Update README.md
Browse files
README.md
CHANGED
|
@@ -85,6 +85,18 @@ PYTHONPATH=./ deepspeed hydit/train_large_deepspeed.py \
|
|
| 85 |
"$@"
|
| 86 |
```
|
| 87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
## Inference
|
| 89 |
|
| 90 |
### Using Gradio
|
|
|
|
| 85 |
"$@"
|
| 86 |
```
|
| 87 |
|
| 88 |
+
Recommended parameter settings
|
| 89 |
+
|
| 90 |
+
| Parameter | Description | Recommended Parameter Value | Note|
|
| 91 |
+
|:---------------:|:---------:|:---------------------------------------------------:|:--:|
|
| 92 |
+
| `--batch_size` | Training batch size | 1 | Depends on GPU memory|
|
| 93 |
+
| `--grad-accu-steps` | Size of gradient accumulation | 2 | |
|
| 94 |
+
| `--rank` | Rank of lora | 64 | 8-128 are all possible|
|
| 95 |
+
| `--max-training-steps` | Training steps | 2000 | Varies with the amount of training data, about 2000 steps are enough for 100 images|
|
| 96 |
+
| `--lr` | Learning rate | 0.0001 | | |
|
| 97 |
+
|
| 98 |
+
|
| 99 |
+
|
| 100 |
## Inference
|
| 101 |
|
| 102 |
### Using Gradio
|