Skip to content

Commit

Permalink
docs: include docs on installing deepspeed w/ cpuadam
Browse files Browse the repository at this point in the history
Signed-off-by: Oleg S <[email protected]>
  • Loading branch information
RobotSail committed Nov 12, 2024
1 parent b812604 commit ea9748f
Showing 1 changed file with 28 additions and 1 deletion.
29 changes: 28 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -121,7 +121,34 @@ allow you to customize aspects of the ZeRO stage 2 optimizer.

For more information about DeepSpeed, see [deepspeed.ai](https://www.deepspeed.ai/)

#### `FSDPOptions`
#### DeepSpeed with CPU Offloading

To use DeepSpeed with CPU offloading, you'll usually encounter an issue indicating that the optimizer needed to use the Adam optimizer on CPU doesn't exist. To resolve this, please follow the following steps:

**Rebuild DeepSpeed with CPUAdam**:

You'll need to rebuild DeepSpeed in order for the optimizer to be present:

```bash
# uninstall deepspeed & reinstall with the flags for installing CPUAdam
pip uninstall deepspeed
DS_BUILD_CPU_ADAM=1 DS_BUILD_UTILS=1 pip install deepspeed --no-deps
```

**Ensure `-lcurand` is linked correctly**:

A problem that we commonly encounter is that the `-lcurand` linker will not be present when
DeepSpeed recompiles. To resolve this, you will need to find the location of the `libcurand.so` file in your machine and ensure it's present in `/usr/lib64`:

```bash
sudo ln -s /usr/local/cuda/lib64/libcurand.so.10 /usr/lib64/libcurand.so
```

> ![NOTE]
> The libcurand file may be located elswhere on your machine. To find it, you can use the following command:
> `find / -name 'libcurand*.so*' 2>/dev/null`

Check failure on line 149 in README.md

View workflow job for this annotation

GitHub Actions / markdown-lint

Trailing spaces

README.md:149:46 MD009/no-trailing-spaces Trailing spaces [Expected: 0 or 2; Actual: 1] https://github.com/DavidAnson/markdownlint/blob/v0.35.0/doc/md009.md
### `FSDPOptions`

Like DeepSpeed, we only expose a number of parameters for you to modify with FSDP.
They are listed below:
Expand Down

0 comments on commit ea9748f

Please sign in to comment.