|
gptkbp:instance_of
|
gptkb:Tensor_Flow
|
|
gptkbp:application
|
computer vision
natural language processing
reinforcement learning
speech recognition
image classification
recommendation systems
|
|
gptkbp:architecture
|
symmetric
custom ASIC
|
|
gptkbp:availability
|
gptkb:Google_Cloud_Platform
|
|
gptkbp:cloud_integration
|
gptkb:Google_Cloud_AI
|
|
gptkbp:compatibility
|
gptkb:Linux
|
|
gptkbp:connects
|
high-speed network
|
|
gptkbp:data_type
|
int64
uint32
uint64
int32
uint8
float32
bfloat16
int16
int8
uint16
|
|
gptkbp:designed_by
|
gptkb:Google
|
|
gptkbp:first_generation
|
4th generation
|
|
gptkbp:form_factor
|
PCIe card
|
|
gptkbp:has_units
|
128
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
TPU v4
|
|
gptkbp:is_scalable
|
high
|
|
gptkbp:notable_feature
|
high throughput
|
|
gptkbp:number_of_cores
|
256
|
|
gptkbp:performance
|
gptkb:AI_Benchmark
gptkb:Cloud_TPU
gptkb:MLPerf
low
up to 275 teraflops
Deep Learning Benchmark
Tensor Flow Benchmark
|
|
gptkbp:power_consumption
|
200 watts
|
|
gptkbp:ram
|
32 GB HBM
|
|
gptkbp:release_year
|
gptkb:2021
|
|
gptkbp:successor
|
gptkb:TPU_v3
|
|
gptkbp:support
|
gptkb:C++
gptkb:Python
gptkb:R
gptkb:Go
gptkb:Java_Script
|
|
gptkbp:supports
|
gptkb:Tensor_Flow
|
|
gptkbp:target_market
|
AI researchers
|
|
gptkbp:used_for
|
gptkb:machine_learning
|
|
gptkbp:bfsParent
|
gptkb:TPU_v3
gptkb:Tensor_Processing_Units_(TPUs)
gptkb:TPUs
|
|
gptkbp:bfsLayer
|
5
|