mirrors
/
moco-v3
mirror of https://github.com/facebookresearch/moco-v3.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

start readme

pull/3/head
Xinlei Chen 2021-07-08 05:04:30 -07:00
parent 4bbfb6633a
commit 87e487c6bb
1 changed files with 6 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
README.md
View File

@ -23,9 +23,7 @@ Similar to MoCo, only **multi-gpu**, **DistributedDataParallel** training is sup
Below we exemplify several MoCo v3 pre-training commands covering different model architectures, training epochs, single-/multi-node, etc.
<details>
<summary>
ResNet-50, 100-Epoch, 2-Node.
</summary>
<summary>ResNet-50, 100-Epoch, 2-Node.</summary>
This is the *default* setting for most hyper-parameters. With a batch size of 4096, the training fits into 2 nodes with a total of 16 Volta 32G GPUs.
@ -46,9 +44,7 @@ python main_moco.py \
</details>
<details>
<summary>
ResNet-50, 300-Epoch, 2-Node.
</summary>
<summary>ResNet-50, 300-Epoch, 2-Node.</summary>
On the first node, run:
```
@ -62,9 +58,7 @@ On the second node, run the same command as above, with `--rank 1`.
</details>
<details>
<summary>
ResNet-50, 1000-Epoch, 2-Node.
</summary>
<summary>ResNet-50, 1000-Epoch, 2-Node.</summary>
On the first node, run:
```
@ -78,10 +72,9 @@ On the second node, run the same command as above, with `--rank 1`.
</details>
<details>
<summary>
ViT-Small, 100-Epoch, 1-Node.
</summary>
With a batch size of 1024, ViT-Small fits into a single node of 8 Volta 32G GPUs.
<summary>ViT-Small, 100-Epoch, 1-Node.</summary>
With a batch size of 1024, ViT-Small fits into a single node of 8 Volta 32G GPUs:
```
python main_moco.py \