"...git@developer.sourcefind.cn:renzhc/diffusers_dcu.git" did not exist on "6c60e430eeb050535aee8854c6ad649b162b62a9"
Unverified Commit 842fb72d authored by Anton Obukhov's avatar Anton Obukhov Committed by GitHub
Browse files

Update README.md

update readme with more precise formulations
parent c721a32b
...@@ -8,13 +8,13 @@ ...@@ -8,13 +8,13 @@
![License](https://img.shields.io/pypi/l/torch-discounted-cumsum) ![License](https://img.shields.io/pypi/l/torch-discounted-cumsum)
This repository implements an efficient parallel algorithm for the computation of discounted cumulative sums This repository implements an efficient parallel algorithm for the computation of discounted cumulative sums
and a Python package with differentiable bindings to PyTorch. The `cumsum` operation is frequently seen in data science and a Python package with differentiable bindings to PyTorch. The discounted `cumsum` operation is frequently seen in
domains concerned with time series, including Reinforcement Learning (RL). data science domains concerned with time series, including Reinforcement Learning (RL).
The traditional sequential algorithm performs the computation of the output elements in a loop. For an input of size The traditional sequential algorithm performs the computation of the output elements in a loop. For an input of size
`N`, it requires `O(N)` operations and takes `O(N)` time steps to complete. `N`, it requires `O(N)` operations and takes `O(N)` time steps to complete.
The proposed parallel algorithm requires a total of `O(N log N)` operations, but takes only `O(log N)` time, which is a The proposed parallel algorithm requires a total of `O(N log N)` operations, but takes only `O(log N)` time steps, which is a
considerable trade-off in many applications involving large inputs. considerable trade-off in many applications involving large inputs.
Features of the parallel algorithm: Features of the parallel algorithm:
...@@ -71,8 +71,8 @@ N = 8 ...@@ -71,8 +71,8 @@ N = 8
K = 2 K = 2
gamma = 0.99 gamma = 0.99
x = torch.ones(1, N).cuda() x = torch.ones(1, N).cuda()
y_inf = discounted_cumsum_right(x, gamma) y_N = discounted_cumsum_right(x, gamma)
y_K = y_inf - (gamma ** K) * torch.cat((y_inf[:, K:], torch.zeros(1, K).cuda()), dim=1) y_K = y_N - (gamma ** K) * torch.cat((y_N[:, K:], torch.zeros(1, K).cuda()), dim=1)
print(y_K) print(y_K)
``` ```
...@@ -87,16 +87,17 @@ tensor([[1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.0000]], ...@@ -87,16 +87,17 @@ tensor([[1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.0000]],
## Parallel Algorithm ## Parallel Algorithm
For the sake of simplicity, the algorithm is explained for `N=16`. For the sake of simplicity, the algorithm is explained for `N=16`.
The processing is performed in-place in the input vector in `log2 N` stages. Each stage updates `N / 2` positions. Each The processing is performed in-place in the input vector in `log2 N` stages. Each stage updates `N / 2` positions in parallel
stage is characterized by the size of the group of sequential elements being updated, which is computed as `2 ** stage`. (that is, in a single time step, provided unrestricted parallelism). A stage is characterized by the size of the group of
sequential elements being updated, which is computed as `2 ^ (stage - 1)`.
The group stride is always twice larger than the group size. The elements updated during the stage are highlighted with The group stride is always twice larger than the group size. The elements updated during the stage are highlighted with
the respective stage color in the figure below. Here input elements are denoted with their position id in hex, and the respective stage color in the figure below. Here input elements are denoted with their position id in hex, and the
2-symbols updated elements indicate the range of indices over which the discounted partial sum has been computed. elements tagged with two symbols indicate the range over which the discounted partial sum is computed upon stage completion.
Each element update includes an in-place addition of a discounted element, which follows the last Each element update includes an in-place addition of a discounted element, which follows the last
updated element in the group. The discount factor is computed as gamma raised to the power of the distance between the updated element in the group. The discount factor is computed as gamma raised to the power of the distance between the
updated and the discounted elements. In the figure below, this operation is denoted with slanted arrows with a greek updated and the discounted elements. In the figure below, this operation is denoted with tilted arrows with a greek
gamma tag. After the last stage finishes, the output is written in place of the input. gamma tag. After the last stage completes, the output is written in place of the input.
![Parallel Algorithm](doc/img/algorithm.png) ![Parallel Algorithm](doc/img/algorithm.png)
...@@ -112,7 +113,7 @@ reversed direction of summation. ...@@ -112,7 +113,7 @@ reversed direction of summation.
The parallel algorithm produces a more numerically-stable output than the sequential algorithm using the same scalar The parallel algorithm produces a more numerically-stable output than the sequential algorithm using the same scalar
data type. data type.
The comparison is performed between 3 runs with identical inputs ([code](tests/test.py)). The first run casts inputs to The comparison is performed between 3 runs with identical inputs ([code](tests/test.py#L116)). The first run casts inputs to
double precision and obtains the output reference using the sequential algorithm. Next, we run both sequential and double precision and obtains the output reference using the sequential algorithm. Next, we run both sequential and
parallel algorithms with the same inputs cast to single precision and compare the results to the reference. The parallel algorithms with the same inputs cast to single precision and compare the results to the reference. The
comparison is performed using the `L_inf` norm, which is just the maximum of per-element discrepancies. comparison is performed using the `L_inf` norm, which is just the maximum of per-element discrepancies.
...@@ -123,10 +124,10 @@ With 10000-element non-zero-centered input (such as all elements are 1.0), the e ...@@ -123,10 +124,10 @@ With 10000-element non-zero-centered input (such as all elements are 1.0), the e
## Speed-up ## Speed-up
We tested 3 implementations of the algorithm with the same 100000-element input ([code](tests/test.py)): We tested 3 implementations of the algorithm with the same 100000-element input ([code](tests/test.py#L154)):
1. Sequential in PyTorch (as in 1. Sequential in PyTorch on CPU (as in
[REINFORCE](https://github.com/pytorch/examples/blob/87d9a1e/reinforcement_learning/reinforce.py#L66-L68)) [REINFORCE](https://github.com/pytorch/examples/blob/87d9a1e/reinforcement_learning/reinforce.py#L66-L68)) (Intel Xeon CPU, DGX-1)
2. Sequential in C++ (Intel Xeon CPU, DGX-1) 2. Sequential in C++ on CPU (Intel Xeon CPU, DGX-1)
3. Parallel in CUDA (NVIDIA P-100, DGX-1) 3. Parallel in CUDA (NVIDIA P-100, DGX-1)
The observed speed-ups are as follows: The observed speed-ups are as follows:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment