int32_t LightDocSampler::OldProposalFreshSampleInfer(LDADocument *doc)
{
DocInit(doc);
int num_token = doc->size();
int32_t &cursor = doc->get_cursor();
int32_t token_sweeped = 0;
cursor = 0;
while (cursor < num_token)
{
++token_sweeped;
int32_t w = doc->Word(cursor);
int32_t s = doc->Topic(cursor); // old topic
int t = Sample2WordFirstInfer(doc, w, s, s); // new topic
if (s != t)
{
doc->SetTopic(cursor, t);
doc_topic_counter_.inc(s, -1);
doc_topic_counter_.inc(t, 1);
}
cursor++;
}
return token_sweeped;
}
void LightDocSampler::InferOneDoc(LDADocument* doc, ModelSlice& word_topic_table,
petuum::ClientSummaryRow& summary_row, AliasSlice& alias_table)
{
DocInit(doc);
int num_token = doc->size();
int32_t slice_id = word_topic_table.SliceId();
int32_t slice_last_word = word_topic_table.LastWord();
int32_t& cursor = doc->get_cursor();
if (slice_id == 0) cursor = 0;
for (; cursor != doc->size(); ++cursor) {
int32_t word = doc->Word(cursor);
if (word > slice_last_word)
break;
int32_t old_topic = doc->Topic(cursor);
int32_t new_topic = InferWordFirst(doc, word, old_topic, old_topic,
word_topic_table, summary_row, alias_table);
if (old_topic != new_topic) {
doc_topic_counter_.inc(old_topic, -1);
doc_topic_counter_.inc(new_topic, 1);
doc->SetTopic(cursor, new_topic);
}
}
}
int32_t LightDocSampler::SampleOneDoc(LDADocument *doc,
ModelSlice& word_topic_table,
petuum::ClientSummaryRow& summary_row,
AliasSlice& alias_table,
std::vector<std::unique_ptr<petuum::DeltaArray>>& word_topic_delta_vec,
petuum::SummaryDelta& summary_delta)
{
DocInit(doc);
int num_token = doc->size();
int32_t num_sampling = 0;
int32_t num_sampling_changed = 0;
int32_t slice_id = word_topic_table.SliceId();
int32_t slice_last_word = word_topic_table.LastWord();
int32_t& cursor = doc->get_cursor();
if (slice_id == 0) cursor = 0;
for (; cursor != doc->size(); ++cursor) {
int32_t word = doc->Word(cursor);
if (word > slice_last_word)
break;
++num_sampling;
++num_sampling_;
int32_t old_topic = doc->Topic(cursor);
int32_t new_topic = Sample2WordFirst(doc, word, old_topic, old_topic,
word_topic_table, summary_row, alias_table);
if (old_topic != new_topic) {
int32_t shard_id = word % word_topic_delta_vec.size();
word_topic_delta_vec[shard_id]->Update(word, old_topic, -1);
doc_topic_counter_.inc(old_topic, -1);
summary_delta.Update(old_topic, -1);
word_topic_delta_vec[shard_id]->Update(word, new_topic, 1);
doc_topic_counter_.inc(new_topic, 1);
summary_delta.Update(new_topic, 1);
doc->SetTopic(cursor, new_topic);
++num_sampling_changed_;
++num_sampling_changed;
}
}
return num_sampling;
}
int32_t LightDocSampler::OldProposalFreshSample(LDADocument *doc)
{
DocInit(doc);
int num_token = doc->size();
int32_t &cursor = doc->get_cursor();
int32_t token_sweeped = 0;
cursor = 0;
while (cursor < num_token)
{
++token_sweeped;
int32_t w = doc->Word(cursor);
int32_t s = doc->Topic(cursor); // old topic
int t = Sample2WordFirst(doc, w, s, s); // new topic
if (s != t)
{
word_topic_delta wtd;
int32_t shard_id = w % num_threads_;
wtd.word = w;
wtd.topic = s;
wtd.delta = -1;
word_topic_delta_[shard_id].push_back(wtd);
wtd.topic = t;
wtd.delta = +1;
word_topic_delta_[shard_id].push_back(wtd);
--delta_summary_row_[s];
++delta_summary_row_[t];
doc->SetTopic(cursor, t);
doc_topic_counter_.inc(s, -1);
doc_topic_counter_.inc(t, 1);
}
cursor++;
}
return token_sweeped;
}
