Update README.md
Browse files
README.md
CHANGED
|
@@ -30,10 +30,6 @@ Code: [github.com/leszkolukasz/diffusers](https://github.com/leszkolukasz/diffus
|
|
| 30 |
| `exp7/` | Pattern-dataset experiments | 2D patterns | Gaussian 8×8, 8×4, 6×6, Spiral5, Beehive |
|
| 31 |
| `edm2/` | External reference | ImageNet-64 | Official NVIDIA EDM2-S checkpoint |
|
| 32 |
|
| 33 |
-
Experiment 7 is not explicitly numbered in the thesis; its results are the 2D
|
| 34 |
-
pattern experiments included in the thesis materials.
|
| 35 |
-
|
| 36 |
-
The `exp1/` files keep their original `exp1v2_` filename prefix.
|
| 37 |
|
| 38 |
## File types
|
| 39 |
|
|
@@ -43,8 +39,6 @@ For each experiment:
|
|
| 43 |
- `raw_*` — raw model weights,
|
| 44 |
- `trainer_*` — training state (epoch, optimizer and scaler).
|
| 45 |
|
| 46 |
-
Model checkpoints contain a `state_dict` and an embedded `config`.
|
| 47 |
-
|
| 48 |
## Usage
|
| 49 |
|
| 50 |
Download a checkpoint:
|
|
@@ -63,10 +57,10 @@ Load it with the thesis code:
|
|
| 63 |
```bash
|
| 64 |
uv run main.py generate \
|
| 65 |
--model-path "$path" \
|
| 66 |
-
--schedule
|
| 67 |
-
--sampling-schedule
|
| 68 |
-
--equation
|
| 69 |
-
--solver
|
| 70 |
--n-steps 32 \
|
| 71 |
--n-samples 1
|
| 72 |
```
|
|
|
|
| 30 |
| `exp7/` | Pattern-dataset experiments | 2D patterns | Gaussian 8×8, 8×4, 6×6, Spiral5, Beehive |
|
| 31 |
| `edm2/` | External reference | ImageNet-64 | Official NVIDIA EDM2-S checkpoint |
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
## File types
|
| 35 |
|
|
|
|
| 39 |
- `raw_*` — raw model weights,
|
| 40 |
- `trainer_*` — training state (epoch, optimizer and scaler).
|
| 41 |
|
|
|
|
|
|
|
| 42 |
## Usage
|
| 43 |
|
| 44 |
Download a checkpoint:
|
|
|
|
| 57 |
```bash
|
| 58 |
uv run main.py generate \
|
| 59 |
--model-path "$path" \
|
| 60 |
+
--schedule <schedule> \
|
| 61 |
+
--sampling-schedule <sampling-schedule> \
|
| 62 |
+
--equation <equation> \
|
| 63 |
+
--solver <solver> \
|
| 64 |
--n-steps 32 \
|
| 65 |
--n-samples 1
|
| 66 |
```
|