Commits

Branch selector

 master

User selector

amitaga

Datepicker

All time

Commit History

Commits on Feb 25, 2016

• Fix a memshare bug to prevent non-shareable matrices to be allocated from the shared matrix pool

  amitaga

  committed
  34de7e5

Commits on Feb 24, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreader_allowlargeutterances

  amitaga

  committed
  7db9ee7

Commits on Feb 23, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreader_allowlargeutterances

  amitaga

  committed
  0235291

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreader_allowlargeutterances

  amitaga

  committed
  fd6a076

Commits on Feb 22, 2016

• Increase the supported number of frames per utternace to 2^16 since our frame randomization implementation no longer requires framerefs for entire corpus to be in memory and we can accomodate a lar…

  Show description for bf5ea19
  amitaga

  committed
  bf5ea19

Commits on Feb 18, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/evalBugFix

  amitaga

  committed
  3a8d7be

• Fixed a bug in EvalWriter.h where the output matrix was being inadvertently transferred to CPU during SaveData

  amitaga

  committed
  0899cee

Commits on Feb 17, 2016

• Fixed CNTKEval bug preventing eval from running on GPU

  amitaga

  committed
  d39709f

Commits on Feb 11, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  9386b8d

• Addressed CR feedback

  amitaga

  committed
  e511c9d

Commits on Feb 10, 2016

• Temporarily retained old memory inefficient frame randomization for side by side testing. The old randomization is off by default and can be switched on using a config setting minimizeReaderMemoryF…

  Show description for 58c768b
  amitaga

  committed
  58c768b

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  3021ad0

• Fixed linux build

  amitaga

  committed
  7baf28a

• Replace incorrect use of bad_exception with runtime_error exceptions

  amitaga

  committed
  2ec9d7a

Commits on Feb 9, 2016

• Write GPUMatrix to file in chunked manner to avoid creating very large intermediate CPU buffers

  amitaga

  committed
  409327a

• Added new handler to print call stack on allocation failure

  amitaga

  committed
  439707a

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  57fd382

• Removed fsync call from fflushOrDie since its is causing issues when writing philly models to HDFS

  amitaga

  committed
  5945c39

• Write checkpoint file before writing model file, so that a checkpoint file is guaranteed to exist for the latest successfully written model file

  amitaga

  committed
  e9f5ca4

Commits on Feb 8, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  5a67cf2

• Disabled HTKMLFReader unit tests HTKMLFReaderSimpleDataLoop7 and HTKMLFReaderSimpleDataLoop17 as they use invalid config setting combination of Truncated=true and frameMode=true

  amitaga

  committed
  c214980

• HTKMLFReader memory optimization: Reimplemented frame mode randomization to keep just a window of randomized frames in memory instead of having a full array of all frames in the corpus

  amitaga

  committed
  a2b99a4

Commits on Feb 6, 2016

• Added a check for inconsistent frameMode and Truncated settings. Also addresed some minor CR feedback.

  amitaga

  committed
  f321625

Commits on Feb 5, 2016

• Added a workaround for the HDFS FUSE implementation bug of failure to delete destination file when the destination file specified to rename function already exists

  amitaga

  committed
  84e9db1

Commits on Feb 4, 2016

• Fixed redundant writing of the .0 model file when no PreCompute is performed and the model is loaded from a checkpoint

  amitaga

  committed
  d3e636c

• Added guards to all model saving calls to prevent mutiple parallel workers from trying to write the model file concurrently

  amitaga

  committed
  14881a2

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  27fd4e0

Commits on Feb 3, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  78587f5

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  c21b233

Commits on Feb 2, 2016

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  8c38137

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/devicePlacement

  amitaga

  committed
  0dbd36e

• HTKMLFReader memory optimizations: Added string interning to the parsedpath type and changed logicalpath member type from utf16 to utf8

  amitaga

  committed
  292ec03

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  6f9a81c

• Change number of bits used for utterance index in frameref to 11

  amitaga

  committed
  f628e52

• Merge branch 'master' of https://github.com/Microsoft/CNTK into amitaga/htkmlfreaderMemoryOptimizations

  amitaga

  committed
  86a356f

Pagination

PreviousNext