
+models.EncoderDecoderModel.input_vocabulary = %VOCABULARY
+models.EncoderDecoderModel.label_smoothing = %LABEL_SMOOTHING
+models.EncoderDecoderModel.loss_normalizing_factor = %LOSS_NORMALIZING_FACTOR
+models.EncoderDecoderModel.module = @network.Transformer()
+models.EncoderDecoderModel.optimizer_def = %OPTIMIZER
+models.EncoderDecoderModel.output_vocabulary = %VOCABULARY
+models.EncoderDecoderModel.z_loss = %Z_LOSS
+
+# Parameters for partitioning.PjitPartitioner:
+# ==============================================================================
+partitioning.PjitPartitioner.logical_axis_rules = \
+    @partitioning.standard_logical_axis_rules()
+partitioning.PjitPartitioner.model_parallel_submesh = None
+partitioning.PjitPartitioner.num_partitions = 1
+
+# Parameters for utils.RestoreCheckpointConfig:
+# ==============================================================================
+utils.RestoreCheckpointConfig.dtype = 'float32'
+utils.RestoreCheckpointConfig.mode = 'specific'
+utils.RestoreCheckpointConfig.path = %INITIAL_CHECKPOINT_PATH
+
+# Parameters for utils.SaveCheckpointConfig:
+# ==============================================================================
+utils.SaveCheckpointConfig.dtype = 'float32'
+utils.SaveCheckpointConfig.keep = None
+utils.SaveCheckpointConfig.period = 1000
+utils.SaveCheckpointConfig.save_dataset = False
+
+# Parameters for seqio.SentencePieceVocabulary:
+# ==============================================================================
+seqio.SentencePieceVocabulary.sentencepiece_model_file = \
+    'gs://t5-data/vocabs/mc4.250000.100extra/sentencepiece.model'
+
+# Parameters for network.T5Config:
+# ==============================================================================
+network.T5Config.dropout_rate = %DROPOUT_RATE
+network.T5Config.dtype = 'bfloat16'
+network.T5Config.emb_dim = 1024
+network.T5Config.head_dim = 64
+network.T5Config.logits_via_embedding = False
+network.T5Config.mlp_activations = ('gelu', 'linear')
+network.T5Config.mlp_dim = 2816
+network.T5Config.num_decoder_layers = 24
+network.T5Config.num_encoder_layers = 24
+network.T5Config.num_heads = 16
+network.T5Config.vocab_size = 250112
+
+# Parameters for train_script.train:
+# ==============================================================================
+train_script.train.checkpoint_cfg = @utils.CheckpointConfig()
+train_script.train.eval_period = 1000
+train_script.train.eval_steps = 20
+train_script.train.infer_eval_dataset_cfg = None
+train_script.train.model = %MODEL
+train_script.train.model_dir = %MODEL_DIR
+train_script.train.partitioner = @partitioning.PjitPartitioner()
+train_script.train.random_seed = %RANDOM_SEED
+train_script.train.summarize_config_fn = @gin_utils.summarize_gin_config
+train_script.train.total_steps = %TRAIN_STEPS
+train_script.train.train_dataset_cfg = @train/utils.DatasetConfig()
+train_script.train.train_eval_dataset_cfg = @train_eval/utils.DatasetConfig()
+train_script.train.trainer_cls = @trainer.Trainer
+train_script.train.use_hardware_rng = %USE_HARDWARE_RNG
+
+# Parameters for trainer.Trainer:
+# ==============================================================================
+trainer.Trainer.learning_rate_fn = @utils.create_learning_rate_scheduler()
+trainer.Trainer.num_microbatches = None
+
+# Parameters for network.Transformer:
+# ==============================================================================
+network.Transformer.config = @network.T5Config()

config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "/home/patrick/hugging_face/t5/mt5-large",
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "d_ff": 2816,
+  "d_kv": 64,
+  "d_model": 1024,
+  "decoder_start_token_id": 0,
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "gated-gelu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "num_decoder_layers": 24,
+  "num_heads": 16,
+  "num_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "tie_word_embeddings": false,
+  "tokenizer_class": "T5Tokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.19.2",
+  "use_cache": true,
+  "vocab_size": 250112
+}

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aad7ce9b9ee79caa0ee4265198e00e4d4d828857c1ad7e5b868ae9f5efede2a3
+size 4918349339

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a84f8ab586000eec0208a03d99db5ec9bb66ce522a76ceb3e1f2dd9826ca7f8
+size 4918507641

special_tokens_map.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef78f86560d809067d12bac6c09f19a462cb3af3f54d2b8acbba26e1433125d6
+size 4309802

tf_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13edf9346263c231557abbd4b42fd847571a957f938e4f48b2d23ec2f1c58acc
+size 9760

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93c3578052e1605d8332eb961bc08d72e246071974e4cc54aa6991826b802aa5
+size 16330369

tokenizer_config.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "extra_ids": 0, "additional_special_tokens": null, "special_tokens_map_file": "/home/patrick/.cache/torch/transformers/685ac0ca8568ec593a48b61b0a3c272beee9bc194a3c7241d15dcadb5f875e53.f76030f3ec1b96a8199b2593390c610e76ca8028ef3d24680000619ffb646276", "name_or_path": "/home/perk/models/t5_large_NCC", "sp_model_kwargs": {}, "tokenizer_class": "T5Tokenizer"}

train/events.out.tfevents.1649427185.t1v-n-6662dcba-w-0.112619.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e44d2410fc3d7223b54fa071d5a78a88d7934f0e7ab1746c11a987c1d9f828f7
+size 255034

train/events.out.tfevents.1649573818.t1v-n-6662dcba-w-0.1061176.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63c7f6048300ecbfda83c2954a57fac7da56130e68d2c95f45d19a9455bbe16c
+size 26806

train/events.out.tfevents.1649586888.t1v-n-6662dcba-w-0.1144436.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6c2dff801911010176642606bdc2f10c4422c1e15a146190b06b9de0f4a6ffe
+size 35584

train/events.out.tfevents.1649602337.t1v-n-6662dcba-w-0.1250589.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:436a2f4bc1037a4c3895a3c67079c8f6463e0aff346514cd89e55f4d5774d120
+size 6409

train/events.out.tfevents.1649607746.t1v-n-6662dcba-w-0.1274684.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ed5f5cce204c20d3a876f8e46aeb57501ceb3c047c41033c5ff236ff47664d8
+size 41436

train/events.out.tfevents.1649658060.t1v-n-6662dcba-w-0.1524215.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b91ba5b6892625d11e079a9a50d7d4cf30de61b28fd4b5e5472ffe192968dc0
+size 6409

train/events.out.tfevents.1649659807.t1v-n-6662dcba-w-0.1534682.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df5bb511ba408405803f5ad6e5451bcfdcd8c0b0001aad4de4babdff9975ae0b
+size 21039

train/events.out.tfevents.1649677590.t1v-n-6662dcba-w-0.40663.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32e75200f32866b44faf40456fe3956fcb7b517ab1524bba4d0aa5f10c28c45f
+size 37047

train/events.out.tfevents.1649694013.t1v-n-6662dcba-w-0.152982.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a570ff851865295c4a2f626a1c9734eb717cf1cb3636ffaa658e7c6db314c502
+size 44362

train/events.out.tfevents.1649743837.t1v-n-6662dcba-w-0.403337.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44c01c528c2ed9a89ac41db8d4badd63ff11f2fa137b5dafcbca95b794993564
+size 265275

train/events.out.tfevents.1649873179.t1v-n-6662dcba-w-0.1303239.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc452a418ba02f5057c18e819a8879c4f8f1997f04aa69d2dfd3526c80f51e04
+size 26806

train/events.out.tfevents.1649910184.t1v-n-6662dcba-w-0.1478589.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62d98850eb2b7d574afb3166d0b637a8faba16dd17e3715cb60c44bb9c99da3f
+size 12176

train/events.out.tfevents.1649914809.t1v-n-6662dcba-w-0.1507325.0.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aede02cfd4d02ca6566742ec97767e03d11a3dca15ef8ca59ab7e941ec86a25
+size 48836

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649427186.t1v-n-6662dcba-w-0.112619.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:719ed7b45a3697bf427dd4b39ddf6123aa8bca724689aff2d8afa3080e365569
+size 220671

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649573818.t1v-n-6662dcba-w-0.1061176.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:512c1fca8168a6ba3754e22579967b894e8cbab9d8547e8256583bae68d90d52
+size 17091

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649586888.t1v-n-6662dcba-w-0.1144436.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8849e4c9f484ca7c567291267c64a88f0a983280f164e1d8608704c43f7d08f
+size 24921

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649602337.t1v-n-6662dcba-w-0.1250589.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:349bda43a9f10beeec9ab84414e77b964e4fb1acf92603d2074701909b2b21c2
+size 40

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649607746.t1v-n-6662dcba-w-0.1274684.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc2e5d7675b77393beb3dc3b3c33b9b56fd2e4c6d22a5caeb93bbfb805798720
+size 30141

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649658060.t1v-n-6662dcba-w-0.1524215.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b430662f465b094605589f87d3e3c6b3202e1fe59c1460abbe65b56902b44e94
+size 40

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649659808.t1v-n-6662dcba-w-0.1534682.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c5fb18061e460fcfea20c4acfcb8c4157b68ff9bf177f4d860c3c8fc08b5f6b
+size 13176

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649677590.t1v-n-6662dcba-w-0.40663.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a44ded4b8032a7734bff803695d1e5cf5c075f9adc23ad0e4926a0094c389fa9
+size 26226

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649694013.t1v-n-6662dcba-w-0.152982.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca40bc02662b1b98648a008b062fe5d48ccb90a1165a226e6f5136a0942bcb2c
+size 32751

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649743838.t1v-n-6662dcba-w-0.403337.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd6ff6d52419fa915bc67d930a3aaecc923c1893a867710a556e7ae0241f7773
+size 229806

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649873180.t1v-n-6662dcba-w-0.1303239.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0aa9aa18d41d9367389e2aefdc4a538787a39c9f07d86e9da6730b3a9b677a7
+size 17091

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649910185.t1v-n-6662dcba-w-0.1478589.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0cb64e34a330c8b4bc2cc13d1b38518c405778bb9eb57ab0689cf3271791a0
+size 4041

training_eval/ncc_english_span_corruption_stream/events.out.tfevents.1649914810.t1v-n-6662dcba-w-0.1507325.1.v2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cc99a368dc89e4cb58a90e9122fd4e5de78b37af47fb4223bfc513a8cdddc59
+size 37971