Automatic push from sapienzanlp
Browse files- cfg.yaml +48 -0
- model.ckpt +3 -0
cfg.yaml
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model_name: relik-reader-deberta-small
|
| 2 |
+
project_name: relik-reader
|
| 3 |
+
training:
|
| 4 |
+
seed: 94
|
| 5 |
+
trainer:
|
| 6 |
+
_target_: lightning.Trainer
|
| 7 |
+
devices:
|
| 8 |
+
- 0
|
| 9 |
+
precision: 16-mixed
|
| 10 |
+
max_steps: 50000
|
| 11 |
+
val_check_interval: 1.0
|
| 12 |
+
num_sanity_val_steps: 0
|
| 13 |
+
limit_val_batches: 1
|
| 14 |
+
gradient_clip_val: 1.0
|
| 15 |
+
model:
|
| 16 |
+
model:
|
| 17 |
+
transformer_model: microsoft/deberta-v3-small
|
| 18 |
+
optimizer:
|
| 19 |
+
lr: 0.0001
|
| 20 |
+
warmup_steps: 5000
|
| 21 |
+
total_steps: ${training.trainer.max_steps}
|
| 22 |
+
total_reset: 1
|
| 23 |
+
weight_decay: 0.0
|
| 24 |
+
lr_decay: 0.8
|
| 25 |
+
no_decay_params:
|
| 26 |
+
- bias
|
| 27 |
+
- LayerNorm.weight
|
| 28 |
+
entities_per_forward: 100
|
| 29 |
+
data:
|
| 30 |
+
train_dataset_path: relik/reader/data/train.jsonl
|
| 31 |
+
val_dataset_path: relik/reader/data/testa.jsonl
|
| 32 |
+
train_dataset:
|
| 33 |
+
_target_: relik.reader.relik_reader_data.RelikDataset
|
| 34 |
+
transformer_model: ${model.model.transformer_model}
|
| 35 |
+
materialize_samples: false
|
| 36 |
+
shuffle_candidates: 0.5
|
| 37 |
+
random_drop_gold_candidates: 0.05
|
| 38 |
+
noise_param: 0.0
|
| 39 |
+
for_inference: false
|
| 40 |
+
tokens_per_batch: 4096
|
| 41 |
+
special_symbols: null
|
| 42 |
+
val_dataset:
|
| 43 |
+
_target_: relik.reader.relik_reader_data.RelikDataset
|
| 44 |
+
transformer_model: ${model.model.transformer_model}
|
| 45 |
+
materialize_samples: false
|
| 46 |
+
shuffle_candidates: false
|
| 47 |
+
for_inference: true
|
| 48 |
+
special_symbols: null
|
model.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7469ffb06dde1de88140041e8f16762d4645ae8bb67851ddcb06eaa7d02263c
|
| 3 |
+
size 577168073
|