Update README.md
Browse files
README.md
CHANGED
|
@@ -7,9 +7,9 @@ tags:
|
|
| 7 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 8 |
should probably proofread and complete it, then remove this comment. -->
|
| 9 |
|
| 10 |
-
# smb-vision-
|
| 11 |
|
| 12 |
-
This model is trained from scratch using [VideoMAE](https://huggingface.co/docs/transformers/en/model_doc/videomae) on over
|
| 13 |
|
| 14 |
## Model description
|
| 15 |
|
|
@@ -29,31 +29,25 @@ More information needed
|
|
| 29 |
|
| 30 |
The following hyperparameters were used during training:
|
| 31 |
- learning_rate: 3e-04
|
| 32 |
-
- train_batch_size:
|
| 33 |
- eval_batch_size: 1
|
| 34 |
- seed: 42
|
| 35 |
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
|
| 36 |
- lr_scheduler_type: cosine
|
| 37 |
-
- num_epochs:
|
| 38 |
|
| 39 |
### Training results
|
| 40 |
|
| 41 |
{
|
| 42 |
-
"_runtime":
|
| 43 |
-
"_step":
|
| 44 |
-
"
|
| 45 |
-
"
|
| 46 |
-
"
|
| 47 |
-
"
|
| 48 |
-
"train/
|
| 49 |
-
"train/
|
| 50 |
-
"train/
|
| 51 |
-
"train/learning_rate": 0,
|
| 52 |
-
"train/loss": 0.5736,
|
| 53 |
-
"train_loss": 0.5022664608695041,
|
| 54 |
-
"train_runtime": 54785.1298,
|
| 55 |
-
"train_samples_per_second": 2.527,
|
| 56 |
-
"train_steps_per_second": 0.079
|
| 57 |
}
|
| 58 |
|
| 59 |
|
|
@@ -69,7 +63,7 @@ The following hyperparameters were used during training:
|
|
| 69 |
# load data using `dataload.py`
|
| 70 |
|
| 71 |
model = VideoMAEForPreTraining.from_pretrained(
|
| 72 |
-
standardmodelbio/smb-vision-
|
| 73 |
trust_remote_code=True,
|
| 74 |
)
|
| 75 |
|
|
|
|
| 7 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 8 |
should probably proofread and complete it, then remove this comment. -->
|
| 9 |
|
| 10 |
+
# smb-vision-large-1202
|
| 11 |
|
| 12 |
+
This model is trained from scratch using [VideoMAE](https://huggingface.co/docs/transformers/en/model_doc/videomae) on over 55k CT volumes.
|
| 13 |
|
| 14 |
## Model description
|
| 15 |
|
|
|
|
| 29 |
|
| 30 |
The following hyperparameters were used during training:
|
| 31 |
- learning_rate: 3e-04
|
| 32 |
+
- train_batch_size: 16
|
| 33 |
- eval_batch_size: 1
|
| 34 |
- seed: 42
|
| 35 |
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
|
| 36 |
- lr_scheduler_type: cosine
|
| 37 |
+
- num_epochs: 10.0
|
| 38 |
|
| 39 |
### Training results
|
| 40 |
|
| 41 |
{
|
| 42 |
+
"_runtime": 2641.091489502,
|
| 43 |
+
"_step": 399,
|
| 44 |
+
"_timestamp": 1733187755.3146417,
|
| 45 |
+
"_wandb.runtime": 2660,
|
| 46 |
+
"train/epoch": 8.425414364640885,
|
| 47 |
+
"train/global_step": 18300,
|
| 48 |
+
"train/grad_norm": 0.04110511764883995,
|
| 49 |
+
"train/learning_rate": 0.0001624558726951691,
|
| 50 |
+
"train/loss": 0.4292
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 51 |
}
|
| 52 |
|
| 53 |
|
|
|
|
| 63 |
# load data using `dataload.py`
|
| 64 |
|
| 65 |
model = VideoMAEForPreTraining.from_pretrained(
|
| 66 |
+
standardmodelbio/smb-vision-large,
|
| 67 |
trust_remote_code=True,
|
| 68 |
)
|
| 69 |
|