Issues
- 5
exception during pcfg resampling
#72 opened by tmills - 2
out of memory error in workers
#71 opened by lifengjin - 0
no batchsize larger than 32 should be used
#70 opened by lifengjin - 0
build column matrix directly
#38 opened by tmills - 0
- 16
Sentence jobs not finishing at end of iter
#47 opened by coryshain - 2
Bad sample error
#48 opened by coryshain - 0
- 0
- 1
thrust errors after split-merge iterations.
#65 opened by tmills - 0
allow +fork/-join setting to d>1 sentences
#23 opened by tmills - 0
fix log probs on infinite sampler
#27 opened by tmills - 1
workers hang without logging anything
#31 opened by tmills - 0
separate cuda code from header files
#69 opened by tmills - 0
- 0
organize scripts directory
#67 opened by tmills - 0
have cython compile to temp directories
#68 opened by tmills - 1
bad memory access
#50 opened by lifengjin - 1
- 1
guess the connectivity
#52 opened by lifengjin - 0
cuda segfault?
#61 opened by lifengjin - 0
swap branches
#66 opened by lifengjin - 0
- 29
issue with incrementing counts in gpu worker
#41 opened by tmills - 6
increment_counts incorrectly reads parse
#63 opened by tmills - 3
num_tokens?
#62 opened by lifengjin - 7
p factoring merge
#53 opened by lifengjin - 0
- 2
sharing models is slow due to network filesystems
#54 opened by tmills - 0
implement factored model in cpu workers
#57 opened by tmills - 1
- 1
- 0
work servers bind to loopback device on ubuntu
#51 opened by tmills - 0
reduce state space used by f/j variables
#37 opened by tmills - 0
cythonize workers completely
#39 opened by tmills - 0
hid_seq too large?
#45 opened by lifengjin - 4
merge of gpu branches and master
#43 opened by lifengjin - 4
tons of pos
#44 opened by lifengjin - 2
Possible memory leak in c++/gpu version
#42 opened by tmills - 0
Distribute model building
#36 opened by tmills - 0
Implement hierarchical priors for d>1
#34 opened by tmills - 0
fix counting for batch processing
#35 opened by tmills - 1
develop depth d sampler
#28 opened by tmills - 0
debug d2 transitions
#33 opened by tmills - 2
use gold pos for initialization
#29 opened by tmills - 0
have model compiler write models
#30 opened by tmills - 0
use scipy sparse matrices when possible
#32 opened by tmills - 0
- 0
- 0
separate worker code from inference code
#25 opened by tmills