# log-anomaly-detection

**Repository Path**: ezi_admin/log-anomaly-detection

## Basic Information

- **Project Name**: log-anomaly-detection
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-04-15
- **Last Updated**: 2025-04-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# log-anomaly-detection



## Test Evaluation:

**result #1**
- BERT params: set #1
- Log params: set #1
- LogAnomaly params: set #1
```shell
[('session_acc', 0.76193), ('f1', 0.11699), ('recall', 0.29118), ('precision', 0.0732)]
```

## Training Experiments

### BERT params:

**set #1**
```python
"attention_probs_dropout_prob": 0.1,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_position_embeddings": 514,
"model_type": "bert",
"num_attention_heads": 12,
"num_hidden_layers": 6,
"pad_token_id": 0,
"type_vocab_size": 1,
"vocab_size": 52000

```

### Log params:

**set #1**
```python
time_interval = 60
window_size = 500
stepping_size = 50
```

### LogAnomaly params:

**set #1**
```python
batch_size = 2
hidden_size = 256
epoches = 8
```


---


## Runtime Log

### BERT (based on full BGL log)

```shell
...
'loss': 0.6645126495896838, 'learning_rate': 7.797171700713524e-07, 'epoch': 7.8752452527885834, 'step': 1160000}
Saving model checkpoint to ./bgl/checkpoint-1160000
Configuration saved in ./bgl/checkpoint-1160000/config.json
Model weights saved in ./bgl/checkpoint-1160000/pytorch_model.bin
Deleting older checkpoint [bgl/checkpoint-1140000] due to args.save_total_limit
...
Iteration: 100%|███████████████████████████████████████████| 147297/147297 [4:01:58<00:00, 10.15it/s]
Epoch: 100%|█████████████████████████████████████████████████████| 8/8 [32:19:13<00:00, 14544.17s/it]
```