Repository navigation
Converter data: preset-L map revision v1.1 (L's 13-entry dec0_values) - #116
Merged
Merged
Conversation
… "preset": "L") data/preset-L.v1.map.json now holds the spec side's revision v1.1, byte for byte: perm dec0_values is L's own 13-entry table (L has no latent gate), stage dec0's note is updated, and the map names its preset. No tensor reads the perm, so the converted tensors and stage digests are unchanged. The file path stays, so every loader keeps working; without --preset the map now reads as preset L. v1.1 has LF endings, so the line-ending check's CR LF exception goes. Tests: the 13-entry table, the messages with and without --preset; M's map keeps its 17 entries. Refs #89 Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Model L (#89): the spec side's preset-L map revision v1.1 replaces v1 in
data/preset-L.v1.map.json, byte for byte asdelivered. Compared with v1, three things change:
dec0_valuesis L's own 13-entry table[0, 1, 6, 7, 8, 9, 10, 11, 12, 2, 3, 4, 5]. Role order is A0, A1,A2, B0, B1, B2, Cx, Cy, Crho, latent candidate 0..3; role i reads decoder-block channel
dec0_values[i]. L has nolatent gate. v1 carried M's 17-entry table.
dec0's note says the same."preset": "L": the map names its preset (Converter: the map's preset by name (no change for M, K, F) #90's field).No tensor reads the perm, so the converter's tensors and stage digests for L are unchanged. Only the map file's own
hash, recorded in a converted file's header, changes.
name,
test_l_runtime_load.py, and the converter tests. Without--presetthe map now reads as preset L, since itnames its preset.
--preset L(Converter: the preset-L map, and --preset to name a map's preset #106) is still accepted, because it agrees with the field. Other presets are refused.tools/ci/check_line_endings.pydrops the CR LF exception that v1 needed.clean/converter/tests/run_tests.pytest_l_real_map:"preset": "L", anddec0_valueshas 13 entries, a permutation of 0..12;--preset;clean/converter/README.md,data/README.md, the converter's usage text and a comment inmap.hfollow.Not in this pull request: the reference table's own copy of the value positions (
clean/l/reference/table.pyDEC0_VALUES, still M's 17 on main). The open Model L pull request that brings L's 13-entry table there mirrors thismap.
Refs #89
Tested
Local, Windows, CPU only, on main 72c32fc plus this commit:
ctest(unit, decompression-vectors, synthetic, synthetic5): 4/4 passed.run_tests.py: 5210 checks passed, 0 failed (5207 before; the new ones are the 13-entry table, M's 17 and thekey-less map).
clean/l/runtime/tests/test_l_runtime_load.py(reads this map's stages, names, shapes and types): 5 tests OK.python tools/ci/check_line_endings.py: 1175 files, 0 problems. Public gates (secrets, personal data, licence,chunks) pass.
Not tested