jl commited on
Commit ·
3b7817c
1
Parent(s): 244e29e
init models research purposes
Browse files- AlteredShield.pth +3 -0
- BaseShield.pth +3 -0
- alter_config.json +22 -0
- base_config.json +34 -0
AlteredShield.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abbe88b5f9a0ee32941ce9495e381c471ba5e31b88e20f1a0b7cd5d382d4ada6
|
| 3 |
+
size 899265658
|
BaseShield.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2caf179c7ae25b2c6bfae58d3210d29c4eec1791e67cc39188d96094fc5590c
|
| 3 |
+
size 879191355
|
alter_config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "ConcatModelWithRationale",
|
| 3 |
+
"hatebert_model": "GroNLP/hateBERT",
|
| 4 |
+
"rationale_model": "bert-base-uncased",
|
| 5 |
+
"cnn_num_filters": 128,
|
| 6 |
+
"cnn_kernel_sizes": [2, 3, 4],
|
| 7 |
+
"adapter_dim": 128,
|
| 8 |
+
"max_length": 128,
|
| 9 |
+
"num_labels": 2,
|
| 10 |
+
"architecture": {
|
| 11 |
+
"temporal_cnn": "TemporalCNN",
|
| 12 |
+
"msa_cnn": "MultiScaleAttentionCNN",
|
| 13 |
+
"projection": "ProjectionMLP"
|
| 14 |
+
},
|
| 15 |
+
"training_info": {
|
| 16 |
+
"batch_size": 8,
|
| 17 |
+
"learning_rate": 1e-5,
|
| 18 |
+
"weight_decay": 0.05,
|
| 19 |
+
"dropout": 0.5,
|
| 20 |
+
"best_val_loss": 0.27
|
| 21 |
+
}
|
| 22 |
+
}
|
base_config.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "BaseShield",
|
| 3 |
+
"architecture": "HateBERT + Rationale BERT with CLS Concatenation",
|
| 4 |
+
"version": "1.0.0",
|
| 5 |
+
|
| 6 |
+
"model_config": {
|
| 7 |
+
"hatebert_model": "GroNLP/hateBERT",
|
| 8 |
+
"rationale_model": "bert-base-uncased",
|
| 9 |
+
"hidden_size": 768,
|
| 10 |
+
"freeze_additional_model": true,
|
| 11 |
+
"num_labels": 2,
|
| 12 |
+
"projection_config": {
|
| 13 |
+
"input_size": 1536,
|
| 14 |
+
"hidden_size": 256,
|
| 15 |
+
"output_size": 2,
|
| 16 |
+
"dropout": 0.3
|
| 17 |
+
}
|
| 18 |
+
},
|
| 19 |
+
|
| 20 |
+
"training_config": {
|
| 21 |
+
"batch_size": 8,
|
| 22 |
+
"learning_rate": 1e-5,
|
| 23 |
+
"weight_decay": 0.0,
|
| 24 |
+
"epochs": 3,
|
| 25 |
+
"max_length": 128,
|
| 26 |
+
"optimizer": "AdamW",
|
| 27 |
+
"warmup_steps": 100,
|
| 28 |
+
"early_stopping": {
|
| 29 |
+
"enabled": true,
|
| 30 |
+
"patience": 3,
|
| 31 |
+
"metric": "loss"
|
| 32 |
+
}
|
| 33 |
+
}
|
| 34 |
+
}
|