Design Doc: Distributed Lookup Table Operator
A lookup table operator in PaddlePaddle where the table could be out of the memory of a computer.
A lookup table operator is well-used in deep learning for learning the representation, or the embedding, of symbols.
The Forward Algorithm
The forward algorithm of the lookup table is a multiplication of the input vector x and the lookup table matrix W:
$$y = x * W$$
When x is a sparse vector of symbols, the above multiplication simplifies into looking up rows in W that correspond to symbols in x, denoted by W(x). Please be aware that W could be huge and out of the memory, so we'd need a distributed storage service, which supports the lookup of rows.
The following figure illustrates the multiplication of x with two non-zero elements, or say, two symbols, and a lookup table W:
The Backward Algorithm
The backward algorithm computes W'(x) using W(x). W'(x) has the same scale of size as W(x) and is much smaller than W.
To optimize W given W', we can do simple SGD update:
$$W = f(W') = \lambda * W'$$
or some more sophisticated algorithms that rely on both W' and W:
$$W = f(W, W')$$
The following figure illustrates the backward pass of the lookup operator:
Distributed Storage Service
The forward algorithm requires a distributed storage service for W. The backward algorithm prefers that the storage system can apply the optimization algorithm on W. The following two sections describe two solutions -- the former doesn't require that the storage service can do optimization, the latter does.
Storage Service Doesn't Optimize
In this design, we use highly-optimized distributed storage, e.g., memcached, as the storage service, and we run the optimization algorithm on parameter servers of PaddlePaddle. The following figure illustrates the training process.
Each trainer runs the forward and backward passes using their local data:
- In the forward pass, when a trainer runs the forward algorithm of a lookup operator, it retrieves W(x) from the storage service.
- The trainer computes W'(x) in the backward pass using W(x).
During the global update process:
- Each trainer uploads its W'(x) to parameter servers.
- The parameter server runs the optimization algorithm, e.g., the Adam optimization algorithm, which requires that
- The parameter server retrieves W(x) from memcached, and
- The parameter server pushes $Delta W(x)=f(W(x), lambda sum_j W'(x))$ to memcached, where $f$ denotes the optimization algorithm.
Storage Service Does Optimize
This design is very similar to the above one, except that the optimization algorithm $f$ runs on the storage service.
- Pro: parameter servers do not retrieve W(x) from the storage service, thus saves half network communication.
- Con: the storage service needs to be able to run the optimization algorithm.
Let us do the "storage service does not optimize" solution first, as a baseline at least, because it is easier to use a well-optimized distributed storage service like memcached. We can do the "storage service does optimize" solution later or at the same time, which, if implemented carefully, should have better performance than the former.