Update CPU benchmark results (#200)

For sparse linear regression model
awslabs · Oct 16, 2018 · bf5ebee · bf5ebee
1 parent 49c41b2
commit bf5ebee
Show file tree

Hide file tree

Showing 2 changed files with 13 additions and 13 deletions.
diff --git a/benchmark/sparse/linear_regression/inference_benchmark.md b/benchmark/sparse/linear_regression/inference_benchmark.md
@@ -19,11 +19,11 @@
 ##### Speed
 | Instance Type | GPUs  | Batch Size  | Keras-MXNet Time/Batch) | Keras-TensorFlow Time/Batch)  |
 |-----|-----|-----|-----|-----|
-| C5.8XLarge |   0  | 64  | 64.2 sec | 64.3 sec
-| C5.8XLarge |   0  | 128 | 32.1 sec | 31.4 sec 
-| C5.8XLarge |   0  | 256 | 16.5 sec | 16.3 sec
-| C5.8XLarge |   0  | 512 | 8.3 sec | 8.8 sec 
-| C5.8XLarge |   0  | 1024 | 4.3 sec | 4.4 sec
+| C5.8XLarge |   0  | 64  | 1.4 sec | 1.3 sec
+| C5.8XLarge |   0  | 128 | 0.9 sec | 0.7 sec 
+| C5.8XLarge |   0  | 256 | 0.6 sec | 0.5 sec
+| C5.8XLarge |   0  | 512 | 0.4 sec | 0.4 sec 
+| C5.8XLarge |   0  | 1024 | 0.3 sec | 0.3 sec
 
 #### Memory Consumed
 | Instance Type | GPUs  | Batch Size | Keras-MXNet (MB) | Keras-TensorFlow (MB)  |
@@ -78,4 +78,4 @@ Run the file as `python run_sparse_benchmark.py`, by default the benchmark runs
 MXNet supports sparse data in 2 NDArray formats - CSRNDArray and RowSparseNDArray which are defined in `mxnet.ndarray.sparse` package
 For further details on MXNet Sparse NDArray API check [documentation related to MXNet Sparse](https://mxnet.incubator.apache.org/api/python/ndarray/sparse.html)
 
-Keras Input layer supports sparse data by setting a boolean placeholder value - check document for [Keras Input layer](https://keras.io/layers/core/#input)
+Keras Input layer supports sparse data by setting a boolean placeholder value - check document for [Keras Input layer](https://keras.io/layers/core/#input)
diff --git a/benchmark/sparse/linear_regression/training_benchmark.md b/benchmark/sparse/linear_regression/training_benchmark.md
@@ -21,17 +21,17 @@
 ###### Note speed calculated using an average of 5 runs
 | Instance Type | GPUs  | Batch Size  | Keras-MXNet (Time/Epoch) | Keras-TensorFlow (Time/Epoch)  |
 |-----|-----|-----|-----|-----|
-| C5.8XLarge |   0  | 64  | 715.4 sec | 676.6 sec
-| C5.8XLarge |   0  | 128 | 346.2 sec | 341.1 sec 
-| C5.8XLarge |   0  | 256 | 168.9 sec | 165.5 sec
-| C5.8XLarge |   0  | 512 | 89.3 sec | 83.8 sec 
-| C5.8XLarge |   0  | 1024 | 50.5 sec | 48.6 sec
+| C5.8XLarge |   0  | 64  | 25.0 sec | 20.3 sec
+| C5.8XLarge |   0  | 128 | 14.6 sec | 11.9 sec 
+| C5.8XLarge |   0  | 256 | 8.9 sec | 7.8 sec
+| C5.8XLarge |   0  | 512 | 5.5 sec | 5.8 sec 
+| C5.8XLarge |   0  | 1024 | 4.4 sec | 5.6 sec
 
 #### CPU Memory Consumed
 | Instance Type | GPUs  | Batch Size | Keras-MXNet (CPU Memory (MB)) | Keras-TensorFlow (CPU Memory (MB))  |
 |-----|-----|-----|-----|-----|
 | C5.8XLarge |   0  | 64  | 1331.3 | 1578.2 |
-| C5.8XLarge |   0  | 128 | 1387.3 | 1472.1 
+| C5.8XLarge |   0  | 128 | 1387.3 | 1472.1 |
 | C5.8XLarge |   0  | 256 | 1431.3 | 1511.2  |
 | C5.8XLarge |   0  | 512 | 1458.2 | 1506.5 |
 | C5.8XLarge |   0  | 1024 | 1247.8 | 1329.1 |
@@ -104,4 +104,4 @@ Run the file as `python run_sparse_benchmark.py`, by default the benchmark runs
 MXNet supports sparse data in 2 NDArray formats - CSRNDArray and RowSparseNDArray which are defined in `mxnet.ndarray.sparse` package
 For further details on MXNet Sparse NDArray API check [documentation related to MXNet Sparse](https://mxnet.incubator.apache.org/api/python/ndarray/sparse.html)
 
-Keras Input layer supports sparse data by setting a boolean placeholder value - check document for [Keras Input layer](https://keras.io/layers/core/#input)
+Keras Input layer supports sparse data by setting a boolean placeholder value - check document for [Keras Input layer](https://keras.io/layers/core/#input)