is it possible to save the learnt vector and matrix somehow?

This project is a good sample to learn parameter server. However, we are still waiting for distributed implement.

In case it is useful to anyone, managed to build and run the example on recent osx (clang) with a couple of minor patches. See https://github.com/cottrell/difacto/tree/osx

709eb4b
b7ec0c4

$ gcc --version
Configured with: --prefix=/Library/Developer/CommandLineTools/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.3.0 (clang-703.0.29)
Target: x86_64-apple-darwin15.3.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin

Can't find row_block.h file in data directory.

Hi, guys:
As has been introduced in the paper 'DiFacto — Distributed Factorization Machines', Difacto's communication relies on the Parameter Server. From the project's dictionary, we see that kv_app.h encapsulates the van logic to supply some higher but simplistic interface Pull, Push;
However we don't see any function call of Push and Pull in apps such as BCD, LBFGS, SGD etc., which is called by store_local's Pull through the function Get. So, I am puzzeled what is relationship between the kv_app(KVWorker, KVServer) and Apps(BCD, LBFGS, SGD). Can anybody help me ?

Hello, when I try to run quick start demo, I came to this

[root@algo064 difacto]# build/difacto data_in=data/gisette_scale val_data=data/gisette_scale.t lr=.02 V_dim=2 V_lr=.001 batch_size=1000
[11:28:34] src/main.cc:27: Unrecognized keyword argument for task = train
[11:28:34] src/main.cc:29:  - val_data = data/gisette_scale.t
[11:28:34] src/sgd/sgd_learner.cc:35: Start epoch 0
[11:28:34] ./src/common/spmm.h:177: Check failed: (sp + k) <= (max_len) 
Aborted

Any idea how to fix that ?

./src/common/./kv_match.h:141:3: error: no matching function for call to 'KVMatch'

we are still working on porting the codes from https://github.com/dmlc/wormhole/tree/master/learn/difacto and also adding new learning algorithms:

• logit / fm loss: src/loss/
• minibatch sgd: src/sgd/
• block coordinate descent: src/bcd
• l-bfgs: src/lbfgs
• distributed by ps-lite: src/store/store_dist.h
• use disk as data cache: class DataStoreDisk
• save model
• predict
• user doc
• developer doc

When I run sample code, there is always an error

terminate called after throwing an instance of 'dmlc::ParamError'
what(): Required parameter batch_size of int is not presented
Aborted (core dumped)

I think maybe you should set default in parameter batch_size.

[16:06:54] include/dmlc/././logging.h:208: [16:06:54] src/io.cc:57: unknown filesystem protocol viewfs:// terminate called after throwing an instance of 'dmlc::Error' what(): [16:06:54] src/io.cc:57: unknown filesystem protocol viewfs:// Exception in thread Thread-1: Traceback (most recent call last): File "/usr/lib64/python2.7/threading.py", line 811, in __bootstrap_inner self.run() File "/usr/lib64/python2.7/threading.py", line 764, in run self.__target(*self.__args, **self.__kwargs) File "/data/home/GPkuaibaofst/model_train/train/bin/tracker/dmlc_tracker/tracker.py", line 363, in <lambda> target=(lambda: subprocess.check_call(self.cmd, env=env, shell=True)), args=()) File "/usr/lib64/python2.7/subprocess.py", line 542, in check_call raise CalledProcessError(retcode, cmd)
I am wondering if difacto will support viewfs ?

I trained the model by following command:
build/difacto data_in=data/gisette_scale val_data=data/gisette_scale.t lr=.02 V_dim=2 V_lr=.001 batch_size=10240 model_out=./fm.model task=train learner=lbfgs
when it's done, I can't find the output model, where is it ?

Although I wasn't able to build on MacOS, the project built on Linux. However, I try to follow the example steps, and it breaks when I try to build a model:

build/difacto data_in=data/gisette_scale val_data=data/gisette_scale.t lr=.02 V_dim=2 V_lr=.001

The error is:

terminate called after throwing an instance of 'dmlc::ParamError'
  what():  Required parameter batch_size of int is not presented
Aborted (core dumped)

@mli in the readme file, I see the example of running code on one machine,but how to run it on different machines ? I do appreciate if you can complete the docs and show us the way run difacto on more than 2 machines . In my opinion, the command shoud contains a file which is the ip list ,etc . The needs of run it on cluster is urgent in industry , your reply will be a greate favor to us .