Delete the files REPORT_* Hieu Sent from my flying horse
On 4 May 2012, at 02:14 PM, Craig Wampler <[email protected]> wrote: > So, I had deleted the 'evaluation' folder, in hopes that it would re-run the > test portion and regenerate the output. It generates an entire plan (the > graphical ghostview plan thing), and looking through the log (attached) it > appears to think that the report is there, and it doesn't need to generate > it, but it's not there. > > So how do I tell it that it needs to generate the report? > > Thanks! > Craig > > On Thu, May 3, 2012 at 11:23 PM, Philipp Koehn <[email protected]> wrote: > Hi, > > you did the right thing here: > > [EVALUATION:europarl-small] > raw-input = $test-dir/newstest2009.$input-extension > raw-reference = $test-dir/newstest2009.$output-extension > > [EVALUATION:wasp0] > raw-input = $data-dir/0/test/test.$input-extension > raw-reference = $data-dir/0/test/test.$output-extension > > [EVALUATION:wasp1] > raw-input = $data-dir/1/test/test.$input-extension > raw-reference = $data-dir/1/test/test.$output-extension > > Why did this not work? What happened? > > The results will be summarize in one report file. > > If you run with different settings (different language models), > then this will need to be a different run of the experiment, but it > will be able to re-use much from the first run (such as the entire > translation model training pipeline). > > -phi > > On Thu, May 3, 2012 at 6:21 PM, Craig Wampler <[email protected]> wrote: > > Hi, > > > > I've followed the EMS walkthrough, and I was able to run an experiment. The > > training/tuning took about 6 hours. I would now like to test it against > > multiple test sets. How do I either > > > > a) modify the config file > > b) delete the right files/folders in my evaluation folder > > > > To get it to re-run and re-report. > > > > Ideally, I'd run the command once against my 10 test sets, and end up with > > 10 reports (with different names), without having to re-train and re-tune > > against he same training and tuning sets. > > > > I'd also like to be able to use two different language models (so a total of > > 20 evaluations). > > > > I've attached a copy of my config. If you look in the evaluation section, on > > line 198 is the evaluation that ran. I then tried adding the two more test > > evaluations (lines 202 and 206), but that didn't work. I also tried deleting > > the steps folder. All it does is pop up an execution plan that says it's > > going to regenerate the report, but it doesn't even regenerate the report. > > > > Thanks! > > > > _______________________________________________ > > Moses-support mailing list > > [email protected] > > http://mailman.mit.edu/mailman/listinfo/moses-support > > > > <log.ems-not-working.txt> > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
