From cdb7f2b26e717bbfe84dbd73f949c6b06ca2b2d7 Mon Sep 17 00:00:00 2001 From: Chris Dyer Date: Mon, 25 Nov 2013 00:14:16 -0500 Subject: remove dead code, add adagrad crf learner --- .gitignore | 2 + decoder/Makefile.am | 7 - decoder/apply_fsa_models.h | 65 ---- decoder/apply_models.cc | 10 +- decoder/cfg.cc | 656 ----------------------------------- decoder/cfg.h | 382 -------------------- decoder/cfg_binarize.h | 90 ----- decoder/cfg_format.h | 134 ------- decoder/cfg_options.h | 79 ----- decoder/cfg_test.cc | 98 ------ decoder/decoder.cc | 99 +----- decoder/decoder.h | 13 +- decoder/ff_factory.cc | 47 +-- decoder/ff_factory.h | 38 +- decoder/ff_klm.cc | 3 +- decoder/hg_cfg.h | 54 --- decoder/sentence_metadata.h | 3 +- decoder/viterbi.cc | 29 +- mteval/external_scorer.h | 3 +- mteval/ns_ext.cc | 2 +- mteval/ns_ter.cc | 2 +- mteval/scorer.h | 4 +- tests/run-system-tests.pl | 4 +- tests/tools/flex-diff.pl | 43 +++ training/crf/Makefile.am | 4 + training/crf/mpi_adagrad_optimize.cc | 355 +++++++++++++++++++ training/crf/mpi_batch_optimize.cc | 6 +- utils/Makefile.am | 8 +- utils/small_vector.h | 12 +- utils/small_vector_test.cc | 12 + utils/sv_test.cc | 24 ++ utils/swap_pod.h | 23 -- utils/value_array.h | 9 +- utils/weights.cc | 10 +- 34 files changed, 521 insertions(+), 1809 deletions(-) delete mode 100644 decoder/apply_fsa_models.h delete mode 100644 decoder/cfg.cc delete mode 100644 decoder/cfg.h delete mode 100644 decoder/cfg_binarize.h delete mode 100644 decoder/cfg_format.h delete mode 100644 decoder/cfg_options.h delete mode 100644 decoder/cfg_test.cc delete mode 100644 decoder/hg_cfg.h create mode 100755 tests/tools/flex-diff.pl create mode 100644 training/crf/mpi_adagrad_optimize.cc create mode 100644 utils/sv_test.cc delete mode 100644 utils/swap_pod.h diff --git a/.gitignore b/.gitignore index 5f573137..0dd30d25 100644 --- a/.gitignore +++ b/.gitignore @@ -37,6 +37,7 @@ decoder/Makefile.in decoder/bin/ decoder/cdec decoder/dict_test +decoder/sv_test decoder/ff_test decoder/grammar_test decoder/hg_test @@ -162,6 +163,7 @@ training/liblbfgs/bin/ training/liblbfgs/ll_test training/model1 training/mpi_batch_optimize +training/mpi_adagrad_optimize training/mpi_compute_cllh training/mpi_em_optimize training/mpi_extract_features diff --git a/decoder/Makefile.am b/decoder/Makefile.am index 8280b22c..b735756d 100644 --- a/decoder/Makefile.am +++ b/decoder/Makefile.am @@ -32,13 +32,8 @@ EXTRA_DIST = test_data rule_lexer.ll libcdec_a_SOURCES = \ JSON_parser.h \ aligner.h \ - apply_fsa_models.h \ apply_models.h \ bottom_up_parser.h \ - cfg.h \ - cfg_binarize.h \ - cfg_format.h \ - cfg_options.h \ csplit.h \ decoder.h \ earley_composer.h \ @@ -74,7 +69,6 @@ libcdec_a_SOURCES = \ freqdict.h \ grammar.h \ hg.h \ - hg_cfg.h \ hg_intersect.h \ hg_io.h \ hg_remove_eps.h \ @@ -105,7 +99,6 @@ libcdec_a_SOURCES = \ bottom_up_parser.cc \ cdec.cc \ cdec_ff.cc \ - cfg.cc \ csplit.cc \ decoder.cc \ earley_composer.cc \ diff --git a/decoder/apply_fsa_models.h b/decoder/apply_fsa_models.h deleted file mode 100644 index 6561c70c..00000000 --- a/decoder/apply_fsa_models.h +++ /dev/null @@ -1,65 +0,0 @@ -#ifndef _APPLY_FSA_MODELS_H_ -#define _APPLY_FSA_MODELS_H_ - -#include -#include -#include "feature_vector.h" -#include "named_enum.h" - -struct FsaFeatureFunction; -struct Hypergraph; -struct SentenceMetadata; -struct HgCFG; - - -#define FSA_BY(X,t) \ - X(t,BU_CUBE,) \ - X(t,BU_FULL,) \ - X(t,EARLEY,) \ - -#define FSA_BY_TYPE FsaBy - -DECLARE_NAMED_ENUM(FSA_BY) - -struct ApplyFsaBy { -/*enum { - BU_CUBE, - BU_FULL, - EARLEY, - N_ALGORITHMS - };*/ - int pop_limit; // only applies to BU_FULL so far - bool IsBottomUp() const { - return algorithm==BU_FULL || algorithm==BU_CUBE; - } - int BottomUpAlgorithm() const; - FsaBy algorithm; - std::string name() const; - friend inline std::ostream &operator << (std::ostream &o,ApplyFsaBy const& c) { - o << c.name(); - if (c.algorithm==BU_CUBE) - o << "("<= 2); int goal_id = num_nodes - 1; int pregoal = goal_id - 1; - int every = 1; - if (num_nodes > 100) every = 10; assert(in.nodes_[pregoal].out_edges_.size() == 1); if (!SILENT) cerr << " "; int has = 0; @@ -563,12 +561,14 @@ struct NoPruningRescorer { int num_nodes = in.nodes_.size(); int goal_id = num_nodes - 1; int pregoal = goal_id - 1; - int every = 1; - if (num_nodes > 100) every = 10; assert(in.nodes_[pregoal].out_edges_.size() == 1); if (!SILENT) cerr << " "; + int has = 0; for (int i = 0; i < in.nodes_.size(); ++i) { - if (!SILENT && i % every == 0) cerr << '.'; + if (!SILENT) { + int needs = (50 * i / in.nodes_.size()); + while (has < needs) { cerr << '.'; ++has; } + } ProcessOneNode(i, i == goal_id); } if (!SILENT) cerr << endl; diff --git a/decoder/cfg.cc b/decoder/cfg.cc deleted file mode 100644 index d6ee651a..00000000 --- a/decoder/cfg.cc +++ /dev/null @@ -1,656 +0,0 @@ -#include "cfg.h" -#include "hg.h" -#include "cfg_format.h" -#include "cfg_binarize.h" -#include "hash.h" -#include "batched_append.h" -#include -#include "fast_lexical_cast.hpp" -//#include "indices_after.h" -#include "show.h" -#include "null_traits.h" - -#define DUNIQ(x) x -#define DBIN(x) -#define DSP(x) x -//SP:binarize by splitting. -#define DCFG(x) IF_CFG_DEBUG(x) - -#undef CFG_FOR_RULES -#define CFG_FOR_RULES(i,expr) \ - for (CFG::NTs::const_iterator n=nts.begin(),nn=nts.end();n!=nn;++n) { \ - NT const& nt=*n; \ - for (CFG::Ruleids::const_iterator ir=nt.ruleids.begin(),er=nt.ruleids.end();ir!=er;++ir) { \ - RuleHandle i=*ir; \ - expr; \ - } \ - } - - -using namespace std; - -typedef CFG::Rule Rule; -typedef CFG::NTOrder NTOrder; -typedef CFG::RHS RHS; -typedef CFG::BinRhs BinRhs; - -/////index ruleids: -void CFG::UnindexRules() { - for (NTs::iterator n=nts.begin(),nn=nts.end();n!=nn;++n) - n->ruleids.clear(); -} - -void CFG::ReindexRules() { - UnindexRules(); - for (int i=0,e=rules.size();i Seen; // 0 = unseen, 1 = seen+finished, 2 = open (for cycle detection; seen but not finished) -enum { UNSEEN=0,SEEN,OPEN }; - -// bottom -> top topo order (rev head->tails topo) -template -struct CFGTopo { -// meaningless efficiency alternative: close over all the args except ni - so they're passed as a single pointer. also makes visiting tail_nts simpler. - CFG const& cfg; - OutOrder outorder; - std::ostream *cerrp; - CFGTopo(CFG const& cfg,OutOrder const& outorder,std::ostream *cerrp=&std::cerr) - : cfg(cfg),outorder(outorder),cerrp(cerrp) // closure over args - , seen(cfg.nts.size()) { } - - Seen seen; - void operator()(CFG::NTHandle ni) { - char &seenthis=seen[ni]; - if (seenthis==UNSEEN) { - seenthis=OPEN; - - CFG::NT const& nt=cfg.nts[ni]; - for (CFG::Ruleids::const_iterator i=nt.ruleids.begin(),e=nt.ruleids.end();i!=e;++i) { - Rule const& r=cfg.rules[*i]; - r.visit_rhs_nts(*this); // recurse. - } - - *outorder++=ni; // dfs finishing time order = reverse topo. - seenthis=SEEN; - } else if (cerrp && seenthis==OPEN) { - std::ostream &cerr=*cerrp; - cerr<<"WARNING: CFG Topo order attempt failed: NT "; - cfg.print_nt_name(cerr,ni); - cerr<<" already reached from goal(top) "; - cfg.print_nt_name(cerr,cfg.goal_nt); - cerr<<". Continuing to reorder, but it's not fully topological.\n"; - } - } - -}; - -template -void DoCFGTopo(CFG const& cfg,CFG::NTHandle goal,O const& o,std::ostream *w=0) { - CFGTopo ct(cfg,o,w); - ct(goal); -} - -}//ns - -// you would need to do this only if you didn't build from hg, or you Binarize without bin_topo option. note: this doesn't sort the list of rules; it's assumed that if you care about the topo order you'll iterate over nodes. -void CFG::OrderNTsTopo(NTOrder *o_,std::ostream *cycle_complain) { - NTOrder &o=*o_; - o.resize(nts.size()); - DoCFGTopo(*this,goal_nt,o.begin(),cycle_complain); -} - - -/////sort/uniq: -namespace { -RHS null_rhs(1,INT_MIN); - -//sort -struct ruleid_best_first { - CFG::Rules const* rulesp; - bool operator()(int a,int b) const { // true if a >(prob for ruleid) b - return (*rulesp)[b].p < (*rulesp)[a].p; - } -}; - -//uniq -struct prob_pos { - prob_pos() {} - prob_pos(prob_t prob,int pos) : prob(prob),pos(pos) {} - prob_t prob; - int pos; - bool operator <(prob_pos const& o) const { return prob > BestRHS; // faster to use trie? maybe. - BestRHS bestp; // once inserted, the position part (output index) never changes. but the prob may be improved (overwrite ruleid at that position). - HASH_MAP_EMPTY(bestp,null_rhs); - Ruleids &adj=nts[ni].ruleids; - Ruleids oldadj=adj; - int newpos=0; - for (int i=0,e=oldadj.size();i!=e;++i) { // this beautiful complexity is to ensure that adj' is a subsequence of adj (without duplicates) - int ri=oldadj[i]; - Rule const& r=rules[ri]; - prob_pos pi(r.p,newpos); - prob_pos &oldpi=get_default(bestp,r.rhs,pi); - if (oldpi.pos==newpos) {// newly inserted - adj[newpos++]=ri; - } else { - SHOWP(DUNIQ,"Uniq duplicate: ") SHOW4(DUNIQ,oldpi.prob,pi.prob,oldpi.pos,newpos); - SHOW(DUNIQ,ShowRule(ri)); - SHOW(DUNIQ,ShowRule(adj[oldpi.pos])); - if (oldpi.prob::min(),std::numeric_limits::min()); - -// index i >= N.size()? then it's in M[i-N.size()] -//WordID first,WordID second, -string BinStr(BinRhs const& b,CFG::NTs const& N,CFG::NTs const& M) -{ - int nn=N.size(); - ostringstream o; -#undef BinNameOWORD -#define BinNameOWORD(w) \ - do { \ - int n=w; if (n>0) o << TD::Convert(n); \ - else { \ - int i=-n; \ - if (i -struct null_for; - - -template <> -struct null_for { - static BinRhs null; -}; - -template <> -struct null_for { - static RHS null; -}; - -template <> -BinRhs null_traits::xnull(std::numeric_limits::min(),std::numeric_limits::min()); - -template <> -RHS null_traits::xnull(1,std::numeric_limits::min()); -*/ - -template -struct add_virtual_rules { - typedef CFG::RuleHandle RuleHandle; - typedef CFG::NTHandle NTHandle; - CFG::NTs &nts,new_nts; - CFG::Rules &rules, new_rules; -// above will be appended at the end, so we don't have to worry about iterator invalidation - WordID newnt; //negative of NTHandle, or positive => unary lexical item (not to binarize). fit for rhs of a rule - RuleHandle newruleid; - typedef HASH_MAP > R2L; - R2L rhs2lhs; // an rhs maps to this -virtntid, or original id if length 1 - bool name_nts; - add_virtual_rules(CFG &cfg,bool name_nts=false) : nts(cfg.nts),rules(cfg.rules),newnt(-nts.size()),newruleid(rules.size()),name_nts(name_nts) { - HASH_MAP_EMPTY(rhs2lhs,null_traits::xnull); - } - NTHandle get_virt(Rhs const& r) { - NTHandle nt=get_default(rhs2lhs,r,newnt); - SHOW(DBIN,newnt) SHOWP(DBIN,"bin="<") SHOW(DBIN,nt); - if (newnt==nt) { - create(r); - } - return nt; - } - inline void set_nt_name(Rhs const& r) { - if (name_nts) - new_nts.back().from.nt=BinName(r,nts,new_nts); - } - inline void create_nt(Rhs const& rhs) { - new_nts.push_back(CFG::NT(newruleid++)); - set_nt_name(rhs); - } - inline void create_rule(Rhs const& rhs) { - new_rules.push_back(CFG::Rule(-newnt,rhs)); - --newnt; - } - inline void create_adding(Rhs const& rhs) { - NTHandle nt=get_default(rhs2lhs,rhs,newnt); - assert(nt==newnt); - create(rhs); - } - inline void create(Rhs const& rhs) { - SHOWP(DSP,"Create ") SHOW3(DSP,newnt,newruleid,BinStr(rhs,nts,new_nts)) - create_nt(rhs); - create_rule(rhs); - assert(newruleid==rules.size()+new_rules.size());assert(-newnt==nts.size()+new_nts.size()); - } - - ~add_virtual_rules() { - append_rules(); - } - void append_rules() { - // marginally more efficient - batched_append_swap(nts,new_nts); - batched_append_swap(rules,new_rules); - } - inline bool have(Rhs const& rhs,NTHandle &h) const { - if (rhs.size()==1) { // stop creating virtual unary rules. - h=rhs[0]; - return true; - } - typename R2L::const_iterator i=rhs2lhs.find(rhs); - if (i==rhs2lhs.end()) - return false; - h=i->second; - return true; - } - //HACK: prevent this for instantiating for BinRhs. we need to use rule index because we'll be adding rules before we can update. - // returns 1 per replaced NT (0,1, or 2) - inline std::string Str(Rhs const& rhs) const { - return BinStr(rhs,nts,new_nts); - } - - template - int split_rhs(RHSi &rhs,bool only_free=false,bool only_reusing_1=false) { - typedef WordID const* WP; - //TODO: don't actually build substrings of rhs; define key type that stores ref to rhs in new_nts by index (because it may grow), and also allows an int* [b,e) range to serve as key (i.e. don't insert the latter type of key). - int n=rhs.size(); - if (n<=2) return 0; - int longest1=1; // all this other stuff is not uninitialized when used, based on checking this and other things (it's complicated, learn to prove theorems, gcc) - int mid=n/2; - int best_k; - enum {HAVE_L=-1,HAVE_NONE=0,HAVE_R=1}; - int have1=HAVE_NONE; // will mean we already have some >1 length prefix or suffix as a virt. (it's free). if we have both we use it immediately and return. - NTHandle ntr,ntl; - NTHandle bestntr,bestntl; - WP b=&rhs.front(),e=b+n; - WP wk=b; - SHOWM3(DSP,"Split",Str(rhs),only_free,only_reusing_1); - int rlen=n; - for (int k=1;k1) { SHOWM3(DSP,"Have l",k,n,Str(l)) } - Rhs r(wk,e); - if (have(r,ntr)) { - SHOWM3(DSP,"Have r too",k,n,Str(r)) - rhs.resize(2); - rhs[0]=ntl; - rhs[1]=ntr; - return 2; - } else if (k>longest1) { - longest1=k; - have1=HAVE_L; - bestntl=ntl; - best_k=k; - } - } else if (rlen>longest1) { // > or >= favors l or r branching, maybe. who cares. - Rhs r(wk,e); - if (have(r,ntr)) { - longest1=rlen; - if (rlen>1) { SHOWM3(DSP,"Have r (only) ",k,n,Str(r)) } - have1=HAVE_R; - bestntr=ntr; - best_k=k; - } - } - //TODO: swap order of consideration (l first or r?) depending on pre/post midpoint? one will be useless to check for beating the longest single match so far. check that second - - } - // now we know how we're going to split the rule; what follows is just doing the actual splitting: - - if (only_free) { - if (have1==HAVE_NONE) - return 0; - if (have1==HAVE_L) { - rhs.erase(rhs.begin()+1,rhs.begin()+best_k); //erase [1..best_k) - rhs[0]=bestntl; - } else { - assert(have1==HAVE_R); - rhs.erase(rhs.begin()+best_k+1,rhs.end()); // erase (best_k..) - rhs[best_k]=bestntr; - } - return 1; - } - /* now we have to add some new virtual rules. - some awkward constraints: - - 1. can't resize rhs until you save copy of l or r split portion - - 2. can't create new rule until you finished modifying rhs (this is why we store newnt then create). due to vector push_back invalidation. perhaps we could bypass this by reserving sufficient space first before a splitting pass (# rules and nts created is <= 2 * # of rules being passed over) - - */ - if (have1==HAVE_NONE) { // default: split down middle. - DSP(assert(longest1==1)); - WP m=b+mid; - if (n%2==0) { - WP i=b; - WP j=m; - for (;i!=m;++i,++j) - if (*i!=*j) goto notleqr; - // [...mid]==[mid...]! - RHS l(b,m); // 1. // this is equal to RHS(m,e). - rhs.resize(2); - rhs[0]=rhs[1]=newnt; //2. - create_adding(l); - return 1; // only had to create 1 total when splitting down middle when l==r - } - notleqr: - if (only_reusing_1) return 0; - best_k=mid; // rounds down - if (mid==1) { - RHS r(m,e); //1. - rhs.resize(2); - rhs[1]=newnt; //2. - create_adding(r); - return 1; - } else { - Rhs l(b,m); - Rhs r(m,e); // 1. - rhs.resize(2); - rhs[0]=newnt; - rhs[1]=newnt-1; // 2. - create_adding(l); - create_adding(r); - return 2; - } - } - WP best_wk=b+best_k; - //we build these first because adding rules may invalidate the underlying pointers (we end up binarizing already split virt rules)!. - //wow, that decision (not to use index into new_nts instead of pointer to rhs), while adding new nts to it really added some pain. - if (have1==HAVE_L) { - Rhs r(best_wk,e); //1. - rhs.resize(2); - rhs[0]=bestntl; - DSP(assert(best_wkb+1)); // because we would have returned having both if lhs was singleton - Rhs l(b,best_wk); //1. - rhs.resize(2); - rhs[0]=newnt; //2. - rhs[1]=bestntr; - create_adding(l); - } - return 1; - } -}; - -}//ns - -void CFG::BinarizeSplit(CFGBinarize const& b) { - add_virtual_rules v(*this,b.bin_name_nts); - CFG_FOR_RULES(i,v.split_rhs(rules[i].rhs,false,false)); - Rules &newr=v.new_rules; -#undef CFG_FOR_VIRT -#define CFG_FOR_VIRT(r,expr) \ - for (int i=0,e=newr.size();i0) - BinarizeThresh(b); - if (b.bin_split) - BinarizeSplit(b); - if (b.bin_l2r) - BinarizeL2R(false,b.bin_name_nts); - if (b.bin_topo) //TODO: more efficient (at least for l2r) maintenance of order? - OrderNTsTopo(); - -} - -namespace { -} - -void CFG::BinarizeThresh(CFGBinarize const& b) { - throw runtime_error("TODO: some fancy linked list thing - see NOTES.partial.binarize"); -} - - -void CFG::BinarizeL2R(bool bin_unary,bool name) { - add_virtual_rules v(*this,name); -cerr << "Binarizing left->right " << (bin_unary?"real to unary":"stop at binary") < > bin2lhs; // we're going to hash cons rather than build an explicit trie from right to left. - HASH_MAP_EMPTY(bin2lhs,null_bin_rhs); - // iterate using indices and not iterators because we'll be adding to both nts and rules list? we could instead pessimistically reserve space for both, but this is simpler. also: store original end of nts since we won't need to reprocess newly added ones. - int rhsmin=bin_unary?0:1; - //NTs new_nts; - //Rules new_rules; - //TODO: this could be factored easily into in-place (append to new_* like below) and functional (nondestructive copy) versions (copy orig to target and append to target) -// int newnt=nts.size(); // we're going to store binary rhs with -nt to keep distinct from words (>=0) -// int newruleid=rules.size(); - BinRhs bin; - CFG_FOR_RULES(ruleid, -/* for (NTs::const_iterator n=nts.begin(),nn=nts.end();n!=nn;++n) { - NT const& nt=*n; - for (Ruleids::const_iterator ir=nt.ruleids.begin(),er=nt.ruleids.end();ir!=er;++ir) { - RuleHandle ruleid=*ir;*/ -// SHOW2(DBIN,ruleid,ShowRule(ruleid)); - Rule & rule=rules[ruleid]; - RHS &rhs=rule.rhs; // we're going to binarize this while adding newly created rules to new_... - if (rhs.empty()) continue; - int r=rhs.size()-2; // loop below: [r,r+1) is to be reduced into a (maybe new) binary NT - if (rhsmin<=r) { // means r>=0 also - bin.second=rhs[r+1]; - int bin_to; // the replacement for bin -// assert(newruleid==rules.size()+new_rules.size());assert(newnt==nts.size()+new_nts.size()); - // also true at start/end of loop: - for (;;) { // pairs from right to left (normally we leave the last pair alone) - bin.first=rhs[r]; - bin_to=v.get_virt(bin); -/* bin_to=get_default(bin2lhs,bin,v.newnt); -// SHOW(DBIN,r) SHOW(DBIN,newnt) SHOWP(DBIN,"bin="<") SHOW(DBIN,bin_to); - if (v.newnt==bin_to) { // it's new! - new_nts.push_back(NT(newruleid++)); - //now newnt is the index of the last (after new_nts is appended) nt. bin is its rhs. bin_to is its lhs - new_rules.push_back(Rule(newnt,bin)); - ++newnt; - if (name) new_nts.back().from.nt=BinName(bin,nts,new_nts); - } -*/ - bin.second=bin_to; - --r; - if (r const& rule_rhs=target_side?er.e():er.f(); - int nr=rule_rhs.size(); - RHS &rhs_out=cfgr.rhs; - rhs_out.resize(nr); - Hypergraph::TailNodeVector const& tails=e.tail_nodes_; - int nvar=0; - //split out into separate target_side, source_side loops? - for (int j=0;j0) - rhs_out[j]=w; - else { - int n=nt_index(nvar,tails,target_side,w); - ++nvar; - if (push_weights) crp*=np[n]; - rhs_out[j]=-n; - } - } - assert(nvar==er.Arity()); - assert(nvar==tails.size()); - } -} - -void CFG::Clear() { - rules.clear(); - nts.clear(); - goal_nt=-1; - hg_=0; -} - -namespace { -CFGFormat form; -} - -void CFG::PrintRule(std::ostream &o,RuleHandle rulei,CFGFormat const& f) const { - Rule const& r=rules[rulei]; - f.print_lhs(o,*this,r.lhs); - f.print_rhs(o,*this,r.rhs.begin(),r.rhs.end()); - f.print_features(o,r.p,r.f); - IF_CFG_TRULE(if (r.rule) o<0) then -w is the NT index. also, non-synchronous ops like binarization make sense. hg is a somewhat bulky encoding of non-synchronous forest - - using indices to refer to NTs saves space (32 bit index vs 64 bit pointer) and allows more efficient ancillary maps for e.g. chart info (if we used pointers to actual node structures, it would be tempting to add various void * or other slots for use by mapped-during-computation ephemera) - */ - -#include -#include -#include -#include "feature_vector.h" -#include "small_vector.h" -#include "wordid.h" -#include "tdict.h" -#include "trule.h" -#include "prob.h" -//#include "int_or_pointer.h" -#include "small_vector.h" -#include "nt_span.h" -#include -#include "indices_after.h" -#include - -class Hypergraph; -class CFGFormat; // #include "cfg_format.h" -class CFGBinarize; // #include "cfg_binarize.h" - -#undef CFG_MUST_EQ -#define CFG_MUST_EQ(f) if (!(o.f==f)) return false; - -struct CFG { - typedef int RuleHandle; - typedef int NTHandle; - typedef SmallVector RHS; // same as in trule rhs: >0 means token, <=0 means -node index (not variable index) - typedef std::vector Ruleids; - - void print_nt_name(std::ostream &o,NTHandle n) const { - o << nts[n].from << n; - } - std::string nt_name(NTHandle n) const { - std::ostringstream o; - print_nt_name(o,n); - return o.str(); - } - void print_rhs_name(std::ostream &o,WordID w) const { - if (w<=0) print_nt_name(o,-w); - else o< BinRhs; - - struct Rule { - std::size_t hash_impl() const { - using namespace boost; - std::size_t h=lhs; - hash_combine(h,rhs); - hash_combine(h,p); - hash_combine(h,f); - return h; - } - bool operator ==(Rule const &o) const { - CFG_MUST_EQ(lhs) - CFG_MUST_EQ(rhs) - CFG_MUST_EQ(p) - CFG_MUST_EQ(f) - return true; - } - inline bool operator!=(Rule const& o) const { return !(o==*this); } - - // for binarizing - no costs/probs - Rule() : lhs(-1) { } - bool is_null() const { return lhs<0; } - void set_null() { lhs=-1; rhs.clear();f.clear(); IF_CFG_TRULE(rule.reset();) } - - Rule(int lhs,BinRhs const& binrhs) : lhs(lhs),rhs(2),p(1) { - rhs[0]=binrhs.first; - rhs[1]=binrhs.second; - } - Rule(int lhs,RHS const& rhs) : lhs(lhs),rhs(rhs),p(1) { - } - - int lhs; // index into nts - RHS rhs; - prob_t p; // h unused for now (there's nothing admissable, and p is already using 1st pass inside as pushed toward top) - SparseVector f; // may be empty, unless copy_features on Init - IF_CFG_TRULE(TRulePtr rule;) - int size() const { // for stats only - return rhs.size(); - } - void Swap(Rule &o) { - using namespace std; - swap(lhs,o.lhs); - swap(rhs,o.rhs); - swap(p,o.p); - swap(f,o.f); - IF_CFG_TRULE(swap(rule,o.rule);) - } - friend inline void swap(Rule &a,Rule &b) { - a.Swap(b); - } - - template - void visit_rhs_nts(V &v) const { - for (RHS::const_iterator i=rhs.begin(),e=rhs.end();i!=e;++i) { - WordID w=*i; - if (w<=0) - v(-w); - } - } - template - void visit_rhs_nts(V const& v) const { - for (RHS::const_iterator i=rhs.begin(),e=rhs.end();i!=e;++i) { - WordID w=*i; - if (w<=0) - v(-w); - } - } - - template - void visit_rhs(V &v) const { - for (RHS::const_iterator i=rhs.begin(),e=rhs.end();i!=e;++i) { - WordID w=*i; - if (w<=0) - v.visit_nt(-w); - else - v.visit_t(w); - } - } - - // returns 0 or 1 (# of non null rules in this rule). - template - bool reorder_from(O &order,NTHandle removed=-1) { - for (RHS::iterator i=rhs.begin(),e=rhs.end();i!=e;++i) { - WordID &w=*i; - if (w<=0) { - int oldnt=-w; - NTHandle newnt=(NTHandle)order[oldnt]; // e.g. unsigned to int (-1) conversion should be ok - if (newnt==removed) { - set_null(); - return false; - } - w=-newnt; - } - } - return true; - } - }; - - struct NT { - NT() { } - explicit NT(RuleHandle r) : ruleids(1,r) { } - std::size_t hash_impl() const { using namespace boost; return hash_value(ruleids); } - bool operator ==(NT const &o) const { - return ruleids==o.ruleids; // don't care about from - } - inline bool operator!=(NT const& o) const { return !(o==*this); } - Ruleids ruleids; // index into CFG rules with lhs = this NT. aka in_edges_ - NTSpan from; // optional name - still needs id to disambiguate - void Swap(NT &o) { - using namespace std; - swap(ruleids,o.ruleids); - swap(from,o.from); - } - friend inline void swap(NT &a,NT &b) { - a.Swap(b); - } - }; - - CFG() : hg_() { uninit=true; } - - // provided hg will have weights pushed up to root - CFG(Hypergraph const& hg,bool target_side=true,bool copy_features=false,bool push_weights=true) { - Init(hg,target_side,copy_features,push_weights); - } - bool Uninitialized() const { return uninit; } - void Clear(); - bool Empty() const { return nts.empty(); } - void UnindexRules(); // save some space? - void ReindexRules(); // scan over rules and rebuild NT::ruleids (e.g. after using UniqRules) - int UniqRules(NTHandle ni); // keep only the highest prob rule for each rhs and lhs=nt - doesn't remove from Rules; just removes from nts[ni].ruleids. keeps the same order in this sense: for a given signature (rhs), that signature's first representative in the old ruleids will become the new position of the best. as a consequence, if you SortLocalBestFirst() then UniqRules(), the result is still best first. but you may also call this on unsorted ruleids. returns number of rules kept - inline int UniqRules() { - int nkept=0; - for (int i=0,e=nts.size();i!=e;++i) nkept+=UniqRules(i); - return nkept; - } - int rules_size() const { - const int sz=rules.size(); - int sum=sz; - for (int i=0;i NTOrder; // a list of nts, in definition-before-use order. - - //perhaps: give up on templated Order and move the below to .cc (NTOrder should be fine) - - // post: iterating nts 0,1... the same as order[0],order[1],... ; return number of non-null rules (these aren't actually deleted) - // pre: order is (without duplicates) a range of NTHandle - template - int ReorderNTs(Order const& order) { - using namespace std; - int nn=nts.size(); -#if 0 - NTs newnts(order.size()); // because the (sub)permutation order may have e.g. 1<->4 - int ni=0; - for (typename Order::const_iterator i=order.begin(),e=order.end();i!=e;++i) { - assert(*i - int RemapRules(NTHandleRemap const& remap_nti,NTHandle removed=-1) { - int n_non_null=0; - for (int i=0,e=rules.size();i - void VisitRuleIds(V &v) { - for (int i=0,e=nts.size();i - void VisitRuleIds(V const& v) { - for (int i=0,e=nts.size();i - void VisitRulesUnindexed(V const &v) { - for (int i=0,e=rules.size();irhs. - // you would need to do this only if you didn't build from hg, or you Binarize without bin_topo option. - // note: this doesn't sort the list of rules; it's assumed that if you care about the topo order you'll iterate over nodes. - // cycle_complain means to warn in case of back edges. it's not necessary to prevent inf. loops. you get some order that's not topo if there are loops. starts from goal_nt, of course. - - void OrderNTsTopo(std::ostream *cycle_complain=0) { - NTOrder o; - OrderNTsTopo(&o,cycle_complain); - ReorderNTs(o); - } - - void BinarizeL2R(bool bin_unary=false,bool name_nts=false); - void Binarize(CFGBinarize const& binarize_options); // see cfg_binarize.h for docs - void BinarizeSplit(CFGBinarize const& binarize_options); - void BinarizeThresh(CFGBinarize const& binarize_options); // maybe unbundle opts later - - typedef std::vector NTs; - NTs nts; - typedef std::vector Rules; - Rules rules; - int goal_nt; - prob_t goal_inside,pushed_inside; // when we push viterbi weights to goal, we store the removed probability in pushed_inside -protected: - bool uninit; - Hypergraph const* hg_; // shouldn't be used for anything, esp. after binarization - // rules/nts will have same index as hg edges/nodes -}; - -inline std::size_t hash_value(CFG::Rule const& r) { - return r.hash_impl(); -} - -inline std::size_t hash_value(CFG::NT const& r) { - return r.hash_impl(); -} - -inline void swap(CFG &a,CFG &b) { - a.Swap(b); -} - -std::ostream &operator<<(std::ostream &o,CFG const &x); - -#endif diff --git a/decoder/cfg_binarize.h b/decoder/cfg_binarize.h deleted file mode 100644 index ae06f8bf..00000000 --- a/decoder/cfg_binarize.h +++ /dev/null @@ -1,90 +0,0 @@ -#ifndef CFG_BINARIZE_H -#define CFG_BINARIZE_H - -#include - -/* - binarization: decimate rhs of original rules until their rhs have been reduced to length 2 (or 1 if bin_unary). also decimate rhs of newly binarized rules until length 2. newly created rules are all binary (never unary/nullary). - - bin_name_nts: nts[i].from will be initialized, including adding new names to TD - - bin_l2r: right-branching (a (b c)) means suffixes are shared. if requested, the only other option that matters is bin_unary - - otherwise, greedy binarization: the pairs that are most frequent in the rules are binarized, one at a time. this should be done efficiently: each pair has a count of and list of its left and right adjacent pair+count (or maybe a non-count collapsed list of adjacent instances). this can be efficiently updated when a pair is chosen for replacement by a new virtual NT. - */ - -struct CFGBinarize { - int bin_thresh; - bool bin_l2r; - int bin_unary; - bool bin_name_nts; - bool bin_topo; - bool bin_split; - int split_passes,split_share1_passes,split_free_passes; - template // template to support both printable_opts and boost nonprintable - void AddOptions(Opts *opts) { - opts->add_options() - ("cfg_binarize_threshold", defaulted_value(&bin_thresh),"(if >0) repeatedly binarize CFG rhs bigrams which appear at least this many times, most frequent first. resulting rules may be 1,2, or >2-ary. this happens before the other types of binarization.") -// ("cfg_binarize_unary_threshold", defaulted_value(&bin_unary),"if >0, a rule-completing production A->BC may be binarized as A->U U->BC if U->BC would be used at least this many times. this happens last.") - ("cfg_binarize_split", defaulted_value(&bin_split),"(DeNero et al) for each rule until binarized, pick a split point k of L->r[0..n) to make rules L->V1 V2, V1->r[0..k) V2->r[k..n), to minimize the number of new rules created") - ("cfg_split_full_passes", defaulted_value(&split_passes),"pass through the virtual rules only (up to) this many times (all real rules will have been split if not already binary)") - ("cfg_split_share1_passes", defaulted_value(&split_share1_passes),"after the full passes, for up to this many times split when at least 1 of the items has been seen before") - ("cfg_split_free_passes", defaulted_value(&split_free_passes),"only split off from virtual nts pre/post nts that already exist - could check for interior phrases but after a few splits everything should be tiny already.") - ("cfg_binarize_l2r", defaulted_value(&bin_l2r),"force left to right (a (b (c d))) binarization (ignore _at threshold)") - ("cfg_binarize_name_nts", defaulted_value(&bin_name_nts),"create named virtual NT tokens e.g. 'A12+the' when binarizing 'B->[A12] the cat'") - ("cfg_binarize_topo", defaulted_value(&bin_topo),"reorder nonterminals after binarization to maintain definition before use (topological order). otherwise the virtual NTs will all appear after the regular NTs") - ; - } - void Validate() { - if (bin_thresh>0&&!bin_l2r) { -// std::cerr<<"\nWARNING: greedy binarization not yet supported; using l2r (right branching) instead.\n"; -// bin_l2r=true; - } - if (false && bin_l2r && bin_split) { // actually, split may be slightly incomplete due to finite number of passes. - std::cerr<<"\nWARNING: l2r and split are both complete binarization and redundant. Using split.\n"; - bin_l2r=false; - } - - } - - bool Binarizing() const { - return bin_split || bin_l2r || bin_thresh>0; - } - void set_defaults() { - bin_split=false; - bin_topo=false; - bin_thresh=0; - bin_unary=0; - bin_name_nts=true; - bin_l2r=false; - split_passes=10;split_share1_passes=0;split_free_passes=10; - } - CFGBinarize() { set_defaults(); } - void print(std::ostream &o) const { - o<<'('; - if (!Binarizing()) - o << "Unbinarized"; - else { - if (bin_unary) - o << "unary-sharing "; - if (bin_thresh) - o<<"greedy bigram count>="<right"; - else - o << "DeNero greedy split"; - if (bin_name_nts) - o << " named-NTs"; - if (bin_topo) - o<<" preserve-topo-order"; - } - o<<')'; - } - friend inline std::ostream &operator<<(std::ostream &o,CFGBinarize const& me) { - me.print(o); return o; - } - -}; - - -#endif diff --git a/decoder/cfg_format.h b/decoder/cfg_format.h deleted file mode 100644 index d12da261..00000000 --- a/decoder/cfg_format.h +++ /dev/null @@ -1,134 +0,0 @@ -#ifndef CFG_FORMAT_H -#define CFG_FORMAT_H - -#include -#include -#include "wordid.h" -#include "feature_vector.h" -#include "program_options.h" - -struct CFGFormat { - bool identity_scfg; - bool features; - bool logprob_feat; - bool comma_nt; - bool nt_span; - std::string goal_nt_name; - std::string nt_prefix; - std::string logprob_feat_name; - std::string partsep; - bool goal_nt() const { return !goal_nt_name.empty(); } - template // template to support both printable_opts and boost nonprintable - void AddOptions(Opts *opts) { - //using namespace boost::program_options; - //using namespace std; - opts->add_options() - ("identity_scfg",defaulted_value(&identity_scfg),"output an identity SCFG: add an identity target side - '[X12] ||| [X13,1] a ||| [1] a ||| feat= ...' - the redundant target '[1] a |||' is omitted otherwise.") - ("features",defaulted_value(&features),"print the CFG feature vector") - ("logprob_feat",defaulted_value(&logprob_feat),"print a LogProb=-1.5 feature irrespective of --features.") - ("logprob_feat_name",defaulted_value(&logprob_feat_name),"alternate name for the LogProb feature") - ("cfg_comma_nt",defaulted_value(&comma_nt),"if false, omit the usual [NP,1] ',1' variable index in the source side") - ("goal_nt_name",defaulted_value(&goal_nt_name),"if nonempty, the first production will be '[goal_nt_name] ||| [x123] ||| LogProb=y' where x123 is the actual goal nt, and y is the pushed prob, if any") - ("nt_prefix",defaulted_value(&nt_prefix),"NTs are [123] where 123 is the node number starting at 0, and the highest node (last in file) is the goal node in an acyclic hypergraph") - ("nt_span",defaulted_value(&nt_span),"prefix A(i,j) for NT coming from hypergraph node with category A on span [i,j). this is after --nt_prefix if any") - ; - } - - void print(std::ostream &o) const { - o<<"["; - if (identity_scfg) - o<<"Identity SCFG "; - if (features) - o<<"+Features "; - if (logprob_feat) - o< - void print_source_nt(std::ostream &o,CFG const&cfg,int id,int position=1) const { - o<<'['; - print_nt_name(o,cfg,id); - if (comma_nt) o<<','< - void print_nt_name(std::ostream &o,CFG const& cfg,int id) const { - o< - void print_lhs(std::ostream &o,CFG const& cfg,int id) const { - o<<'['; - print_nt_name(o,cfg,id); - o<<']'; - } - - template - void print_rhs(std::ostream &o,CFG const&cfg,Iter begin,Iter end) const { - o<0) o << TD::Convert(w); - else print_source_nt(o,cfg,-w,++pos); - } - if (identity_scfg) { - o<0) o << TD::Convert(w); - else o << '['<<++pos<<']'; - } - } - } - - void print_features(std::ostream &o,prob_t p,SparseVector const& fv=SparseVector()) const { - bool logp=(logprob_feat && p!=prob_t::One()); - if (features || logp) { - o << partsep; - if (logp) - o << logprob_feat_name<<'='< // template to support both printable_opts and boost nonprintable - void AddOptions(Opts *opts) { - opts->add_options() - ("cfg_output", defaulted_value(&out),"write final target CFG (before FSA rescoring) to this file") - ("source_cfg_output", defaulted_value(&source_out),"write source CFG (after prelm-scoring, prelm-prune) to this file") - ("cfg_unbin_output", defaulted_value(&unbin_out),"write pre-binarization CFG to this file") //TODO: - ("cfg_uniq", defaulted_value(&uniq),"in case of duplicate rules, keep only the one with highest prob") - - ; - binarize.AddOptions(opts); - format.AddOptions(opts); - } - void Validate() { - format.Validate(); - binarize.Validate(); - } - void maybe_output_source(Hypergraph const& hg) { - if (source_out.empty()) return; - std::cerr<<"Printing source CFG to "< -#include -#include "cfg.h" -#include "hg_test.h" -#include "cfg_options.h" -#include "show.h" - -/* TODO: easiest way to get meaningful confirmations that things work: implement conversion back to hg, and compare viterbi/inside etc. stats for equality to original hg. or you can define CSHOW_V and see lots of output */ - -using namespace boost; - -#define CSHOW_V 0 - -#if CSHOW_V -# define CSHOWDO(x) x; -#else -# define CSHOWDO(x) -#endif -#define CSHOW(x) CSHOWDO(cerr<<#x<<'='< HgW; // hg file,weights - -struct CFGTest : public TestWithParam { - string hgfile; - Hypergraph hg; - CFG cfg; - CFGFormat form; - SparseVector weights; - - static void JsonFN(Hypergraph &hg,CFG &cfg,SparseVector &featw,std::string file - ,std::string const& wts="Model_0 1 EgivenF 1 f1 1") - { - istringstream ws(wts); - EXPECT_TRUE(ws>>featw); - CSHOW(featw) - std::string path(boost::unit_test::framework::master_test_suite().argc == 2 ? boost::unit_test::framework::master_test_suite().argv[1] : TEST_DATA); - HGSetup::JsonTestFile(&hg,path,file); - hg.Reweight(featw); - cfg.Init(hg,true,true,false); - } - static void SetUpTestCase() { - } - static void TearDownTestCase() { - } - CFGTest() { - hgfile=GetParam().first; - JsonFN(hg,cfg,weights,hgfile,GetParam().second); - CSHOWDO(cerr<<"\nCFG Test: ") - CSHOW(hgfile); - form.nt_span=true; - form.comma_nt=false; - } - ~CFGTest() { } -}; - -TEST_P(CFGTest,Binarize) { - CFGBinarize b; - b.bin_name_nts=1; - CFG cfgu=cfg; - EXPECT_EQ(cfgu,cfg); - int nrules=cfg.rules.size(); - CSHOWDO(cerr<<"\nUniqing: "<=0) { - int f=i<<1; - b.bin_l2r=1; - b.bin_unary=(f>>=1)&1; - b.bin_topo=(f>>=1)&1; - uniq=(f>>=1)&1; - } else - b.bin_l2r=0; - CFG cc=uniq?cfgu:cfg; - CSHOW("\nBinarizing "<<(uniq?"uniqued ":"")<<": "< "< #include -#include "program_options.h" #include "stringlib.h" #include "weights.h" #include "filelib.h" @@ -49,13 +48,6 @@ namespace std { using std::tr1::unordered_map; } #include "hg_io.h" #include "aligner.h" -#undef FSA_RESCORING -#ifdef FSA_RESCORING -#include "hg_cfg.h" -#include "apply_fsa_models.h" -#include "cfg_options.h" -#endif - #ifdef CP_TIME clock_t CpTime::time_; void CpTime::Add(clock_t x){time_+=x;} @@ -140,21 +132,6 @@ inline boost::shared_ptr make_ff(string const& ffp,bool verbose return pf; } -#ifdef FSA_RESCORING -inline boost::shared_ptr make_fsa_ff(string const& ffp,bool verbose_feature_functions,char const* pre="") { - string ff, param; - SplitCommandAndParam(ffp, &ff, ¶m); - cerr << "FSA Feature: " << ff; - if (param.size() > 0) cerr << " (with config parameters '" << param << "')\n"; - else cerr << " (no config parameters)\n"; - boost::shared_ptr pf = fsa_ff_registry.Create(ff, param); - if (!pf) exit(1); - if (verbose_feature_functions && !SILENT) - cerr<<"State is "<state_bytes()<<" bytes for "< translator; boost::shared_ptr > init_weights; // weights used with initial parse vector > pffs; -#ifdef FSA_RESCORING - CFGOptions cfg_options; - vector > fsa_ffs; - vector fsa_names; -#endif boost::shared_ptr > rng; int sample_max_trans; bool aligner_mode; @@ -324,7 +296,6 @@ struct DecoderImpl { SparseVector acc_vec; // accumulate gradient double acc_obj; // accumulate objective int g_count; // number of gradient pieces computed - int pop_limit; bool csplit_output_plf; bool write_gradient; // TODO Observer bool feature_expectations; // TODO Observer @@ -372,6 +343,7 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream ("weights,w",po::value(),"Feature weights file (initial forest / pass 1)") ("feature_function,F",po::value >()->composing(), "Pass 1 additional feature function(s) (-L for list)") ("intersection_strategy,I",po::value()->default_value("cube_pruning"), "Pass 1 intersection strategy for incorporating finite-state features; values include Cube_pruning, Full, Fast_cube_pruning, Fast_cube_pruning_2") + ("cubepruning_pop_limit,K",po::value()->default_value(200), "Max number of pops from the candidate heap at each node") ("summary_feature", po::value(), "Compute a 'summary feature' at the end of the pass (before any pruning) with name=arg and value=inside-outside/Z") ("summary_feature_type", po::value()->default_value("node_risk"), "Summary feature types: node_risk, edge_risk, edge_prob") ("density_prune", po::value(), "Pass 1 pruning: keep no more than this many times the number of edges used in the best derivation tree (>=1.0)") @@ -380,6 +352,7 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream ("weights2",po::value(),"Optional pass 2") ("feature_function2",po::value >()->composing(), "Optional pass 2") ("intersection_strategy2",po::value()->default_value("cube_pruning"), "Optional pass 2") + ("cubepruning_pop_limit2",po::value()->default_value(200), "Optional pass 2") ("summary_feature2", po::value(), "Optional pass 2") ("density_prune2", po::value(), "Optional pass 2") ("beam_prune2", po::value(), "Optional pass 2") @@ -387,18 +360,14 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream ("weights3",po::value(),"Optional pass 3") ("feature_function3",po::value >()->composing(), "Optional pass 3") ("intersection_strategy3",po::value()->default_value("cube_pruning"), "Optional pass 3") + ("cubepruning_pop_limit3",po::value()->default_value(200), "Optional pass 3") ("summary_feature3", po::value(), "Optional pass 3") ("density_prune3", po::value(), "Optional pass 3") ("beam_prune3", po::value(), "Optional pass 3") -#ifdef FSA_RESCORING - ("fsa_feature_function,A",po::value >()->composing(), "Additional FSA feature function(s) (-L for list)") - ("apply_fsa_by",po::value()->default_value("BU_CUBE"), "Method for applying fsa_feature_functions - BU_FULL BU_CUBE EARLEY") //+ApplyFsaBy::all_names() -#endif ("add_pass_through_rules,P","Add rules to translate OOV words as themselves") ("k_best,k",po::value(),"Extract the k best derivations") ("unique_k_best,r", "Unique k-best translation list") - ("cubepruning_pop_limit,K",po::value()->default_value(200), "Max number of pops from the candidate heap at each node") ("aligner,a", "Run as a word/phrase aligner (src & ref required)") ("aligner_use_viterbi", "If run in alignment mode, compute the Viterbi (rather than MAP) alignment") ("goal",po::value()->default_value("S"),"Goal symbol (SCFG & FST)") @@ -446,10 +415,6 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream ("remove_intersected_rule_annotations", "After forced decoding is completed, remove nonterminal annotations (i.e., the source side spans)"); // ob.AddOptions(&opts); -#ifdef FSA_RESCORING - po::options_description cfgo(cfg_options.description()); - cfg_options.AddOptions(&cfgo); -#endif po::options_description clo("Command line options"); clo.add_options() ("config,c", po::value >(&cfg_files), "Configuration file(s) - latest has priority") @@ -459,15 +424,10 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream ; po::options_description dconfig_options, dcmdline_options; -#ifdef FSA_RESCORING - dconfig_options.add(opts).add(cfgo); -#else dconfig_options.add(opts); -#endif dcmdline_options.add(dconfig_options).add(clo); if (argc) { - argv_minus_to_underscore(argc,argv); po::store(parse_command_line(argc, argv, dcmdline_options), conf); if (conf.count("compgen")) { print_options(cout,dcmdline_options); @@ -511,10 +471,6 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream if (conf.count("list_feature_functions")) { cerr << "Available feature functions (specify with -F; describe with -u FeatureName):\n"; ff_registry.DisplayList(); //TODO -#ifdef FSA_RESCORING - cerr << "Available FSA feature functions (specify with --fsa_feature_function):\n"; - fsa_ff_registry.DisplayList(); // TODO -#endif cerr << endl; exit(1); } @@ -574,9 +530,6 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream if (conf.count("weights")) Weights::InitFromFile(str("weights",conf), init_weights.get()); - // cube pruning pop-limit: we may want to configure this on a per-pass basis - pop_limit = conf["cubepruning_pop_limit"].as(); - if (conf.count("extract_rules")) { if (!DirectoryExists(conf["extract_rules"].as())) MkDirP(conf["extract_rules"].as()); @@ -620,6 +573,9 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream if (conf.count(dp)) { rp.density_prune = conf[dp].as(); } int palg = (has_stateful ? 1 : 0); // if there are no stateful featueres, default to FULL string isn = "intersection_strategy" + StringSuffixForRescoringPass(pass); + string spl = "cubepruning_pop_limit" + StringSuffixForRescoringPass(pass); + unsigned pop_limit = 200; + if (conf.count(spl)) { pop_limit = conf[spl].as(); } if (LowercaseString(str(isn.c_str(),conf)) == "full") { palg = 0; } @@ -686,21 +642,6 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream else assert(!"error"); -#ifdef FSA_RESCORING - store_conf(conf,"fsa_feature_function",&fsa_names); - for (int i=0;i1) { - //FIXME: support N fsa ffs. - cerr<<"Only the first fsa FF will be used (FIXME).\n"; - fsa_ffs.resize(1); - } - if (!fsa_ffs.empty()) { - cerr<<"FSA: "; - show_all_features(fsa_ffs,*init_weights,cerr,cerr,true,true); - } -#endif - if (late_freeze) { cerr << "Late freezing feature set (use --no_freeze_feature_set to prevent)." << endl; FD::Freeze(); // this means we can't see the feature names of not-weighted features @@ -720,10 +661,6 @@ DecoderImpl::DecoderImpl(po::variables_map& conf, int argc, char** argv, istream oracle.show_derivation=conf.count("show_derivations"); remove_intersected_rule_annotations = conf.count("remove_intersected_rule_annotations"); -#ifdef FSA_RESCORING - cfg_options.Validate(); -#endif - if (conf.count("extract_rules")) { stringstream ss; ss << sent_id; @@ -840,7 +777,7 @@ bool DecoderImpl::Decode(const string& input, DecoderObserver* o) { HypergraphIO::WriteTarget(conf["show_target_graph"].as(), sent_id, forest); } if (conf.count("incremental_search")) { - incremental->Search(pop_limit, forest); + incremental->Search(conf["cubepruning_pop_limit"].as(), forest); } if (conf.count("show_target_graph") || conf.count("incremental_search")) { o->NotifyDecodingComplete(smeta); @@ -851,9 +788,6 @@ bool DecoderImpl::Decode(const string& input, DecoderObserver* o) { const RescoringPass& rp = rescoring_passes[pass]; const vector& cur_weights = *rp.weight_vector; if (!SILENT) cerr << endl << " RESCORING PASS #" << (pass+1) << " " << rp << endl; -#ifdef FSA_RESCORING - cfg_options.maybe_output_source(forest); -#endif string passtr = "Pass1"; passtr[4] += pass; forest.Reweight(cur_weights); @@ -949,25 +883,6 @@ bool DecoderImpl::Decode(const string& input, DecoderObserver* o) { string fullbp = "beam_prune" + StringSuffixForRescoringPass(pass); string fulldp = "density_prune" + StringSuffixForRescoringPass(pass); maybe_prune(forest,conf,fullbp.c_str(),fulldp.c_str(),passtr,srclen); - -#ifdef FSA_RESCORING - HgCFG hgcfg(forest); - cfg_options.prepare(hgcfg); - - if (!fsa_ffs.empty()) { - Timer t("Target FSA rescoring:"); - if (!has_late_models) - forest.Reweight(pass0_weights); - Hypergraph fsa_forest; - assert(fsa_ffs.size()==1); - ApplyFsaBy cfg(str("apply_fsa_by",conf),pop_limit); - if (!SILENT) cerr << "FSA rescoring with "<describe()<& last_weights = (rescoring_passes.empty() ? *init_weights : *rescoring_passes.back().weight_vector); diff --git a/decoder/decoder.h b/decoder/decoder.h index 79c7a602..8039a42b 100644 --- a/decoder/decoder.h +++ b/decoder/decoder.h @@ -25,9 +25,10 @@ private: class SentenceMetadata; class Hypergraph; -struct DecoderImpl; +class DecoderImpl; -struct DecoderObserver { +class DecoderObserver { + public: virtual ~DecoderObserver(); virtual void NotifyDecodingStart(const SentenceMetadata& smeta); virtual void NotifySourceParseFailure(const SentenceMetadata& smeta); @@ -37,9 +38,10 @@ struct DecoderObserver { virtual void NotifyDecodingComplete(const SentenceMetadata& smeta); }; -struct Grammar; // TODO once the decoder interface is cleaned up, - // this should be somewhere else -struct Decoder { +class Grammar; // TODO once the decoder interface is cleaned up, + // this should be somewhere else +class Decoder { + public: Decoder(int argc, char** argv); Decoder(std::istream* config_file); bool Decode(const std::string& input, DecoderObserver* observer = NULL); @@ -49,6 +51,7 @@ struct Decoder { std::vector& CurrentWeightVector(); const std::vector& CurrentWeightVector() const; + // this sets the current sentence ID void SetId(int id); ~Decoder(); const boost::program_options::variables_map& GetConf() const { return conf; } diff --git a/decoder/ff_factory.cc b/decoder/ff_factory.cc index 25d37648..f45d8695 100644 --- a/decoder/ff_factory.cc +++ b/decoder/ff_factory.cc @@ -7,26 +7,15 @@ using boost::shared_ptr; using namespace std; -UntypedFactory::~UntypedFactory() { } +// global ff registry +FFRegistry ff_registry; -namespace { -std::string const& debug_pre="debug"; -} +UntypedFactory::~UntypedFactory() { } void UntypedFactoryRegistry::clear() { reg_.clear(); } -bool UntypedFactoryRegistry::parse_debug(std::string & p) { - int pl=debug_pre.size(); - bool space=false; - bool debug=match_begin(p,debug_pre)&& - (p.size()==pl || (space=(p[pl]==' '))); - if (debug) - p.erase(0,debug_pre.size()+space); - return debug; -} - bool UntypedFactoryRegistry::have(std::string const& ffname) { return reg_.find(ffname)!=reg_.end(); } @@ -54,34 +43,18 @@ void UntypedFactoryRegistry::Register(const string& ffname, UntypedFactory* fact } -void UntypedFactoryRegistry::Register(UntypedFactory* factory) -{ +void UntypedFactoryRegistry::Register(UntypedFactory* factory) { Register(factory->usage(false,false),factory); } -/*FIXME: I want these to go in ff_factory.cc, but extern etc. isn't workign right: - ../decoder/libcdec.a(ff_factory.o): In function `~UntypedFactory': -/nfs/topaz/graehl/ws10smt/decoder/ff_factory.cc:9: multiple definition of `global_ff_registry' -mr_vest_generate_mapper_input.o:/nfs/topaz/graehl/ws10smt/vest/mr_vest_generate_mapper_input.cc:307: first defined here -*/ -FsaFFRegistry fsa_ff_registry; -FFRegistry ff_registry; - -/* -#include "null_deleter.h" -boost::shared_ptr global_fsa_ff_registry(&fsa_ff_registry,null_deleter()); -boost::shared_ptr global_ff_registry(&ff_registry,null_deleter()); -*/ - -void ff_usage(std::string const& n,std::ostream &out) -{ +void ff_usage(std::string const& n,std::ostream &out) { bool have=ff_registry.have(n); if (have) - cout<<"FF "< -struct FactoryBase : public UntypedFactory { +class FactoryBase : public UntypedFactory { + public: typedef FF F; typedef boost::shared_ptr FP; @@ -40,7 +40,8 @@ struct FactoryBase : public UntypedFactory { /* see cdec_ff.cc for example usage: this create concrete factories to be registered */ template -struct FFFactory : public FactoryBase { +class FFFactory : public FactoryBase { + public: FP Create(std::string param) const { FF *ret=new FF(param); return FP(ret); @@ -51,18 +52,6 @@ struct FFFactory : public FactoryBase { }; -// same as above, but we didn't want to require a typedef e.g. Parent in FF class, and template typedef isn't available -template -struct FsaFactory : public FactoryBase { - FP Create(std::string param) const { - FF *ret=new FF(param); - return FP(ret); - } - virtual std::string usage(bool params,bool verbose) const { - return FF::usage(params,verbose); - } -}; - struct UntypedFactoryRegistry { std::string usage(std::string const& ffname,bool params=true,bool verbose=true) const; bool have(std::string const& ffname); @@ -70,7 +59,6 @@ struct UntypedFactoryRegistry { void Register(const std::string& ffname, UntypedFactory* factory); void Register(UntypedFactory* factory); void clear(); - static bool parse_debug(std::string & param_in_out); // returns true iff param starts w/ debug (and remove that prefix from param) protected: typedef boost::shared_ptr FactoryP; typedef std::map Factmap; @@ -92,26 +80,16 @@ struct FactoryRegistry : public UntypedFactoryRegistry { Factmap::const_iterator it = reg_.find(ffname); if (it == reg_.end()) throw std::runtime_error("I don't know how to create feature "+ffname); - bool debug=parse_debug(param); - if (debug) - cerr<<"debug enabled for "<(*it->second).Create(param); return res; } }; typedef FactoryRegistry FFRegistry; -typedef FactoryRegistry FsaFFRegistry; -extern FsaFFRegistry fsa_ff_registry; -inline FsaFFRegistry & global_fsa_ff_registry() { return fsa_ff_registry; } extern FFRegistry ff_registry; -inline FFRegistry & global_ff_registry() { return ff_registry; } +inline FFRegistry& global_ff_registry() { return ff_registry; } -void ff_usage(std::string const& name,std::ostream &out=std::cout); +void ff_usage(std::string const& name,std::ostream& out=std::cerr); -/* -extern boost::shared_ptr global_fsa_ff_registry; -extern boost::shared_ptr global_ff_registry; -*/ #endif diff --git a/decoder/ff_klm.cc b/decoder/ff_klm.cc index c8ca917a..339a10c3 100644 --- a/decoder/ff_klm.cc +++ b/decoder/ff_klm.cc @@ -187,6 +187,7 @@ class KLanguageModelImpl { // this assumes no target words on final unary -> goal rule. is that ok? // for (n-1 left words) and (n-1 right words) double FinalTraversalCost(const void* state_void, double* oovs) { + *oovs = 0; const BoundaryAnnotatedState &annotated = *static_cast(state_void); if (add_sos_eos_) { // rules do not produce , so do it here assert(!annotated.seen_bos); @@ -344,7 +345,7 @@ void KLanguageModel::TraversalFeaturesImpl(const SentenceMetadata& /* sme const Hypergraph::Edge& edge, const vector& ant_states, SparseVector* features, - SparseVector* estimated_features, + SparseVector* /*estimated_features*/, void* state) const { double est = 0; double oovs = 0; diff --git a/decoder/hg_cfg.h b/decoder/hg_cfg.h deleted file mode 100644 index b90aca47..00000000 --- a/decoder/hg_cfg.h +++ /dev/null @@ -1,54 +0,0 @@ -#ifndef HG_CFG_H -#define HG_CFG_H - -#include "cfg.h" - -class Hypergraph; - -// in case you might want the CFG whether or not you apply FSA models: -struct HgCFG { - void set_defaults() { - have_cfg=binarized=have_features=uniqed=false; - want_features=true; - } - HgCFG(Hypergraph const& ih) : ih(ih) { - set_defaults(); - } - Hypergraph const& ih; - CFG cfg; - bool have_cfg; - bool have_features; - bool want_features; - void InitCFG(CFG &to) { - to.Init(ih,true,want_features,true); - } - bool binarized; - bool uniqed; - CFG &GetCFG() - { - if (!have_cfg) { - have_cfg=true; - InitCFG(cfg); - } - return cfg; - } - void GiveCFG(CFG &to) { - if (!have_cfg) - InitCFG(to); - else { - cfg.VisitRuleIds(*this); - have_cfg=false; - to.Clear(); - swap(to,cfg); - } - } - void operator()(int ri) const { - } - CFG const& GetCFG() const { - assert(have_cfg); - return cfg; - } -}; - - -#endif diff --git a/decoder/sentence_metadata.h b/decoder/sentence_metadata.h index 52586331..f2a779f4 100644 --- a/decoder/sentence_metadata.h +++ b/decoder/sentence_metadata.h @@ -9,7 +9,8 @@ struct DocScorer; // deprecated, will be removed struct Score; // deprecated, will be removed -struct SentenceMetadata { +class SentenceMetadata { + public: friend class DecoderImpl; SentenceMetadata(int id, const Lattice& ref) : sent_id_(id), diff --git a/decoder/viterbi.cc b/decoder/viterbi.cc index 9e381ac6..9204ad04 100644 --- a/decoder/viterbi.cc +++ b/decoder/viterbi.cc @@ -1,6 +1,8 @@ -#include "fast_lexical_cast.hpp" #include "viterbi.h" +#include +#include +#include "fast_lexical_cast.hpp" #include #include #include "hg.h" @@ -110,30 +112,7 @@ string JoshuaVisualizationString(const Hypergraph& hg) { return TD::GetString(tmp); } - -//TODO: move to appropriate header if useful elsewhere -/* - The simple solution like abs(f1-f2) <= e does not work for very small or very big values. This floating-point comparison algorithm is based on the more confident solution presented by Knuth in [1]. For a given floating point values u and v and a tolerance e: - -| u - v | <= e * |u| and | u - v | <= e * |v| -defines a "very close with tolerance e" relationship between u and v - (1) - -| u - v | <= e * |u| or | u - v | <= e * |v| -defines a "close enough with tolerance e" relationship between u and v - (2) - -Both relationships are commutative but are not transitive. The relationship defined by inequations (1) is stronger that the relationship defined by inequations (2) (i.e. (1) => (2) ). Because of the multiplication in the right side of inequations, that could cause an unwanted underflow condition, the implementation is using modified version of the inequations (1) and (2) where all underflow, overflow conditions could be guarded safely: - -| u - v | / |u| <= e and | u - v | / |v| <= e -| u - v | / |u| <= e or | u - v | / |v| <= e - (1`) -(2`) -*/ -#include -#include -inline bool close_enough(double a,double b,double epsilon) -{ +inline bool close_enough(double a,double b,double epsilon) { using std::fabs; double diff=fabs(a-b); return diff<=epsilon*fabs(a) || diff<=epsilon*fabs(b); diff --git a/mteval/external_scorer.h b/mteval/external_scorer.h index a28fb920..85535655 100644 --- a/mteval/external_scorer.h +++ b/mteval/external_scorer.h @@ -24,7 +24,8 @@ class ScoreServer { int c2p[2]; }; -struct ScoreServerManager { +class ScoreServerManager { + public: static ScoreServer* Instance(const std::string& score_type); private: static std::map > servers_; diff --git a/mteval/ns_ext.cc b/mteval/ns_ext.cc index 956708af..1e7e2bc1 100644 --- a/mteval/ns_ext.cc +++ b/mteval/ns_ext.cc @@ -118,7 +118,7 @@ void ExternalMetric::ComputeSufficientStatistics(const std::vector& hyp, } float ExternalMetric::ComputeScore(const SufficientStats& stats) const { - eval_server->ComputeScore(stats.fields); + return eval_server->ComputeScore(stats.fields); } ExternalMetric::ExternalMetric(const string& metric_name, const std::string& command) : diff --git a/mteval/ns_ter.cc b/mteval/ns_ter.cc index 680fb7b4..00b6eb01 100644 --- a/mteval/ns_ter.cc +++ b/mteval/ns_ter.cc @@ -298,7 +298,7 @@ class TERScorerImpl { } bool CalculateBestShift(const vector& cur, - const vector& hyp, + const vector& /*hyp*/, float curerr, const vector& path, vector* new_hyp, diff --git a/mteval/scorer.h b/mteval/scorer.h index bb1e89ae..8d986612 100644 --- a/mteval/scorer.h +++ b/mteval/scorer.h @@ -103,7 +103,7 @@ class DocScorer { virtual int size() const { return scorers_.size(); } virtual ScorerP operator[](size_t i) const { return scorers_[i]; } - virtual void update(const std::string& ref) {} + virtual void update(const std::string& /*ref*/) {} private: std::vector scorers_; }; @@ -124,7 +124,7 @@ class DocStreamScorer : public DocScorer { { Init(type,ref_files,src_file,verbose); } - ScorerP operator[](size_t i) const { return scorer; } + ScorerP operator[](size_t /*i*/) const { return scorer; } int size() const { return 1; } void update(const std::string& ref); private: diff --git a/tests/run-system-tests.pl b/tests/run-system-tests.pl index 8555ef78..324763ae 100755 --- a/tests/run-system-tests.pl +++ b/tests/run-system-tests.pl @@ -11,6 +11,7 @@ my $TEMP_DIR = tempdir( CLEANUP => 1 ); my $DECODER = "$script_dir/../decoder/cdec"; my $FILTER = "$script_dir/tools/filter-stderr.pl"; my $COMPARE_STATS = "$script_dir/tools/compare-statistics.pl"; +my $XDIFF = "$script_dir/tools/flex-diff.pl"; die "Can't find $DECODER" unless -f $DECODER; die "Can't execute $DECODER" unless -x $DECODER; @@ -18,6 +19,7 @@ die "Can't find $FILTER" unless -f $FILTER; die "Can't execute $FILTER" unless -x $FILTER; die "Can't find $COMPARE_STATS" unless -f $COMPARE_STATS; die "Can't execute $COMPARE_STATS" unless -x $COMPARE_STATS; +die "Can't execute $XDIFF" unless -x $XDIFF; my $TEST_DIR = "$script_dir/system_tests"; opendir DIR, $TEST_DIR or die "Can't open $TEST_DIR: $!"; @@ -62,7 +64,7 @@ for my $test (@tests) { } else { die unless -f "$TEMP_DIR/stdout"; my $failed = 0; - run3 "diff gold.stdout $TEMP_DIR/stdout"; + run3 "$XDIFF gold.stdout $TEMP_DIR/stdout"; if ($? != 0) { print STDERR " FAILED differences in output!\n"; $failed = 1; diff --git a/tests/tools/flex-diff.pl b/tests/tools/flex-diff.pl new file mode 100755 index 00000000..245e9a1b --- /dev/null +++ b/tests/tools/flex-diff.pl @@ -0,0 +1,43 @@ +#!/usr/bin/perl -w +use strict; +use IPC::Run3; + +# this file abstracts away from differences due to different hash +# functions that lead to different orders of features, n-best entries, +# etc. + +die "Usage: $0 file1.txt file2.txt\n" unless scalar @ARGV == 2; +my $tmpa = "tmp.$$.a"; +my $tmpb = "tmp.$$.b"; +create_sorted($ARGV[0], $tmpa); +create_sorted($ARGV[1], $tmpb); + +my $failed = 0; +run3 "diff $tmpa $tmpb"; +if ($? != 0) { + run3 "diff $ARGV[0] $ARGV[1]"; + $failed = 1; +} + +unlink $tmpa; +unlink $tmpb; + +exit $failed; + +sub create_sorted { + my ($in, $out) = @_; + open A, "sort $in|" or die "Can't read $in: $!"; + open AA, ">$out" or die "Can't write $out: $!"; + while() { + chomp; + s/^\s*//; + s/\s*$//; + my @cs = split //; + @cs = sort @cs; + my $o = join('', @cs); + print AA "$o\n"; + } + close AA; + close A; +} + diff --git a/training/crf/Makefile.am b/training/crf/Makefile.am index 4a8c30fd..cd82161f 100644 --- a/training/crf/Makefile.am +++ b/training/crf/Makefile.am @@ -1,5 +1,6 @@ bin_PROGRAMS = \ mpi_batch_optimize \ + mpi_adagrad_optimize \ mpi_compute_cllh \ mpi_extract_features \ mpi_extract_reachable \ @@ -10,6 +11,9 @@ bin_PROGRAMS = \ mpi_baum_welch_SOURCES = mpi_baum_welch.cc mpi_baum_welch_LDADD = ../../decoder/libcdec.a ../../klm/search/libksearch.a ../../mteval/libmteval.a ../../utils/libutils.a ../../klm/lm/libklm.a ../../klm/util/libklm_util.a ../../klm/util/double-conversion/libklm_util_double.a -lz +mpi_adagrad_optimize_SOURCES = mpi_adagrad_optimize.cc cllh_observer.cc cllh_observer.h +mpi_adagrad_optimize_LDADD = ../../training/utils/libtraining_utils.a ../../decoder/libcdec.a ../../klm/search/libksearch.a ../../mteval/libmteval.a ../../utils/libutils.a ../../klm/lm/libklm.a ../../klm/util/libklm_util.a ../../klm/util/double-conversion/libklm_util_double.a -lz + mpi_online_optimize_SOURCES = mpi_online_optimize.cc mpi_online_optimize_LDADD = ../../training/utils/libtraining_utils.a ../../decoder/libcdec.a ../../klm/search/libksearch.a ../../mteval/libmteval.a ../../utils/libutils.a ../../klm/lm/libklm.a ../../klm/util/libklm_util.a ../../klm/util/double-conversion/libklm_util_double.a -lz diff --git a/training/crf/mpi_adagrad_optimize.cc b/training/crf/mpi_adagrad_optimize.cc new file mode 100644 index 00000000..05e7c35f --- /dev/null +++ b/training/crf/mpi_adagrad_optimize.cc @@ -0,0 +1,355 @@ +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +#include "config.h" +#include "stringlib.h" +#include "verbose.h" +#include "cllh_observer.h" +#include "hg.h" +#include "prob.h" +#include "inside_outside.h" +#include "ff_register.h" +#include "decoder.h" +#include "filelib.h" +#include "online_optimizer.h" +#include "fdict.h" +#include "weights.h" +#include "sparse_vector.h" +#include "sampler.h" + +#ifdef HAVE_MPI +#include +#include +namespace mpi = boost::mpi; +#endif + +using namespace std; +namespace po = boost::program_options; + +bool InitCommandLine(int argc, char** argv, po::variables_map* conf) { + po::options_description opts("Configuration options"); + opts.add_options() + ("weights,w",po::value(), "Initial feature weights") + ("training_data,d",po::value(), "Training data corpus") + ("test_data,t",po::value(), "(optional) Test data") + ("decoder_config,c",po::value(), "Decoder configuration file") + ("minibatch_size_per_proc,s", po::value()->default_value(8), + "Number of training instances evaluated per processor in each minibatch") + ("max_passes", po::value()->default_value(20.0), "Maximum number of passes through the data") + ("max_walltime", po::value(), "Walltime to run (in minutes)") + ("write_every_n_minibatches", po::value()->default_value(100), "Write weights every N minibatches processed") + ("random_seed,S", po::value(), "Random seed") + ("regularization,r", po::value()->default_value("none"), + "Regularization 'none', 'l1', or 'l2'") + ("regularization_strength,C", po::value(), "Regularization strength") + ("eta,e", po::value()->default_value(1.0), "Initial learning rate (eta)"); + po::options_description clo("Command line options"); + clo.add_options() + ("config", po::value(), "Configuration file") + ("help,h", "Print this help message and exit"); + po::options_description dconfig_options, dcmdline_options; + dconfig_options.add(opts); + dcmdline_options.add(opts).add(clo); + + po::store(parse_command_line(argc, argv, dcmdline_options), *conf); + if (conf->count("config")) { + ifstream config((*conf)["config"].as().c_str()); + po::store(po::parse_config_file(config, dconfig_options), *conf); + } + po::notify(*conf); + + if (conf->count("help") || !conf->count("training_data") || !conf->count("decoder_config")) { + cerr << dcmdline_options << endl; + return false; + } + return true; +} + +void ReadTrainingCorpus(const string& fname, int rank, int size, vector* c, vector* order) { + ReadFile rf(fname); + istream& in = *rf.stream(); + string line; + int id = 0; + while(getline(in, line)) { + if (id % size == rank) { + c->push_back(line); + order->push_back(id); + } + ++id; + } +} + +static const double kMINUS_EPSILON = -1e-6; + +struct TrainingObserver : public DecoderObserver { + void Reset() { + acc_grad.clear(); + acc_obj = 0; + total_complete = 0; + } + + virtual void NotifyDecodingStart(const SentenceMetadata&) { + cur_model_exp.clear(); + cur_obj = 0; + state = 1; + } + + // compute model expectations, denominator of objective + virtual void NotifyTranslationForest(const SentenceMetadata&, Hypergraph* hg) { + assert(state == 1); + state = 2; + const prob_t z = InsideOutside, + EdgeFeaturesAndProbWeightFunction>(*hg, &cur_model_exp); + cur_obj = log(z); + cur_model_exp /= z; + } + + // compute "empirical" expectations, numerator of objective + virtual void NotifyAlignmentForest(const SentenceMetadata&, Hypergraph* hg) { + assert(state == 2); + state = 3; + SparseVector ref_exp; + const prob_t ref_z = InsideOutside, + EdgeFeaturesAndProbWeightFunction>(*hg, &ref_exp); + ref_exp /= ref_z; + + double log_ref_z; +#if 0 + if (crf_uniform_empirical) { + log_ref_z = ref_exp.dot(feature_weights); + } else { + log_ref_z = log(ref_z); + } +#else + log_ref_z = log(ref_z); +#endif + + // rounding errors means that <0 is too strict + if ((cur_obj - log_ref_z) < kMINUS_EPSILON) { + cerr << "DIFF. ERR! log_model_z < log_ref_z: " << cur_obj << " " << log_ref_z << endl; + exit(1); + } + assert(!std::isnan(log_ref_z)); + ref_exp -= cur_model_exp; + acc_grad += ref_exp; + acc_obj += (cur_obj - log_ref_z); + } + + virtual void NotifyDecodingComplete(const SentenceMetadata&) { + if (state == 3) { + ++total_complete; + } else { + } + } + + void GetGradient(SparseVector* g) const { + g->clear(); +#if HAVE_CXX11 + for (auto& gi : acc_grad) { +#else + for (FastSparseVector::const_iterator it = acc_grad.begin(); it != acc_grad.end(); ++it) { + pair& gi = *it; +#endif + g->set_value(gi.first, -gi.second.as_float()); + } + } + + int total_complete; + SparseVector cur_model_exp; + SparseVector acc_grad; + double acc_obj; + double cur_obj; + int state; +}; + +#ifdef HAVE_MPI +namespace boost { namespace mpi { + template<> + struct is_commutative >, SparseVector > + : mpl::true_ { }; +} } // end namespace boost::mpi +#endif + +class AdaGradOptimizer { + public: + explicit AdaGradOptimizer(double e) : + eta(e), + G() {} + void update(const SparseVector& g, vector* x) { + if (x->size() > G.size()) G.resize(x->size(), 0.0); +#if HAVE_CXX11 + for (auto& gi : g) { +#else + for (SparseVector::const_iterator it = g.begin(); it != g.end(); ++it) { + const pair& gi = *it; +#endif + if (gi.second) { + G[gi.first] += gi.second * gi.second; + (*x)[gi.first] -= eta / sqrt(G[gi.first]) * gi.second; + } + } + } + const double eta; + vector G; +}; + +unsigned non_zeros(const vector& x) { + unsigned nz = 0; + for (unsigned i = 0; i < x.size(); ++i) + if (x[i]) ++nz; + return nz; +} + +int main(int argc, char** argv) { +#ifdef HAVE_MPI + mpi::environment env(argc, argv); + mpi::communicator world; + const int size = world.size(); + const int rank = world.rank(); +#else + const int size = 1; + const int rank = 0; +#endif + if (size > 1) SetSilent(true); // turn off verbose decoder output + register_feature_functions(); + + po::variables_map conf; + if (!InitCommandLine(argc, argv, &conf)) + return 1; + + ReadFile ini_rf(conf["decoder_config"].as()); + Decoder decoder(ini_rf.stream()); + + // load initial weights + vector init_weights; + if (conf.count("input_weights")) + Weights::InitFromFile(conf["input_weights"].as(), &init_weights); + + vector corpus, test_corpus; + vector ids; + ReadTrainingCorpus(conf["training_data"].as(), rank, size, &corpus, &ids); + assert(corpus.size() > 0); + if (conf.count("test_data")) + ReadTrainingCorpus(conf["test_data"].as(), rank, size, &corpus, &ids); + + const unsigned size_per_proc = conf["minibatch_size_per_proc"].as(); + if (size_per_proc > corpus.size()) { + cerr << "Minibatch size must be smaller than corpus size!\n"; + return 1; + } + const double minibatch_size = size_per_proc * size; + + size_t total_corpus_size = 0; +#ifdef HAVE_MPI + reduce(world, corpus.size(), total_corpus_size, std::plus(), 0); +#else + total_corpus_size = corpus.size(); +#endif + + if (rank == 0) + cerr << "Total corpus size: " << total_corpus_size << endl; + + boost::shared_ptr rng; + if (conf.count("random_seed")) + rng.reset(new MT19937(conf["random_seed"].as())); + else + rng.reset(new MT19937); + + double passes_per_minibatch = static_cast(size_per_proc) / total_corpus_size; + + int write_weights_every_ith = conf["write_every_n_minibatches"].as(); + + unsigned max_iteration = conf["max_passes"].as() / passes_per_minibatch; + ++max_iteration; + if (rank == 0) + cerr << "Max passes through data = " << conf["max_passes"].as() << endl + << " --> max minibatches = " << max_iteration << endl; + unsigned timeout = 0; + if (conf.count("max_walltime")) + timeout = 60 * conf["max_walltime"].as(); + vector& lambdas = decoder.CurrentWeightVector(); + if (init_weights.size()) { + lambdas.swap(init_weights); + init_weights.clear(); + } + + AdaGradOptimizer adagrad(conf["eta"].as()); + int iter = -1; + bool converged = false; + + TrainingObserver observer; + ConditionalLikelihoodObserver cllh_observer; + + const time_t start_time = time(NULL); + while (!converged) { +#ifdef HAVE_MPI + mpi::timer timer; +#endif + ++iter; + observer.Reset(); + if (rank == 0) { + converged = (iter == max_iteration); + string fname = "weights.cur.gz"; + if (iter % write_weights_every_ith == 0) { + ostringstream o; o << "weights." << iter << ".gz"; + fname = o.str(); + } + const time_t cur_time = time(NULL); + if (timeout && ((cur_time - start_time) > timeout)) { + converged = true; + fname = "weights.final.gz"; + } + ostringstream vv; + double minutes = (cur_time - start_time) / 60.0; + vv << "total walltime=" << minutes << "min iter=" << iter << " minibatch=" << size_per_proc << " sentences/proc x " << size << " procs. num_feats=" << non_zeros(lambdas) << '/' << FD::NumFeats() << " passes_thru_data=" << (iter * size_per_proc / static_cast(corpus.size())); + const string svv = vv.str(); + cerr << svv << endl; + Weights::WriteToFile(fname, lambdas, true, &svv); + } + + for (int i = 0; i < size_per_proc; ++i) { + int ei = corpus.size() * rng->next(); + int id = ids[ei]; + decoder.SetId(id); + decoder.Decode(corpus[ei], &observer); + } + SparseVector local_grad, g; + observer.GetGradient(&local_grad); +#ifdef HAVE_MPI + reduce(world, local_grad, g, std::plus >(), 0); +#else + g.swap(local_grad); +#endif + local_grad.clear(); + if (rank == 0) { + g /= minibatch_size; + lambdas.resize(FD::NumFeats(), 0.0); // might have seen new features + adagrad.update(g, &lambdas); + Weights::SanityCheck(lambdas); + Weights::ShowLargestFeatures(lambdas); + } +#ifdef HAVE_MPI + broadcast(world, lambdas, 0); + broadcast(world, converged, 0); + world.barrier(); + if (rank == 0) { cerr << " ELAPSED TIME THIS ITERATION=" << timer.elapsed() << endl; } +#endif + } + cerr << "CONVERGED = " << converged << endl; + cerr << "EXITING...\n"; + return 0; +} + diff --git a/training/crf/mpi_batch_optimize.cc b/training/crf/mpi_batch_optimize.cc index 2eff07e4..da1845b1 100644 --- a/training/crf/mpi_batch_optimize.cc +++ b/training/crf/mpi_batch_optimize.cc @@ -97,14 +97,14 @@ struct TrainingObserver : public DecoderObserver { (*g)[it->first] = it->second.as_float(); } - virtual void NotifyDecodingStart(const SentenceMetadata& smeta) { + virtual void NotifyDecodingStart(const SentenceMetadata&) { cur_model_exp.clear(); cur_obj = 0; state = 1; } // compute model expectations, denominator of objective - virtual void NotifyTranslationForest(const SentenceMetadata& smeta, Hypergraph* hg) { + virtual void NotifyTranslationForest(const SentenceMetadata&, Hypergraph* hg) { assert(state == 1); state = 2; const prob_t z = InsideOutside #include #include -#include "swap_pod.h" #include //sizeof(T)/sizeof(T*)>1?sizeof(T)/sizeof(T*):1 @@ -278,8 +277,15 @@ public: return !(a==b); } - void swap(Self& o) { - swap_pod(*this,o); + inline void swap(Self& o) { + const unsigned s=sizeof(SmallVector); + char tmp[s]; + void *pt=static_cast(tmp); + void *pa=static_cast(this); + void *pb=static_cast(&o); + std::memcpy(pt,pa,s); + std::memcpy(pa,pb,s); + std::memcpy(pb,pt,s); } inline std::size_t hash_impl() const { diff --git a/utils/small_vector_test.cc b/utils/small_vector_test.cc index cded4619..a4eb89ae 100644 --- a/utils/small_vector_test.cc +++ b/utils/small_vector_test.cc @@ -82,6 +82,18 @@ BOOST_AUTO_TEST_CASE(LargerThan2) { BOOST_CHECK(cc.size() == 0); } +BOOST_AUTO_TEST_CASE(SwapSV) { + SmallVectorInt v; + SmallVectorInt v2(2, 10); + SmallVectorInt v3(2, 10); + BOOST_CHECK(v2 == v3); + BOOST_CHECK(v != v3); + v.swap(v2); + BOOST_CHECK(v == v3); + SmallVectorInt v4; + BOOST_CHECK(v4 == v2); +} + BOOST_AUTO_TEST_CASE(Small) { SmallVectorInt v; SmallVectorInt v1(1,0); diff --git a/utils/sv_test.cc b/utils/sv_test.cc new file mode 100644 index 00000000..c7ac9e54 --- /dev/null +++ b/utils/sv_test.cc @@ -0,0 +1,24 @@ +#define BOOST_TEST_MODULE WeightsTest +#include +#include +#include "sparse_vector.h" + +using namespace std; + +BOOST_AUTO_TEST_CASE(Equality) { + SparseVector x; + SparseVector y; + x.set_value(1,-1); + y.set_value(1,-1); + BOOST_CHECK(x == y); +} + +BOOST_AUTO_TEST_CASE(Division) { + SparseVector x; + SparseVector y; + x.set_value(1,1); + y.set_value(1,-1); + BOOST_CHECK(!(x == y)); + x /= -1; + BOOST_CHECK(x == y); +} diff --git a/utils/swap_pod.h b/utils/swap_pod.h deleted file mode 100644 index bb9a830d..00000000 --- a/utils/swap_pod.h +++ /dev/null @@ -1,23 +0,0 @@ -#ifndef SWAP_POD_H -#define SWAP_POD_H - -//for swapping objects of the same concrete type where just swapping their bytes will work. will at least work on plain old data. - -#include // not used, but people who use this will want to bring std::swap in anyway -#include - -template -inline void swap_pod(T &a,T &b) { - using namespace std; - const unsigned s=sizeof(T); - char tmp[s]; - void *pt=(void*)tmp; - void *pa=(void*)&a; - void *pb=(void*)&b; - memcpy(pt,pa,s); - memcpy(pa,pb,s); - memcpy(pb,pt,s); -} - - -#endif diff --git a/utils/value_array.h b/utils/value_array.h index 12fc9d87..e59349b5 100644 --- a/utils/value_array.h +++ b/utils/value_array.h @@ -1,8 +1,6 @@ #ifndef VALUE_ARRAY_H #define VALUE_ARRAY_H -//TODO: option for non-constructed version (type_traits pod?), option for small array optimization (if sz < N, store inline in union, see small_vector.h) - #define DBVALUEARRAY(x) x #include @@ -30,8 +28,7 @@ // valarray like in that size is fixed (so saves space compared to vector), but same interface as vector (less resize/push_back/insert, of course) template > -class ValueArray : A // private inheritance so stateless allocator adds no size. -{ +class ValueArray : A { // private inheritance so stateless allocator adds no size. typedef ValueArray Self; public: #if VALUE_ARRAY_OSTREAM @@ -323,14 +320,14 @@ private: //friend class boost::serialization::access; public: template - void save(Archive& ar, unsigned int version) const + void save(Archive& ar, unsigned int /*version*/) const { ar << sz; for (size_type i = 0; i != sz; ++i) ar << at(i); } template - void load(Archive& ar, unsigned int version) + void load(Archive& ar, unsigned int /*version*/) { size_type s; ar >> s; diff --git a/utils/weights.cc b/utils/weights.cc index effdfc5e..1f66c441 100644 --- a/utils/weights.cc +++ b/utils/weights.cc @@ -127,9 +127,11 @@ void Weights::InitSparseVector(const vector& dv, } void Weights::SanityCheck(const vector& w) { - for (unsigned i = 0; i < w.size(); ++i) { - assert(!std::isnan(w[i])); - assert(!std::isinf(w[i])); + for (unsigned i = 1; i < w.size(); ++i) { + if (std::isnan(w[i]) || std::isinf(w[i])) { + cerr << FD::Convert(i) << " has bad weight: " << w[i] << endl; + abort(); + } } } @@ -161,7 +163,7 @@ string Weights::GetString(const vector& w, bool hide_zero_value_features) { ostringstream os; os.precision(17); - int nf = FD::NumFeats(); + const unsigned nf = FD::NumFeats(); for (unsigned i = 1; i < nf; i++) { weight_t val = (i < w.size() ? w[i] : 0.0); if (hide_zero_value_features && val == 0.0) { -- cgit v1.2.3