issues
search
facebookresearch
/
HolisticTraceAnalysis
A library to analyze PyTorch traces.
http://hta.readthedocs.io
MIT License
293
stars
40
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Analyzer reads the log files correctly but doesn't show data
#148
oabuhamdan
opened
4 months ago
0
Migrate memory bandwidth analyzer to HTA
#147
shengfukevin
closed
4 months ago
10
CUPTI Counter Analysis empty
#146
jeromeku
opened
5 months ago
1
Is there any visualization tool for HTA, similar to the visualization interface of tensorboard plugin?
#145
GuWei007
closed
4 months ago
5
[Question] Can HTA work on other trace files (not generated by pytorch) too?
#144
Sarbojit2019
opened
5 months ago
1
add nccl field parser config and test
#143
briancoutinho
closed
5 months ago
2
Fix setting negative weight to 0
#142
pavanky
closed
5 months ago
3
Why are there call_stack.py and trace_call_stack.py at the same time?
#141
zhouyiyuan-mt
opened
5 months ago
0
Evaluate logger.debug only when needed
#140
pavanky
closed
5 months ago
3
Fix CPU kernel start time check when `include_last_profiler_step=True`
#139
jj10306
closed
5 months ago
3
[critical path] Add tolerance for negative one weight due to precision issues, improvements
#138
briancoutinho
closed
5 months ago
4
Enable comm_replay in PARAM by Integrating and Refactoring Comm Code
#137
briancoutinho
closed
5 months ago
2
Add get_stack_of_node to CallGraph
#136
pavanky
closed
5 months ago
9
Add parent to dataframe in call_stack
#135
pavanky
closed
5 months ago
5
Filter out non kernel events for gpu stacks
#134
pavanky
closed
5 months ago
8
Fix sorting of events in call_graph
#133
pavanky
closed
5 months ago
10
0507 add metadata parser ijson
#132
briancoutinho
closed
5 months ago
4
Draft refactor of et replay
#131
briancoutinho
closed
5 months ago
2
correct typos in docs for function get_queue_length_time_series
#130
staugust
closed
5 months ago
3
function get_queue_length_series not found
#129
staugust
closed
4 months ago
0
0501 update gpu kernel filtering
#128
briancoutinho
closed
5 months ago
4
Performance improvements to call_stack
#127
pavanky
closed
6 months ago
2
[critical path] Add save and restore for cp_graph
#126
briancoutinho
closed
5 months ago
8
add attribution for kernel-kernel delay and check for sync on same strea
#125
briancoutinho
closed
6 months ago
3
Estimate TFLOPS of PyTorch Matrix Multiplication Operators from Kineto Trace
#124
fengxizhou
opened
6 months ago
0
Add support for new NCCL naming conventions in the is_comm_kernel method in utils.py
#123
OckermanSethGVSU
closed
6 months ago
3
Categorizing ncclDevKernel_AllReduce_Sum_f32_RING_LL as Computation
#122
OckermanSethGVSU
closed
6 months ago
1
[Critical path] determine previous kernel for an event using the stream
#121
briancoutinho
closed
6 months ago
5
Fix parsing no fwdbwd, add unit test for ns duration and attempt work around
#120
briancoutinho
closed
6 months ago
3
Clarify how traces are collected + Some Minor Documentation Updates
#119
wkaisertexas
opened
7 months ago
2
A faster way to load HTA Trace and create CallGraph
#118
pavanky
closed
7 months ago
13
Critical path analysis - matching the kernel related to cudaEventRecord with stream
#117
briancoutinho
closed
6 months ago
3
[critical path] Add graph validation checks and fix 0 duration stack issue.
#116
briancoutinho
closed
7 months ago
3
[1/n] Optimize performance Critical Path Analysis algorithm for CUDA sync events
#115
briancoutinho
closed
7 months ago
5
Add an interim fix for stack traversal order in HTA
#114
briancoutinho
closed
7 months ago
4
New Trace Callstack processing out of order
#113
briancoutinho
opened
7 months ago
0
Fixing Unary Op evaluate issue
#112
amoghavs
closed
7 months ago
5
Fix Undefined local_symbol_table if "traceEvents" is not in "trace_record"
#111
mkyybx
closed
8 months ago
2
Trace load and json parsing optimizations
#110
briancoutinho
closed
6 months ago
4
update is_comm_kernel check to work with newer nccl versions (and thus generate expected comm/compute overlap numbers)
#109
lessw2020
opened
8 months ago
2
add knob to turn off causal edges
#108
briancoutinho
closed
8 months ago
5
min(arg) is an empty sequence -> issues creating analyzer
#107
wkaisertexas
closed
8 months ago
10
Critical Path Add 0 weight causal edges for kernel launches
#106
briancoutinho
closed
8 months ago
5
Simulating critical paths with optimization heuristics
#105
kvignesh1420
opened
8 months ago
7
HTA expects rank to be specified in a trace file
#104
lishen
closed
8 months ago
3
[critical path] Support Cuda Event synchronization between GPU and CPU
#103
briancoutinho
closed
8 months ago
7
single GPU support
#102
AwePhD
closed
9 months ago
1
Critical path breakdown and summarization
#101
briancoutinho
closed
9 months ago
4
Critical path async memcopy fix
#100
briancoutinho
closed
9 months ago
6
How to plot timeline with the label added by torch.profiler.record_function?
#99
zhouyiyuan-mt
opened
9 months ago
2
Previous
Next