Merge pull request #104 from laekov/faster-doc

Documents for FasterMoE

README.md

@@ -99,8 +99,20 @@ FastMoE's model parallel requires sophiscated parallel strategies that neither P
 Megatron-LM provides. The `fmoe.DistributedGroupedDataParallel` module is
 introduced to replace PyTorch's DDP module.
 
+#### Faster Performance Features
+
+From a PPoPP'22 paper, _FasterMoE: modeling and optimizing training of
+large-scale dynamic pre-trained models_, we have adopted techniques to make
+FastMoE's model parallel much more efficient.
+
+These optimizations are named as **Faster Performance Features**, and can be
+enabled via several environment variables. Their usage and constraints are
+detailed in [a separate document](doc/fastermoe).
+
 ## Citation
 
+For the core FastMoE system.
+
 ```
 @article{he2021fastmoe,
       title={FastMoE: A Fast Mixture-of-Expert Training System}, 
@@ -110,6 +122,27 @@ introduced to replace PyTorch's DDP module.
 }
 ```
 
+For the [faster performance features](doc/fastermoe).
+
+```
+@inproceedings{he2022fastermoe,
+    author = {He, Jiaao and Zhai, Jidong and Antunes, Tiago and Wang, Haojie and Luo, Fuwen and Shi, Shangfeng and Li, Qin},
+    title = {FasterMoE: Modeling and Optimizing Training of Large-Scale Dynamic Pre-Trained Models},
+    year = {2022},
+    isbn = {9781450392044},
+    publisher = {Association for Computing Machinery},
+    address = {New York, NY, USA},
+    url = {https://doi.org/10.1145/3503221.3508418},
+    doi = {10.1145/3503221.3508418},
+    booktitle = {Proceedings of the 27th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming},
+    pages = {120–134},
+    numpages = {15},
+    keywords = {parallelism, distributed deep learning, performance modeling},
+    location = {Seoul, Republic of Korea},
+    series = {PPoPP '22}
+}
+```
+
 ## Troubleshootings / Discussion
 
 If you have any problem using FastMoE, or you are interested in getting involved in developing FastMoE, feel free to join [our slack channel](https://join.slack.com/t/fastmoe/shared_invite/zt-mz0ai6ol-ggov75D62YsgHfzShw8KYw).

doc/fastermoe/README.md

@@ -0,0 +1,98 @@
+Boost the Performance by FasterMoE 
+===
+
+一个中文版见[这篇博客](https://laekov.com.cn/view/181401#howto)
+
+There are three main optimizations in the PPoPP'22 paper _FasterMoE: Modeling
+and Optimizing Training of Large-scale Dynamic Pre-trained Models_. Thanks to
+the contributions of authors of the article, their optimizations are now
+integrated into FastMoE, and can be enabled via switches of environment
+variables. These optimizations can greatly increase the training efficiency of
+FastMoE.
+
+## Smart Scheduling
+
+Recall that in an MoE layer, two `all-to-all`s are performed with the experts'
+computation in-between. In FasterMoE, the `all-to-all`s are broken down using
+a _group-wise exchange_ algorithm. And then, the expert can instantly start
+its jobs as long as a part of input, e.g. tokens from one other worker, is
+ready.
+
+Its effectiveness is revealed in the following timeline. `S` and `R` stand for
+the components of the `all-to-all`s, and `C` stands for computation of the
+expert.
+
+![](smartsch.png)
+
+In FastMoE, to enable smart scheduling, set the environment variable `
+FMOE_FASTER_SCHEDULE_ENABLE` to `1` or `ON`, and it is now by default off.
+
+Please note that there are a few constraints for smart scheduling in the
+current version of FastMoE. `num_expert` has to be `1`, which means only one
+expert can reside on each worker. The input and output features have to be of
+the same length for the experts. This is because the developers of FasterMoE
+only implement this on their prototype, and they are looking for the
+community's efforts to have other cases supported.
+
+To fine-tune the performance of smart scheduling, the environment variable
+`FMOE_FASTER_GROUP_SIZE` stands for the size of worker groups in the
+_Group-wise Exchange_ algorithm. In other words, it is the granularity of the
+schedule. It should be set to a proper value that balance between pipeline
+bubbles and inefficient undersized computation granularity.
+
+## Expert Shadowing
+
+According to observations when training real models, when no limitation is
+placed over expert selection, it follows a skew distribution, which means a few
+experts are much more popular than others. This introduces significant
+performance issue of load imbalance when using FastMoE's model parallel mode.
+
+The authors of FasterMoE proposes the solution that for the hot experts, their
+parameters are broadcast to all workers, namely shadows. With the shadows,
+computation of the hot experts can be performed locally on all workers,
+avoiding the bottleneck of sending so much workload to the workers containing
+the hot experts. Besides, a performance predictor, together with a shadow
+selection algorithm, is used to determine which experts to be shadowed before
+each iteration.
+
+In FastMoE, this feature is enabled by the environment variable
+`FMOE_FASTER_SHADOW_ENABLE`.  For simplicity, this feature is only available
+when smart scheduling is enabled. Besides the constraints of smart scheduling,
+this feature requires the experts to be identical in structure, so that
+parameters can be copied between experts.
+
+A default shadow selection policy is located at
+`fmoe/fastermoe/shadow_policy.py`. If you want to alter the policy, please code
+there and re-install FastMoE. For the default policy, we assume that the
+experts are two-layer MLPs. A few parameters of the policy can be specified by
+the following environment variables for better effectiveness of the shadowing
+mechanism.
+
+* `FMOE_FASTER_GLBPLC_NETBW` is the bandwidth of the interconnection between
+  workers, measured by `GBps`.
+* `FMOE_FASTER_GLBPLC_GPUTP` is the GeMM throughput of the GPUs, measured by
+  `FLOPs`, e.g. `13e12` for NVIDIA V100 PCIe GPUs using fp32.
+* `FMOE_FASTER_GLBPLC_ALPHA` is the fraction of the activation length in the
+  middle of the MLP to the input and output feature length, commonly seen to be
+`2` or `4` in transformers.
+* `FMOE_FASTER_GLBPLC_DMODEL` is the feature length of input and output of the
+  experts. This parameter can be set automatically by FastMoE.
+
+## Topology-aware Gate
+
+The two optimizations above do not change the behavior of the model, while this
+one does. To reduce network congestion when training in distributed system
+with hierarchical network topology, e.g. many GPUs in each of many nodes, the
+number of samples transmitted through the slower upper-level network is
+limited. The overfilling tokens select experts within the same lower-level
+network to reduce the communication overhead.
+
+The example topology-aware gate is implemented as `FasterGate` among FastMoE's
+gates. However, note that it may influence the accuracy of the model. And for
+different training hardware, different topology-aware gates shall be designed
+according to the specific case.
+
+The environment variable `FMOE_TOPO_GPUS_PER_NODE` represents number of GPUs in
+each local network, e.g. each node. And `FMOE_TOPO_OUTGOING_FRACTION` controls
+the fraction of tokens that are allowed to be sent across the upper-level
+network.

doc/readme-cn.md

@@ -95,6 +95,15 @@ FastMoE 的模型并行模式需要专门的并行策略, 而 PyTorch 和 Megatr
 都不支持这样的策略. 因此, 需要使用 `fmoe.DistributedGroupedDataParallel`
 模块来代替 PyTorch 的 DDP 模块.
 
+### 如何训练得更快
+
+在 PPoPP'22 会议上有一篇论文: _FasterMoE: modeling and optimizing training of
+large-scale dynamic pre-trained models_. 我们将文中的技术集成到了 FastMoE 系统中,
+从而提升其模型并行的效率.
+
+这些新特性被命名为 **Faster Performance Features**, 并通过一些环境变量来控制是否
+启用它们. 详见[这篇单独的文档](doc/fastermoe).
+
 ## 答疑 / 讨论
 
 如果您在使用 FastMoE 的过程中有任何疑问, 或您有兴趣参与 FastMoE 的相关工作,

doc/release-note.md

@@ -1,3 +1,19 @@
+## v1.0.0
+
+### FasterMoE
+
+* The new performance boosting features in the PPoPP'22 paper FasterMoE, detailed in the document.
+	* Expert Shadowing.
+	* Smart Scheduling.
+	* Topology-aware gate.
+
+### Bug fixes
+
+* Transformer-XL examples.
+* Compatibility to PyTorch versions.
+* Megatron-LM documents.
+* GShardGate.
+
 ## v0.3.0
 
 ### FMoE core

fmoe/fastermoe/schedule.py

@@ -61,6 +61,8 @@ def stash_fn(params, idx):
         out = _local_gather(local_output_buf, pos_g, out_batch_size,
                 maybe_overlap=False)
 
+        # gib and local_input_buf are necessary, because ctx.gibs are created
+        # based on their memory
         variables = (pos_s, pos_g, local_expert_count, global_expert_count,
                 stored_models, gib, local_input_buf)
 

setup.py

@@ -13,6 +13,7 @@
         'Tiago Antunes', 
         'Jinjun Peng', 
         'Qin Li',
+        'Mingshu Zhai'
 ]
 
 is_rocm_pytorch = False
@@ -37,7 +38,7 @@
 if __name__ == '__main__':
     setuptools.setup(
         name='fastmoe',
-        version='0.3.0',
+        version='1.0.0',
         description='An efficient Mixture-of-Experts system for PyTorch',
         author=', '.join(authors),
         author_email='[email protected]',

tests/README.md

@@ -0,0 +1,7 @@
+FastMoE test
+===
+
+To run unit test, directly run `pytest` in this directory.
+
+`test.sh` is a wrapper script to execute single tests without pytest for
+debugging purpose.