[doc] fix grouped sequence_fields default rule; drop epoch/wall-clock note from train.md#542
Merged
tiankongdeguiji merged 2 commits intoJun 11, 2026
Conversation
…e ops For multi-input feature ops (LookupFeature, ComboFeature, etc.) inside a grouped sequence feature, inputs with input_side == item default to sequence type, not input_side != item. This matches BaseFeature._is_sequence_input and the example on the next line (item:cate -> click_seq__cate, user:kv_cate_cnt -> kv_cate_cnt). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…stamp_interval Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
tiankongdeguiji
changed the title
chengaofei
approved these changes
Jun 11, 2026
WhiteSwan1
added a commit
to WhiteSwan1/TorchEasyRec
that referenced
this pull request
Jun 11, 2026
…t/sid_abstract upstream/master advanced to 3d4d5a8: alibaba#539 (SidRqkmeans) was merged as a squash and alibaba#542 (docs) landed. feat already carries alibaba#539's content (from the earlier pr-539 merge), so this is effectively the alibaba#542 doc update plus reconciling the SID proto/types surface. Conflicts (all add/add on files feat extended for RQ-VAE) resolved by taking feat's superset: - modules/sid/types.py: keep QuantizeForwardMode + ResidualQuantizerOutput (master's alibaba#539 has only QuantizeOutput, which is identical and auto-merged). - protos/models/sid_model.proto: keep SinkhornConfig/ClipConfig/SidRqvae on top of the shared FaissKmeansConfig + SidRqkmeans. - protos/model.proto: keep `SidRqvae sid_rqvae = 600;` (master still has it reserved as a comment).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Two doc fixes:
docs/source/feature/feature.md— fix an inverted condition in the grouped sequence feature docs: for multi-input feature ops (LookupFeature, ComboFeature, etc.) without explicitsequence_fields, inputs withinput_side == itemdefault to sequence type — the doc previously saidinput_side != item.docs/source/usage/train.md— remove the confusing parenthetical「边界按 Unix 时间(epoch/墙钟)对齐，而非训练 epoch」from thesave_checkpoints_timestamp_intervaldescription; it used "epoch" in two senses (Unix epoch vs training epoch) in one sentence. The remaining text already describes the trigger; precise boundary semantics stay documented incheckpoint_util.py.Why
For (1), the stated rule contradicted both the implementation and the doc's own example on the next line:
BaseFeature._is_sequence_inputreturnsside == "item"for multi-input feature classes (tzrec/features/feature.py), mirroring the pyfg C++ rule insequence_feature.cc.item:catemaps to columnclick_seq__cate(sequence, prefixed) whileuser:kv_cate_cntmaps tokv_cate_cnt(non-sequence, unprefixed).Verified live: a grouped
lookup_featurewithmap: "user:kv_cate_cnt",key: "item:cate"and nosequence_fieldsyieldssequence_input_names == ['click_seq__cate'].🤖 Generated with Claude Code