Trained on xsarchitecturalv3com_v3.ckpt
Browse files
last_config/config_file.toml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
[model_arguments]
|
| 2 |
v2 = false
|
| 3 |
v_parameterization = false
|
| 4 |
-
pretrained_model_name_or_path = "/content/pretrained_model/
|
| 5 |
|
| 6 |
[optimizer_arguments]
|
| 7 |
optimizer_type = "AdamW8bit"
|
|
@@ -36,7 +36,7 @@ train_batch_size = 4
|
|
| 36 |
max_token_length = 225
|
| 37 |
mem_eff_attn = false
|
| 38 |
xformers = true
|
| 39 |
-
max_train_steps =
|
| 40 |
max_data_loader_n_workers = 8
|
| 41 |
persistent_data_loader_workers = true
|
| 42 |
gradient_checkpointing = false
|
|
|
|
| 1 |
[model_arguments]
|
| 2 |
v2 = false
|
| 3 |
v_parameterization = false
|
| 4 |
+
pretrained_model_name_or_path = "/content/pretrained_model/xsarchitecturalv3com_v3.ckpt"
|
| 5 |
|
| 6 |
[optimizer_arguments]
|
| 7 |
optimizer_type = "AdamW8bit"
|
|
|
|
| 36 |
max_token_length = 225
|
| 37 |
mem_eff_attn = false
|
| 38 |
xformers = true
|
| 39 |
+
max_train_steps = 3000
|
| 40 |
max_data_loader_n_workers = 8
|
| 41 |
persistent_data_loader_workers = true
|
| 42 |
gradient_checkpointing = false
|