Update README.md
Browse files
README.md
CHANGED
|
@@ -5,4 +5,24 @@ language:
|
|
| 5 |
pipeline_tag: text-generation
|
| 6 |
---
|
| 7 |
|
| 8 |
-
Sharathhebbar24/ssh_1.8B is a 1.8B model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
pipeline_tag: text-generation
|
| 6 |
---
|
| 7 |
|
| 8 |
+
Sharathhebbar24/ssh_1.8B is a 1.8B model
|
| 9 |
+
|
| 10 |
+
The model is a modified version of [qnguyen3/quan-1.8b-chat](https://huggingface.co/qnguyen3/quan-1.8b-chat)
|
| 11 |
+
|
| 12 |
+
## Training hyperparameters
|
| 13 |
+
The following hyperparameters were used during training:
|
| 14 |
+
|
| 15 |
+
learning_rate: 2e-05
|
| 16 |
+
train_batch_size: 2
|
| 17 |
+
eval_batch_size: 2
|
| 18 |
+
seed: 42
|
| 19 |
+
distributed_type: multi-GPU
|
| 20 |
+
num_devices: 4
|
| 21 |
+
gradient_accumulation_steps: 4
|
| 22 |
+
total_train_batch_size: 32
|
| 23 |
+
total_eval_batch_size: 8
|
| 24 |
+
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 25 |
+
lr_scheduler_type: cosine
|
| 26 |
+
lr_scheduler_warmup_steps: 100
|
| 27 |
+
num_epochs: 4
|
| 28 |
+
|