Add model documentation
Browse files
README.md
CHANGED
|
@@ -31,9 +31,9 @@ effectively predict anomalies. We will then attempt build a model that is more
|
|
| 31 |
- Base model: EleutherAI/pythia-14m
|
| 32 |
- Dataset: https://zenodo.org/records/8196385/files/HDFS_v1.zip?download=1 + preprocessed data at honicky/log-analysis-hdfs-preprocessed
|
| 33 |
- Batch size: 4
|
| 34 |
-
- Max sequence length:
|
| 35 |
- Learning rate: 0.0001
|
| 36 |
-
- Training steps:
|
| 37 |
|
| 38 |
## Special Tokens
|
| 39 |
- Added `<|sep|>` token for event ID separation
|
|
@@ -46,6 +46,6 @@ This model is intended for:
|
|
| 46 |
|
| 47 |
## Limitations
|
| 48 |
- Model is specifically trained on HDFS logs and may not generalize to other log formats
|
| 49 |
-
- Limited to the context window size of
|
| 50 |
|
| 51 |
|
|
|
|
| 31 |
- Base model: EleutherAI/pythia-14m
|
| 32 |
- Dataset: https://zenodo.org/records/8196385/files/HDFS_v1.zip?download=1 + preprocessed data at honicky/log-analysis-hdfs-preprocessed
|
| 33 |
- Batch size: 4
|
| 34 |
+
- Max sequence length: 405
|
| 35 |
- Learning rate: 0.0001
|
| 36 |
+
- Training steps: 12000
|
| 37 |
|
| 38 |
## Special Tokens
|
| 39 |
- Added `<|sep|>` token for event ID separation
|
|
|
|
| 46 |
|
| 47 |
## Limitations
|
| 48 |
- Model is specifically trained on HDFS logs and may not generalize to other log formats
|
| 49 |
+
- Limited to the context window size of 405 tokens
|
| 50 |
|
| 51 |
|