DNN Pruning
30
Apr
Compiler Level Optimizations for Accelerated Deep Learning
#CellStratAILab #disrupt4.0 #WeCreateAISuperstars #AlwaysUpskilling Last Saturday (25th Apr ’20), our AI Lab Researcher Darshan G. presented a fabulous hands-on workshop on Tensor Fusion, Accelerated Linear Algebra (XLA) and many other techniques to speed up Deep Learning computations. The following discussion covers many of these techniques. TensorFlow Graph Concepts :- TensorFlow (v1.x) programs generate a DataFlow […]
Tags:
accelerated linear algebra,
Compiler Optimization,
computation graph,
data parallelism,
DNN Pruning,
GEMM,
General Matrix Multiplication,
grappler,
High Level Optimizer,
HLO,
HLO IR,
Intermediate Representation,
kernel fusion,
kernel optimization,
Memory Access Coalescing,
MLIR,
model parallelism,
Multi-Level Intermediate Representation,
Op Fusion,
Remapper,
Sparse Matrix Vector Multiplication,
SpMV,
Tensor Fusion,
tensorflow,
XLA,