Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
updating fsdp docs as per resolution to issue https://github.com/huggingface/accelerate/issues/1054 (#3)
Browse files- updating fsdp docs as per resolution to issue https://github.com/huggingface/accelerate/issues/1054 (48d3c971c8a350515cfb5300ee84457740bbcb25)
Co-authored-by: Sourab Mangrulkar <smangrul@users.noreply.huggingface.co>
code_samples/training_configuration/pytorch_fsdp
CHANGED
@@ -27,16 +27,27 @@ use_cpu: false
|
|
27 |
<pre>
|
28 |
from accelerate import Accelerator
|
29 |
|
30 |
-
|
31 |
-
|
32 |
-
-
|
33 |
-
-
|
34 |
-
|
35 |
-
|
36 |
-
+
|
37 |
-
+optimizer
|
38 |
-
+
|
39 |
-
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
</pre>
|
41 |
##
|
42 |
If the YAML was generated through the `accelerate config` command:
|
|
|
27 |
<pre>
|
28 |
from accelerate import Accelerator
|
29 |
|
30 |
+
def main():
|
31 |
+
accelerator = Accelerator()
|
32 |
+
- model, optimizer, dataloader, scheduler = accelerator.prepare(
|
33 |
+
- model, optimizer, dataloader, scheduler
|
34 |
+
- )
|
35 |
+
+ model = accelerator.prepare(model)
|
36 |
+
+ # Optimizer can be any PyTorch optimizer class
|
37 |
+
+ optimizer = torch.optim.AdamW(params=model.parameters(), lr=lr)
|
38 |
+
+ optimizer, dataloader, scheduler = accelerator.prepare(
|
39 |
+
+ optimizer, dataloader, scheduler
|
40 |
+
+ )
|
41 |
+
|
42 |
+
...
|
43 |
+
|
44 |
+
accelerator.unwrap_model(model).save_pretrained(
|
45 |
+
args.output_dir,
|
46 |
+
is_main_process=accelerator.is_main_process,
|
47 |
+
save_function=accelerator.save,
|
48 |
+
+ state_dict=accelerator.get_state_dict(model)
|
49 |
+
)
|
50 |
+
...
|
51 |
</pre>
|
52 |
##
|
53 |
If the YAML was generated through the `accelerate config` command:
|