JacobLinCool commited on
Commit
09e2da6
·
verified ·
1 Parent(s): cce8292

BarScript C1.3 (sc2_c1_v13_s1234_best) release package: bf16 weights, frozen serving contract, model card

Browse files
Files changed (7) hide show
  1. MODEL_CARD.md +617 -0
  2. config.json +299 -0
  3. model.safetensors +3 -0
  4. preprocessor_config.json +44 -0
  5. release_manifest.json +42 -0
  6. spec.json +89 -0
  7. vocab.json +2080 -0
MODEL_CARD.md ADDED
@@ -0,0 +1,617 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ tags:
4
+ - taiko
5
+ - rhythm-game
6
+ - chart-generation
7
+ - music
8
+ - audio-to-symbolic
9
+ ---
10
+
11
+ # BarScript (experimental preview) — C1.3, denominator augmentation, seed 1234
12
+
13
+ Audio log-mel + a bar grid → a Taiko no Tatsujin chart, decoded under a
14
+ finite-state grammar over bar-scoped plan, skeleton and realization stages.
15
+
16
+ **BarScript** is the release name of this charting model; it is also the name of
17
+ the bar-scoped multi-stage token encoding it decodes into. It is a **preview
18
+ under evaluation**, published so it can be tried, not as a finished model: the
19
+ champion comparison below is `inconclusive`, and §2 lists measured defects — one
20
+ of which fires a pre-registered kill condition. The seven published SoftChart
21
+ 1.x models are a separate, unaffected line.
22
+
23
+ | | |
24
+ |---|---|
25
+ | source checkpoint | `runs/sc2_c1_v13_s1234/best.pt` |
26
+ | sha256 | `44980bf64b1611ea73c1433c41adceff7596cfeece30dfaad5d228eae4a14e6e` |
27
+ | seed / step | 1234 / 96 000 of 100 000 |
28
+ | best validation CE | 0.5769365892960475 |
29
+ | parameters | 9 107 291 |
30
+ | shipped weights | bfloat16 safetensors, 18.2 MB (fp32 checkpoint preserved in the run dir) |
31
+ | training data | `JacobLinCool/taiko-1000-parsed-clean`, revision `b72da4616d643018e81f372cea06ce51349285e0` |
32
+ | label spec | `barscript_labels_v1`, frozen, shipped as `spec.json` (sha1 `cd876768…`) |
33
+ | training flags vs C1 | `--cal-axes big span density` · `--balance-exclude-big` · `--denom-augment 0.5 --denom-augment-mode lattice` |
34
+ | external pretraining | none |
35
+
36
+ **This card is written for THIS checkpoint.** Every number below was measured on
37
+ C1.3's own generations under the serving code this package ships
38
+ (`generate.py` md5 `73b6e49b…`, `min_onset_gap_sec = 0.020`). Nothing is
39
+ inherited from the C1 card. Where an item on the C1 card has no C1.3
40
+ measurement, it is marked **not measured on this arm** rather than carried over.
41
+
42
+ **Read the KNOWN LIMITATIONS section before using any number above it.** This
43
+ model has real, measured defects, one of which fires a pre-registered kill
44
+ condition, and its difficulty ladder is roughly 25 coverage points less nested
45
+ than an authored one.
46
+
47
+ ---
48
+
49
+ ## 0. Status: the champion comparison is `inconclusive`
50
+
51
+ `scripts/sc2_gates.py champion`, three pairings on one serving code base
52
+ (`experiments/sc2_eval/FINAL_CHAMPION.md` §1):
53
+
54
+ | pairing | `champion_verdict` | reason code |
55
+ |---|---|---|
56
+ | C1.2 vs C1 | `inconclusive` | `no_seed_replicate` |
57
+ | **C1.3 vs C1** | **`inconclusive`** | `no_seed_replicate` |
58
+ | C1.3 vs C1.2 | `inconclusive` | `no_seed_replicate` |
59
+
60
+ `MIN_SEEDS_FOR_VERDICT` is 2 in **both** arms and C1.3 has one seed. The rule
61
+ fires before any metric is read. **No arm has been declared champion.**
62
+
63
+ What the clause ledger says, which is one-directional: C1.3 **passes** the
64
+ required §4 clause `plan_token_calibration`, which **C1 fails on both its
65
+ seeds**. No clause is passed by C1 and failed by C1.3. The two `a2` span clauses
66
+ fail on **all four arms** and belong to the campaign, not to this checkpoint.
67
+
68
+ The shipping rationale, the two named compromises and the decision rule for
69
+ choosing a different checkpoint are in
70
+ `experiments/sc2_eval/SHIP_DECISION.md`.
71
+
72
+ Gate items on the 71-chart hard+oni population, this arm against the C1 seed-1234
73
+ arm under identical code:
74
+
75
+ | | C1 | **C1.3** |
76
+ |---|---:|---:|
77
+ | fail / pass / na | 9 / 10 / 6 | **8 / 11 / 6** |
78
+ | newly passing | — | `control_own_axis_density`, `control_named_regression_density_big_count` |
79
+ | newly failing | — | `a2_span_placement_fit` (knife-edge, §2.6) |
80
+
81
+ ---
82
+
83
+ ## 1. What it does
84
+
85
+ * **Input** — 128-bin log-mel at 22 050 Hz (`preprocessor_config.json` freezes
86
+ the exact decode/STFT/mel contract), plus a bar grid with explicit measure
87
+ edges, exact rational meters and per-bar lattice denominators.
88
+ * **Conditions** — course, authored level, density bucket, and the split-axis
89
+ knobs `big_rate` / `span_rate` / `stream` / `sync`.
90
+ * **Output** — a BarScript token stream decoded under `BarscriptFSM`, exported
91
+ as TJA on the 96-slot lattice.
92
+ * **Family decoding** — `generate_chart_family` decodes a song's courses as one
93
+ nested ladder, hardest first, each easier course biased toward its harder
94
+ sibling's onsets. **Off by default**; the campaign's headline population was
95
+ decoded independently (§2.1).
96
+
97
+ Capabilities this checkpoint carries (`config.json:capabilities`): `aux`,
98
+ `beat_head` (hi-res), `hierarchical_ctx`, `axis_knobs`, `sync_token`,
99
+ `stage_emb`, `span_duration_head`, `tempo_head`, `rich_section_stats`. It
100
+ carries **no** `style`, `sibling`, `ctx`, `plan`-prefix, `slot`, `dual`,
101
+ `align`, `mask_infill`, `func_time`, `global_ctx` or `complexity` conditioning;
102
+ requests on those axes reach nothing.
103
+
104
+ ### What works
105
+
106
+ Evidence marks: **●** n ≥ 71 charts with a measured C1 seed band; **◐** n = 19–71
107
+ charts or 8–12 sweep cells, one seed; **○** n < 19 or a CI covering zero.
108
+
109
+ | axis | this checkpoint | reference | mark |
110
+ |---|---|---|---|
111
+ | **Density knob** — median within-song ρ **0.850**, sign p **0.0078**, 8/8 cells positive, **+1.413 nps** first→last, 12 % monotone adjacency | C1 moves the chart +0.024 / +0.151 nps over the same range and **fails** the gate; C1 seed-to-seed Δρ 0.003 (p 1.000) | 8 songs × 5 buckets, oni | ◐ |
112
+ | **`big_rate` knob** — ρ **1.000**, p 0.00049, 91.7 % monotone, and **`calibrated: true`** — mean absolute bucket error **0.533**, within ±1 bucket **93.3 %** | first axis in the whole campaign to clear `calibrated`; C1 0.533 vs 1.200 / 0.967 | 12 songs × 5 buckets, oni | ◐ |
113
+ | **`span_rate` knob** — ρ 0.810, p 0.0117, `calibrated: true`, mean abs err 1.483; and unlike C1 it **no longer drags hit density with it** | C1's span knob flags `onset_nps` / `hit_nps` / `n_onsets` as interference; C1.3 flags only the definitional `span_count` | 12 songs × 5 buckets, oni | ◐ |
114
+ | **Span budget** — 739 spans against the authored 762 (**0.970×**); `span_rate_gen_per_min` gap to authored **0.049** against a C1 seed band of 0.198 | C1 1.160–1.244×, C1.2 1.259× | 97 charts | ● |
115
+ | **FSM guarantees** — **0 unclosed, 0 orphan ends, 0 swallowed hits** over 739 generated spans | holds on all four arms | 97 charts | ● |
116
+ | **Placement does not degrade** — precision 0.6477 (C1 band 0.6480–0.6502), median \|offset\| **0 ms**, `exact_slot_lift_over_null` **1.1755** (above the C1 band), long-song drift tests significant **2** vs C1's 6 / 5 | see §2.3 for why apparent recall drops | 71 charts | ● |
117
+ | **Accent over-emission reduced furthest** — pooled hard+oni big share **0.0873** against C1's 0.1082 / 0.1027 (3.3 seed bands) and an authored 0.0589 | still outside the pre-registered stop window 0.045–0.075 | 71 charts | ● |
118
+ | **Cheaper** — 1 618.6 tokens/chart, 19.22 tokens/bar, `max_window_tokens_p99` **449** | C1 1 731.2 / 20.51 / 515 | 71 charts | ● |
119
+ | **Deployment grid repaired** — on a forced `/16` BPM grid, notes ÷ authored **1.130 / 1.112 / 1.067 / 1.027** and onset F1 **0.333 / 0.374 / 0.466 / 0.580** | C1 1.688 / 1.461 / 1.297 / 1.210 and F1 0.318 / 0.376 / 0.477 / 0.599; C1.3 beats C1.2 on **12 of 12** deployable cells | 13 songs × 4 courses, one seed | ◐ |
120
+ | **Empty-bar declaration restored** — `/16` empty bars 12.82 / 8.60 / 4.55 / 3.14 % against C1.2's 5.96 / 3.23 / 1.08 / 0.33 %, sign-significant on 3 of 4 courses | authored 15.69 / 12.30 / 9.08 / 6.27 % | 13 songs | ◐ |
121
+
122
+ ---
123
+
124
+ ## 2. KNOWN LIMITATIONS
125
+
126
+ Nothing in this section is softened. Where a limitation is invisible to the gate
127
+ suite, that is said.
128
+
129
+ ### 2.1 The difficulty ladder is 25–32 coverage points less nested than authored
130
+
131
+ Coverage = the fraction of the easier chart's notes that have a note in the
132
+ harder chart within tolerance. **Two decode modes, two different numbers, and
133
+ both belong on the record.**
134
+
135
+ | adjacent pair | authored charts | **independent decode** (package default) | **family decode β = 2** (authored grid) | **family decode β = 2** (`/16` deploy grid) |
136
+ |---|---:|---:|---:|---:|
137
+ | easy → normal | 0.970 | **0.658** | 0.730 | 0.726 |
138
+ | normal → hard | 0.979 | **0.667** | 0.797 | 0.811 |
139
+ | hard → oni | 0.985 | **0.660** | 0.822 | 0.834 |
140
+
141
+ Independent-decode figures: `FINAL_CHAMPION.md` §8.1, n = 13 / 13 / 35 songs, one
142
+ seed, same serving code as this package (●/◐). Family-decode figures:
143
+ `C13_PREDICTION.md` §3.3, n = 13 songs, one seed (◐).
144
+
145
+ Five things make this worse than the headline:
146
+
147
+ * **`nesting_coverage` and `hand_agreement` FAIL the family gate on every arm
148
+ measured**, C1 included. `difficulty_monotonicity` passes on all of them.
149
+ * **C1.3 is slightly worse than C1 here.** C1's independent-decode coverage on
150
+ the same population is 0.690 / 0.704 / 0.737; C1.3 is −0.032 / −0.037 / −0.077
151
+ against C1 seed bands of 0.010 / 0.021 / 0.025. The move is small against the
152
+ ~0.30 gap to authored that every arm shares, but it is in the wrong direction.
153
+ * **Hand agreement at coinciding hits is near chance on hard→oni**: C1.3 0.539
154
+ against a marginal-chance null of 0.500 and an authored 0.753. When two
155
+ generated courses agree that a note belongs somewhere, which drum they pick is
156
+ near-independent across courses.
157
+ * **The `/16` figure overstates the model's nesting.** The same checkpoint on a
158
+ `/96` grid drops to 0.649 / 0.730 / 0.752, worse on 11–12 of 13 songs
159
+ (p = .003 / .022 / .022). A coarse lattice manufactures agreement by leaving
160
+ few places to disagree (`C13_PREDICTION.md` §5.2).
161
+ * **`family_bias = 2.0` / `family_hand_bias = 1.5` are PROVISIONAL** — described
162
+ in `generate.py` as logit offsets on a decoder never trained at that setting,
163
+ and never calibrated against the authored target.
164
+
165
+ ### 2.2 The note-type channel carries almost no information
166
+
167
+ 71 charts, **22 260 matched slots** (`FINAL_CHAMPION.md` §5.2, ●).
168
+
169
+ | quantity | this model | its own floor | corpus §2.1 floor |
170
+ |---|---:|---:|---:|
171
+ | 4-way accuracy | **0.4776** | majority **0.5559** | majority 4-way **0.563** |
172
+ | — margin over majority | **−0.0783**, 95 % CI [−0.1022, −0.0530] | | |
173
+ | hand accuracy | **0.5493** | always-don **0.5975** | always-don **0.620** |
174
+ | — margin | **−0.0482**, CI [−0.0721, −0.0229] | | |
175
+ | **MI(generated; authored), 4-way** | **0.01028 bits** | — | **0.79 % of H(authored) = 1.300 bits** |
176
+ | MI, hand | 0.00534 bits | — | 0.55 % of 0.972 bits |
177
+
178
+ * **The model sits below its own majority floor on both endpoints**, and below
179
+ the campaign's corpus baselines (56.3 % / 62.0 %), which are a fixed reference
180
+ line from a separate 59 961-hit census, not this population.
181
+ * **The honest caveat, which cuts the other way.** The published floors are
182
+ *argmax* predictors scored against a temperature-1.0 top-p-0.95 *sample*.
183
+ Against the sampler-appropriate i.i.d. floor (Σ p² = 0.4582 4-way, 0.5190
184
+ hand) this model is **above** baseline by **+0.0194** and **+0.0303** — the
185
+ largest 4-way excess of any arm in the campaign. Both readings belong on the
186
+ record.
187
+ * **MI is flat at ~0.010–0.012 bits on all four arms** (seed band 0.0008). The
188
+ accuracy differences between arms are marginal-matching, not information.
189
+ * Accuracies are conditional on coverage **0.6442** (precision side) /
190
+ **0.5842** (recall side); more than a third of generated hits have no authored
191
+ partner, and coverage is 9 % lower than C1's, so this population is smaller
192
+ and differently selected than C1's.
193
+ * **No inter-charter agreement ceiling exists** for this split — no
194
+ (song, course) carries two independent authored charts — so 100 % is not a
195
+ legitimate target and is not used as one.
196
+ * **Greedy re-decoding was not repeated on this arm.** The C1-era finding that
197
+ greedy closes part of the gap at an unacceptable `motif_reuse` cost is **not
198
+ measured on this arm** and must not be quoted for it.
199
+
200
+ ### 2.3 The note budget is 9 % short, and it is the whole of the apparent timing regression
201
+
202
+ | | C1 (s1234 / s4321) | **C1.3** |
203
+ |---|---:|---:|
204
+ | `onset_precision` | 0.6480 / 0.6502 | **0.6477** |
205
+ | note budget (gen ÷ authored notes) | 1.0046 / 0.9912 | **0.9069** |
206
+ | `onset_recall` | 0.6510 / 0.6445 | **0.5874** |
207
+ | `frac_of_authored_within_jnd` | 0.6463 / 0.6397 | 0.5840 |
208
+ | `type_coverage_recall` | 0.6465 / 0.6401 | 0.5842 |
209
+
210
+ `precision × budget` equals `onset_recall` to four decimals on **every** arm, by
211
+ construction. Those three "timing" rows are one quantity, and it is the note
212
+ budget, not placement: precision is inside the C1 seed band, median offset is
213
+ 0 ms, `exact_slot_lift_over_null` is *above* the band, and significant drift
214
+ tests fall from 6 / 5 to 2.
215
+
216
+ **The shortfall itself is a real regression and its cause is unknown.** It
217
+ appears on both C1.2 and C1.3, so it belongs to the `--cal-axes` /
218
+ `--balance-exclude-big` pair rather than to the denominator augmentation. It is
219
+ the largest practical regression in this release.
220
+
221
+ ### 2.4 Raising density squeezes drumrolls out — this is the compromise a user can hit
222
+
223
+ Median spans/min over eight oni songs, per requested density bucket
224
+ (authored **3.885**; `FINAL_CHAMPION.md` §4.3, 8 songs × 5 buckets, ◐):
225
+
226
+ | requested density | 0 | 3 | 7 | 11 | **15** |
227
+ |---|---:|---:|---:|---:|---:|
228
+ | C1 (s1234) | 3.629 | 3.055 | 4.153 | 2.818 | **2.971** |
229
+ | **C1.3** | 3.550 | 4.413 | 2.172 | 2.727 | **1.279** |
230
+
231
+ ρ(density → `span_per_min`) = **−0.759**, sign p 0.0078, first→last −2.443 /min.
232
+ At the top of the density request C1.3 delivers **33 %** of the authored span
233
+ rate; C1 delivers 76–79 %, C1.2 44 %.
234
+
235
+ * **Do not advertise the density knob at its top setting** until this is fixed.
236
+ * There is a named suspect: `cal_span` never left **0.1667** on either
237
+ challenger, so this is plausibly a calibration term that never trained rather
238
+ than an intrinsic cost of the knob.
239
+ * The mirror image is a *gain*: C1.3's `span_rate` knob no longer drags hit
240
+ density, which C1's does.
241
+
242
+ ### 2.5 Absolute bucket calibration is still wrong on the density axis
243
+
244
+ | axis | exact bucket | within ±1 | mean abs error | mean signed | `calibrated` |
245
+ |---|---:|---:|---:|---:|:--:|
246
+ | density | 5.0 % | 25.0 % | 4.125 | −1.075 | **false** |
247
+ | `big_rate` | **53.3 %** | **93.3 %** | **0.533** | +0.267 | **true** |
248
+ | `span_rate` | — | — | 1.483 | — | true |
249
+
250
+ The density knob orders the request correctly and lands in the wrong bucket,
251
+ under-shooting by ~1 bucket on average — more than C1.2 does. Rank control is
252
+ real; absolute density targeting is not delivered.
253
+
254
+ Note the cost that comes with the `big_rate` calibration win: C1.3 has the
255
+ **shortest** `big_rate` ladder of any arm (first→last +0.159 against C1's
256
+ 0.183 / 0.224), so the top of that request now delivers less accent than C1's
257
+ did.
258
+
259
+ ### 2.6 It fails its own span-quality gate and triggers a pre-registered kill condition
260
+
261
+ 71 charts hard+oni (467 closed spans) and 26 charts easy+normal (272):
262
+
263
+ | axis | hard+oni | easy+normal |
264
+ |---|---|---|
265
+ | `duration_low_tail` | **fail** (rate verdict passes; balloon 0.250 beats = 0.75× the shortest authored balloon, roll 0.250 = 0.60×) | **fail** (balloon 0.250 beats = 0.20× the shortest authored balloon; 45/159 balloons below authored support) |
266
+ | `forced_close` | **fail** — **7 / 467 = 1.499 %** (0 unclosed, 7 clamp-forced) | **fail** — **7 / 272 = 2.574 %** |
267
+ | `over_span_max` | **fail** — one balloon 7.72 s = 1.19× the 6.5 s ceiling | **fail** — one balloon 7.58 s = 1.17× |
268
+ | `placement_fit` | **fail** — AUC **0.560** (CI [0.529, 0.593]) against a 0.56 floor; authored reference 0.668 | pass — AUC 0.604 |
269
+ | `orphans` / `swallowed` / `rate` / `duration` | pass | pass |
270
+ | **`a2_kill_condition_not_triggered`** | **fail — the kill condition FIRED** | fail |
271
+
272
+ The pre-registered kill condition (`SOFTCHART2_DESIGN.md` §4-A2) reads: *rate is
273
+ calibrated but placement/length quality is bad → concede that the bar-scoped
274
+ commitment mechanism is insufficient and escalate to an independent span-state
275
+ objective.* It fires on **all four arms**, including C1. Shipping does not
276
+ discharge it.
277
+
278
+ **`placement_fit` is a knife-edge, not a finding.** C1.2 passes at 0.561 and
279
+ C1.3 fails at 0.560 with CIs overlapping every other arm across their whole
280
+ width. One thousandth of AUC decides the verdict. It is reported, not used.
281
+
282
+ ### 2.7 One span in seven is a micro-span
283
+
284
+ All generated spans, 97 charts, `FINAL_CHAMPION.md` §2 (●):
285
+
286
+ | | n spans | p1 | p5 | p50 | p90 | min | **share < 0.25 s** |
287
+ |---|---:|---:|---:|---:|---:|---:|---:|
288
+ | authored | 762 | 0.177 | 0.273 | 0.963 | 2.419 | 0.083 | **3.8 %** |
289
+ | **this model** | 739 | 0.089 | 0.150 | 0.818 | 2.258 | **0.072** | **14.7 %** |
290
+
291
+ Per course: easy 6.6 %, normal 8.1 %, hard 13.7 %, oni **25.2 %** — authored
292
+ 3.0 / 3.3 / 2.1 / 7.1 %.
293
+
294
+ * C1.3 does **not** inherit C1.2's worsening (16.4 %); 14.7 % is exactly at the
295
+ top of the C1 seed band (13.6–14.7 %). It is not an improvement either.
296
+ * **Mechanism is known and unfixed on every arm.** Duration bucket 0 is open at
297
+ the bottom; 123 of C1.3's 739 spans sit in it, at a median 0.5 beats against
298
+ the authored 0.917.
299
+ * **The shipped floor does not fix it.** `span_min_sec` / `span_min_beats` fired
300
+ 263 times (`span_close_floor`) plus 33 carries across 97 charts and the
301
+ < 0.25 s share is still 14.7 %.
302
+ * **The mitigation is unbuilt.** `span_bucket0_profile` is `null` in this
303
+ package because `scripts/build_span_bucket0_profile.py` has never been run on
304
+ the train split. This is the shortest available fix and it is not in the box.
305
+ * **The gate cannot see it directly.** `a2_span_duration` compares p50/p90/p99
306
+ only and passes; `duration_low_tail` catches the support violation but not the
307
+ rate (its rate verdict passes on this arm).
308
+ * On a `/16` deployment grid the same quantity is 14.7 / 12.4 / **30.6** / 21.8 %
309
+ — hard is untouched by the augmentation and is the worst cell in the release.
310
+
311
+ ### 2.8 The deployment grid has a two-role conflict that no checkpoint dissolves
312
+
313
+ `BAR_DENOM` is simultaneously an input the grid supplies and a **trained output
314
+ token** the model reads as a density announcement. `bar_denoms="supplied"`, the
315
+ shipped default, teacher-forces it.
316
+
317
+ 13 songs × 4 courses, one seed, `--cond authored`, `C13_PREDICTION.md`:
318
+
319
+ | easy / normal / hard / oni | authored grid | **forced `/16`** | `/96` | `/24` |
320
+ |---|---|---|---|---|
321
+ | notes ÷ authored | 0.917 / 0.919 / 0.875 / 0.899 | **1.130 / 1.112 / 1.067 / 1.027** | 0.944 / 0.950 / 0.943 / 1.027 | 1.102 / 1.070 / 0.950 / 0.921 |
322
+ | onset F1 | 0.613 / 0.565 / 0.595 / 0.647 | **0.333 / 0.374 / 0.466 / 0.580** | 0.303 / 0.325 / 0.393 / 0.535 | 0.342 / 0.370 / 0.442 / 0.542 |
323
+ | genuine triplets (den 3/6/12/24) % | 1.13 / 2.48 / 6.34 / 9.43 | **0.00** (arithmetically impossible) | 3.45 / 5.78 / 10.57 / 11.46 | 14.13 / 20.31 / 28.57 / 38.70 |
324
+ | ultra-fine (den 48/96) % | ~0 | 0.00 | **11.53 / 13.22 / 15.62 / 6.60** | 0.00 |
325
+ | nesting (family β = 2) | 0.730 / 0.797 / 0.822 | 0.726 / 0.811 / 0.834 | 0.649 / 0.730 / 0.752 | 0.714 / 0.769 / 0.779 |
326
+
327
+ Authored triplet rates for reference: 4.20 / 5.72 / 7.43 / 9.95 %; ultra-fine
328
+ ~0.00 %.
329
+
330
+ **The recommended deployment configuration is `--grid bpm --grid-bpm-denom 16`
331
+ with the package default `bar_denoms="supplied"`.** Its two costs, stated
332
+ plainly: **triplets are exactly 0 and cannot be otherwise**, and hard micro-spans
333
+ stay at 30.6 % against an authored 2.1 %.
334
+
335
+ Four consequences a deployer must carry:
336
+
337
+ * **Deployment costs 10–46 % of onset F1** relative to the authored grid:
338
+ easy −46 %, normal −34 %, hard −22 %, oni −10 %. Every `--grid authored`
339
+ number in this card is therefore optimistic for a user who supplies only BPM.
340
+ * **`p90 |offset|` is 0 ms on the authored grid and 14.6–36.7 ms on the bpm
341
+ grids.** That residual belongs to the bar edges; no denominator choice touches
342
+ it.
343
+ * **The model's off-dyadic rate is a property of the model, not the lattice.**
344
+ Given any 3-divisible grid it puts 2–3× the authored share off the dyadic
345
+ grid; the grid only decides whether that mass gets called "triplets" (`/24`)
346
+ or "ultra-fine jitter" (`/96`). This is a training-side defect and no
347
+ deployment grid fixes it.
348
+ * **The gate suite cannot score the deployment path.** `robustness.py:629-642`
349
+ raises `ContractError` on uniform-BPM arms, so every gate timing metric is
350
+ `null` for the path the model ships on; the `/16` figures above come from
351
+ `timing.py`, a second implementation that reproduces the gate on the authored
352
+ grid.
353
+
354
+ **The `deploy_bpm_grid` profile in this package is NOT the recommendation.** It
355
+ sets `bar_denoms="model"`, `bar_denom_mask="div96"`, per `DEPLOY_GRID_FIX.md`
356
+ §7 — a recommendation made explicitly conditional on C1.3 failing, which it did
357
+ not. That arm was measured on **C1 only**: density 1.06 / 1.08 / 0.97 / 1.01 and
358
+ nesting 0.635 / 0.706 / 0.778. **It has never been run on this checkpoint.** The
359
+ profile is kept because it is implemented and because the arm is the obvious
360
+ next experiment, not because it is advised here.
361
+
362
+ **One contradiction inside this package, stated so nobody has to find it.**
363
+ `config.json:serving_profiles.deploy_bpm_grid.basis` still quotes
364
+ `DEPLOY_GRID_FIX.md` §7 verbatim, including the words *"Recommended now"*, and
365
+ still carries C1's nesting figures. That string is generated by
366
+ `scripts/sc2_package.py`, which is read-only for this release. **This section
367
+ supersedes it.** The generated string is left intact rather than silently
368
+ diverging from the shipped code.
369
+
370
+ ### 2.9 Playability floor: gaps no human hand can play
371
+
372
+ Recomputed on this arm's own events, 71 charts hard+oni, generated vs authored
373
+ on the same songs (producer `experiments/sc2_eval/ship_c13/playability.py`,
374
+ records `.../records/playability_all_arms.json`; ●):
375
+
376
+ | | this model | authored |
377
+ |---|---:|---:|
378
+ | inter-onset gaps < 40 ms | **102** across **29** charts | 2 across 1 chart |
379
+ | — as a share of adjacent pairs | **0.296 %** (102 / 34 484) | **0.0053 %** (2 / 38 032) |
380
+ | gaps below the series record 29.4118 ms | **19** | 0 |
381
+ | shortest gap | **20.8 ms** (the serving floor) | 39.7 ms |
382
+ | peak burst in any 1 s window | **17 notes** | 15 notes |
383
+
384
+ C1 on the same population: 151 gaps (0.395 %), shortest 16.7 ms, peak 22
385
+ notes/s. C1.3 is better on every row and still **56× the authored rate**.
386
+
387
+ * `min_onset_gap_sec = 0.020` is a **degeneracy guard, not a fix**: it is set
388
+ below the fastest thing the series has shipped (29.4118 ms,
389
+ TAIKO-TONGUE-TWISTER oni, BPM 170, 48th notes) precisely so it cannot refuse a
390
+ chart the domain writes. It fired **125 times** across 97 charts with **0**
391
+ fail-opens. *A mask cannot fix a distribution.*
392
+ * The `generate.py` constant comment quotes the generated sub-40 ms rate as
393
+ ~0.35 % against ~0.045 % authored. **That authored figure does not reconcile
394
+ with the 0.0053 % measured here**, and the two have never been put on the same
395
+ denominator. The excess is 8× on the comment's accounting and 56× on this one.
396
+ * **Good news that replaces a C1-card claim.** The C1 card reported the model
397
+ placing hits in 19.6 % of sub-0.25 s scaffolding bars. Under today's serving
398
+ floors this arm places **0 hits in all 112 such bars** across 97 charts, exactly
399
+ as the authored side does; `degenerate_bar_hits_blocked` fired 12 times with 0
400
+ fail-opens.
401
+
402
+ ### 2.10 Pattern proxies regress against C1 on hard+oni
403
+
404
+ 71 charts, one code version, C1 band = |s1234 − s4321| (`FINAL_CHAMPION.md` §3):
405
+
406
+ | metric (lower is better) | C1 mean | C1 band | **C1.3** | ×band | relative |
407
+ |---|---:|---:|---:|---:|---:|
408
+ | **`motif_reuse_gap`** | 0.3630 | 0.0042 | **0.4222** | +14.1 | **+16.3 %** |
409
+ | `compression_gap` | 0.0976 | 0.0006 | 0.1116 | +23.3 | +14.3 % |
410
+ | `motif_ref_marginal_js` | 0.0331 | 0.0046 | 0.0452 | +2.6 | +36.6 % |
411
+ | `ioi_js_per_chart` | 0.0295 | 0.0006 | **0.0352** | **+9.7** | +19.3 % |
412
+ | `ul_4gram_js_per_chart` | 0.2402 | 0.0055 | 0.2530 | +2.3 | +5.3 % |
413
+ | `class_4gram_js_per_chart` | 0.1337 | 0.0080 | 0.1357 | +0.2 | +1.5 % |
414
+
415
+ The `×band` column overstates the case — several bands are under 1 % of their own
416
+ level — so the relative column is the honest one. On `motif_reuse` the bootstrap
417
+ CIs are nonetheless **disjoint**: C1 [−0.392, −0.334] / [−0.388, −0.334] against
418
+ C1.3 [−0.447, −0.396]. **The regression is real and not seed noise.**
419
+
420
+ Four bounds, all measured, none of them a dismissal:
421
+
422
+ * **It does not happen on easy+normal.** Every pattern proxy there is inside the
423
+ C1 seed band, `motif_reuse_gap` is marginally *better* than C1's mean, and
424
+ `class_4gram_js` is better on both challengers.
425
+ * **The decomposition puts the loss on rhythm and hand, while the accent layer
426
+ improves**: rhythm 0.1251 → 0.1640, hand 0.1774 → 0.2102, accent 0.0606 →
427
+ **0.0480**.
428
+ * **`ioi_js_per_chart` is where the augmentation shows up** — it reads the exact
429
+ rational IOI lattice, and it is C1.3's worst pattern metric relative to C1.2
430
+ (+9.7 band against +1.5).
431
+ * **The proxy reads mostly a channel carrying ~0.010 bits** about the authored
432
+ type (§2.2), and it has **never been validated against a listener**. It is
433
+ quoted because §4 names it, not because it is strong.
434
+
435
+ Realized `motif_reuse` is **0.3165** against an authored **0.7387** — under half
436
+ the authored repetition.
437
+
438
+ ### 2.11 The note-type marginal, and one C1 finding that does NOT reproduce
439
+
440
+ Big notes are **1.48× the authored rate** pooled hard+oni (0.0873 vs 0.0589),
441
+ worst on the easiest course:
442
+
443
+ | course | authored | **C1.3** | C1 (s1234 / s4321) |
444
+ |---|---:|---:|---:|
445
+ | easy | 18.80 % | **26.22 % (1.40×)** | 32.68 % / 29.65 % |
446
+ | normal | 11.75 % | **17.41 % (1.48×)** | 19.21 % / 20.77 % |
447
+ | hard | 7.38 % | **10.61 % (1.44×)** | 13.11 % / 12.73 % |
448
+ | oni | 4.88 % | **7.40 % (1.52×)** | 9.29 % / 8.58 % |
449
+
450
+ The pre-registered stop window (0.045–0.075 pooled hard+oni) is **not met**.
451
+
452
+ **The C1 card's §2.8 claim does not reproduce here and is corrected.** On C1 the
453
+ realized class marginal was 13× closer to the training loss weights than to the
454
+ corpus. On C1.3 it is the other way round: JS(gen ‖ authored) = **0.00588**
455
+ against JS(gen ‖ class-weight prediction) = **0.00746**
456
+ (JS(authored ‖ weights) = 0.01452). `--balance-exclude-big` moved the marginal
457
+ off the loss weights and toward the corpus. The log-log fit of realized share
458
+ against class weight collapses from slope 0.378 (ρ 0.436) on C1 to slope 0.078
459
+ (ρ 0.156) on C1.3.
460
+
461
+ ### 2.12 Fine-lattice, tuplet and syncopation accuracy are much worse than average
462
+
463
+ `robustness.py` timing strata, 71 charts, exact-slot rate (overall **0.5840**):
464
+
465
+ | stratum | n authored | exact-slot | C1 (s1234) |
466
+ |---|---:|---:|---:|
467
+ | lattice step 11.6–23.2 ms | 114 | **0.342** | 0.456 |
468
+ | lattice step 23.2–46.4 ms | 790 | **0.443** | 0.465 |
469
+ | lattice step ≥ 92.9 ms | 24 228 | 0.594 | 0.651 |
470
+ | positions with a denominator divisible by 3 | 3 480 | **0.345** | 0.496 |
471
+ | syncopation band 0 → band 5 | 20 846 → 1 669 | **0.615 → 0.434** | 0.672 → 0.514 |
472
+ | local IOI 25–50 ms | 189 | **0.402** | — |
473
+ | BPM > 250 | 2 570 | **0.499** | — |
474
+
475
+ Every stratum is lower than C1's, which is the note-budget effect of §2.3 acting
476
+ on a per-stratum recall, but the *shape* is the finding: tuplet positions and
477
+ fast lattice steps are 0.24 below the overall rate, and accuracy decays
478
+ monotonically with syncopation.
479
+
480
+ ### 2.13 Smaller, but on the record
481
+
482
+ | finding | number | n | source |
483
+ |---|---|---|---|
484
+ | `long_song_no_drift` fails | **2** of 21 metrics significant at BH q ≤ 0.05 (`dens_signed_err`, `plan_dens_bucket_tvd`) — C1 fails 6 / 5 | 71 charts | `robustness.json` ● |
485
+ | Harness-level generation failures | **6** of 71 song-courses refused (`#BRANCHSTART` and non-representable meters); **1** chart exported `gen_tja = null` — identical to C1 on the same population | 71 charts | `index.json` ● |
486
+ | Slot-export off-lattice events | 8 across 97 charts (C1: 5) — the documented contract path, not a crash | 97 charts | decoder counters ● |
487
+ | `escape_hatch` fires | 91 times across 97 charts (C1: 84) | 97 charts | decoder counters ● |
488
+ | Generation wall time | 9.29 s/chart, 4.33 s per audio-minute — **contended**, the GPU was shared, quoted as a declaration only | 71 charts | ● |
489
+ | Complete-bar rest placement | **not measured on this arm** (C1: Jaccard 0.474, recall 0.623) | — | — |
490
+ | Greedy-decode contrast | **not measured on this arm** | — | — |
491
+ | Micro-span seed sensitivity | **not measured on this arm** (C1: 24.0 % vs 32.4 % across two sampling seeds on 19 oni charts) | — | — |
492
+
493
+ ### 2.14 The evidence base is thinner than it looks
494
+
495
+ * **One seed.** Every number in this card is n = 1 in the training seed. The C1
496
+ seed band quoted throughout is C1's, used as a reproducibility scale; it is a
497
+ range over n = 2 and carries no confidence statement.
498
+ * **easy and normal are evaluated on 13 songs.** The 71-chart population is
499
+ `hard` + `oni` only; the easy/normal population is 26 charts from 13 songs, and
500
+ its seed band is 4–5× wider than hard+oni's.
501
+ * Every difficulty-family and deployment-grid number is **13 songs, one seed**
502
+ (◐). Every knob sweep is **8–12 songs × 5 buckets** (◐).
503
+ * At these sizes a gate `pass` carries little information. Worked example from
504
+ this campaign: the `big_share_of_hits` gate **flips between two C1 training
505
+ seeds** while the failing seed's point estimate is *smaller*. At n = 36 that
506
+ gate cannot rank checkpoints, and no pass/fail on it should be quoted as
507
+ evidence for any arm.
508
+ * **Nothing here is audio-referenced** except `a2_span_placement_fit`. Every
509
+ other quantity is chart-vs-chart on the exact rational lattice or on hit order.
510
+ No timing window and no game judgement parameter appears anywhere in this path.
511
+ * **`best_val` is comparable across C1 / C1.2 / C1.3** (one training cache) and
512
+ **not** comparable to C1.1 or to any arm on a different cache. C1.3's 0.5769 is
513
+ **2.9 % worse** than C1.2's 0.5605; that was the registered trade and the
514
+ deployment side won it.
515
+ * **No plan-neutral arm was generated**, so the `plan_neutral_fallback` §4 clause
516
+ is `na` and the ledger is incomplete by one required clause.
517
+
518
+ ### 2.15 Serving regime: what `motif_constraint: auto` resolves to, and why it matters less here
519
+
520
+ `motif_constraint: "auto"` resolves to **OFF** for this checkpoint: the training
521
+ cache's `barscript_md5` (`634e3dc3…`) differs from the serving `barscript.py`
522
+ (`9afc715a…`), so the gold sequences it trained on never satisfied the MOTIF hard
523
+ constraint. `train_serve_matched = true` — OFF is the train/serve-matched choice
524
+ and it is the right default.
525
+
526
+ **Unlike the C1 package, this card needs no correction for it.** Every number in
527
+ this card was measured with the constraint **OFF**, i.e. under exactly this
528
+ package's resolved default; the 20 evaluation runs all record
529
+ `serving_fsm.motif_constraint: off, verified: true`. In particular
530
+ `motif_ref_marginal_js` here is **0.0452** (hard+oni) and **0.0114**
531
+ (easy+normal), both measured OFF.
532
+
533
+ Two things that still belong on the record:
534
+
535
+ * **The ON/OFF blast radius was measured on C1 only**, where switching the
536
+ constraint on moved `motif_ref_marginal_js` 0.0311 → 0.0189 (−39 %). On the C1
537
+ package the shipped default therefore makes the correct value **0.0311, not the
538
+ 0.0189 in the older tables**. **That contrast has never been measured on
539
+ C1.3**, so no ON-constraint number should be quoted for this checkpoint.
540
+ * Serving under the constraint ON would be train/serve **mismatched** for this
541
+ checkpoint and is not a supported configuration.
542
+
543
+ ---
544
+
545
+ ## 3. Serving contract
546
+
547
+ Every value below is in `config.json:serving`, with its justification in
548
+ `config.json:serving_basis`. Pass them explicitly — the package's
549
+ `serving_kwargs()` does — so the recorded contract is the one that reaches the
550
+ decoder.
551
+
552
+ | parameter | value | one-line basis |
553
+ |---|---|---|
554
+ | `min_onset_gap_sec` | **0.020** | Degeneracy guard, **not** a corpus percentile. Must stay strictly below the series record of 29.4118 ms (BPM 170, 48ths). Two earlier corpus-derived values (0.0395, 0.0300) were both wrong. See §2.9. |
555
+ | `degenerate_bar_sec` | **0.25** | Authored charts place zero hits in any bar under 0.375 s (3 010 bars). Largest round threshold with zero counterexamples and 1.5× margin; identical to the gate's threshold. Blocks hits only, never span geometry. |
556
+ | `span_min_sec` / `span_min_beats` | 0.0833 / 1⁄3 | Authored population minima. Fail-open, counted. Does not fix §2.7. |
557
+ | `span_bucket0_profile` | **null** | Not built — see §2.7. |
558
+ | `family_bias` / `family_hand_bias` / `family_mode` | 2.0 / 1.5 / `bias` | **Provisional**, never calibrated — see §2.1. Family decode is opt-in. |
559
+ | `motif_constraint` | `auto` → resolves **off** here | Train/serve matching by `barscript.py` md5 — see §2.15. |
560
+ | `bar_denoms` / `bar_denom_mask` | `supplied` / `lattice` | The evaluated regime, and the recommended deployment regime on a uniform `/16` grid. The `deploy_bpm_grid` profile exists but is **not** recommended for this checkpoint — see §2.8. |
561
+ | `greedy` / `temperature` / `top_p` | false / 1.0 / 0.95 | As evaluated. |
562
+ | `plan_temperature` | `null` (follows `temperature`) | Never exercised in evaluation; shipped unset rather than tuned. |
563
+
564
+ `_has_sync` **must** be forced on at load. `train.py` records
565
+ `sync_token=False` while the training prefix carries the SYNC slot, so a loader
566
+ that trusts the recorded flag drops the slot and every sync bucket reaches the
567
+ decoder as an identical prefix. The package loader does this and records why in
568
+ `config.json:serving_prefix_fix`. This is a workaround; the fix belongs in
569
+ `train.py`.
570
+
571
+ ### Recommended deployment configuration
572
+
573
+ ```
574
+ --grid bpm --grid-bpm-denom 16 # uniform /16 grid
575
+ bar_denoms = "supplied" # the package default
576
+ motif_constraint = auto -> off # resolved at package time
577
+ family decode: optional; β = 2.0 / 1.5 is provisional
578
+ ```
579
+
580
+ Expected behaviour under it, 13 songs × 4 courses, one seed: notes within
581
+ 13 / 11 / 7 / 3 % of authored; onset F1 0.333 / 0.374 / 0.466 / 0.580;
582
+ **zero triplets**; hard micro-spans ~30 %.
583
+
584
+ ---
585
+
586
+ ## 4. Weights are bfloat16
587
+
588
+ Training ran bf16 autocast and CUDA inference runs bf16 autocast, so fp32
589
+ storage carried no information the forward pass could use — the argument the
590
+ v1.5 release made and verified. Measured cast cost: max absolute delta
591
+ **0.00711** on `frontend.2.weight` (0.349 % of that tensor's max) over 9 107 291
592
+ float elements.
593
+
594
+ This is an argument about the **compute** dtype, not a proof that the two
595
+ checkpoints decode identically. Sampling is chaotic in the logits, so individual
596
+ charts can differ. The paired fp32-vs-bf16 evaluation has **not** been repeated
597
+ for BarScript. The fp32 checkpoint is preserved in the source run directory and
598
+ is the reference for any bit-level comparison.
599
+
600
+ ---
601
+
602
+ ## 5. Provenance
603
+
604
+ | | |
605
+ |---|---|
606
+ | serving code | `generate.py` md5 `73b6e49b00fb60f3229c8539bc7a1029`; all 18 `src/softchart/*.py` md5s in `config.json:code.module_md5` |
607
+ | evaluation harness | `sc2_generate_eval_v2_2_serving_floors`, contract `sc2_eval_v2` |
608
+ | evaluation runs | 20 run indices, 1 028 charts, one `serving_floors` signature, `comparable: true`, 6/6 floor counters present |
609
+ | headline population | 71 charts hard+oni (36 songs) + 26 charts easy+normal (13 songs), `--cond authored`, `--grid authored`, seed 1 |
610
+ | deployment population | 13 songs × 4 courses × 10 arms, `--cond authored`, one serving tree, one signature |
611
+ | label spec | `spec.json`, field-for-field equal to the cache manifest's copy (`spec_parameters_match_cache_manifest: true`) |
612
+ | uploaded | **no** (`release_manifest.json: "uploaded": false`) |
613
+
614
+ Reports: `experiments/sc2_eval/SHIP_DECISION.md`,
615
+ `experiments/sc2_eval/FINAL_CHAMPION.md`,
616
+ `experiments/sc2_eval/C13_PREDICTION.md`,
617
+ `experiments/sc2_eval/C12_EVAL.md`, `experiments/sc2_eval/DEPLOY_GRID_FIX.md`.
config.json ADDED
@@ -0,0 +1,299 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "capabilities": {
3
+ "align": false,
4
+ "aux": true,
5
+ "axis_knobs": true,
6
+ "beat_head": true,
7
+ "complexity": false,
8
+ "ctx": false,
9
+ "dual": false,
10
+ "func_time": false,
11
+ "global_ctx": false,
12
+ "hierarchical_ctx": true,
13
+ "mask_infill": false,
14
+ "plan": false,
15
+ "sibling": false,
16
+ "slot": false,
17
+ "style": false,
18
+ "sync_token": true
19
+ },
20
+ "code": {
21
+ "git": {
22
+ "branch": null,
23
+ "commit": null,
24
+ "dirty": false
25
+ },
26
+ "module_md5": {
27
+ "src/softchart/__init__.py": "d41d8cd98f00b204e9800998ecf8427e",
28
+ "src/softchart/barscript.py": "9afc715a56ac28e9ec5675d64cd0536c",
29
+ "src/softchart/baseline.py": "875b63450c59bb62f5d50965ec7b218c",
30
+ "src/softchart/checkpoint.py": "10e215e86ac77ada9046a164bffa0644",
31
+ "src/softchart/data.py": "d0cf18237bc644f49bf2b62b9eef0ee3",
32
+ "src/softchart/evaluate.py": "72d588b7f0711dfa6d1f49b161172061",
33
+ "src/softchart/generate.py": "73b6e49b00fb60f3229c8539bc7a1029",
34
+ "src/softchart/grid.py": "eba9fddd432bee0b62ddfc13bdefc06d",
35
+ "src/softchart/hf.py": "727db85293862948587c99636ea1d1ef",
36
+ "src/softchart/meter.py": "18ccd6a003086f81276f2a916ada4c64",
37
+ "src/softchart/model.py": "eace5d89abeebf0edcf28d10b5ad3e68",
38
+ "src/softchart/preprocess.py": "b87fa0cc8af9628ae47a2fbe86de266a",
39
+ "src/softchart/proxies.py": "918c72c22d48e4ccb762e0fe814d0bfe",
40
+ "src/softchart/rhythm.py": "11cca7c58809b339f7384dcc0bf1fcaa",
41
+ "src/softchart/tja.py": "360b95e19dabe925bf8eab03fa3fd110",
42
+ "src/softchart/trace.py": "e707f7204f0d86deacca6b6325f031f1",
43
+ "src/softchart/train.py": "7f6323973c26992873717307f6652f91",
44
+ "src/softchart/vocab.py": "0f948a0efc52f5ce60e8219fcaeb136c"
45
+ },
46
+ "torch": "2.9.1+cu128"
47
+ },
48
+ "created_utc": "2026-07-26T00:17:52+00:00",
49
+ "model": {
50
+ "adapter_rank": 0,
51
+ "adapter_rank_ffn": 0,
52
+ "aux": true,
53
+ "beat": "hires",
54
+ "clean_phase": true,
55
+ "d_model": 256,
56
+ "dec_layers": 4,
57
+ "dec_share": null,
58
+ "depth_emb": false,
59
+ "dropout": 0.1,
60
+ "emb_factor": 64,
61
+ "enc_layers": 4,
62
+ "enc_share": null,
63
+ "ffn": 960,
64
+ "func_time": false,
65
+ "global_ctx": false,
66
+ "hierarchical_ctx": true,
67
+ "in_ch": 128,
68
+ "max_tgt": 960,
69
+ "nhead": 8,
70
+ "ptr": false,
71
+ "rich_section_stats": true,
72
+ "section_ffn": 512,
73
+ "section_layers": 2,
74
+ "span_duration_head": true,
75
+ "stage_emb": true,
76
+ "tempo_head": true,
77
+ "unique_layernorm": false,
78
+ "unshare_last_dec": false,
79
+ "vocab_size": 2071
80
+ },
81
+ "release_name": "sc2_c1_v13_s1234_best",
82
+ "schema_version": "softchart2_release_config_v1",
83
+ "serving": {
84
+ "bar_denom_mask": "lattice",
85
+ "bar_denoms": "supplied",
86
+ "degenerate_bar_sec": 0.25,
87
+ "family_bias": 2.0,
88
+ "family_hand_bias": 1.5,
89
+ "family_mode": "bias",
90
+ "greedy": false,
91
+ "min_onset_gap_sec": 0.02,
92
+ "motif_constraint": "auto",
93
+ "plan_temperature": null,
94
+ "span_bucket0_profile": null,
95
+ "span_min_beats": 0.3333333333333333,
96
+ "span_min_sec": 0.0833,
97
+ "temperature": 1.0,
98
+ "top_p": 0.95
99
+ },
100
+ "serving_basis": {
101
+ "bar_denom_mask": "Only consulted when bar_denoms='model'. 'lattice' = divisors of the bar's own TJA export lattice; a superset of 'div96' outside 4/4.",
102
+ "bar_denoms": "'supplied' teacher-forces the grid's per-bar denominators into every [BAR] header. It is the regime this checkpoint was TRAINED and EVALUATED under -- every published number belongs to it -- so it is the DEFAULT, and it is correct when the grid comes from a real authored/traced chart. It is the WRONG value on a synthetic uniform BPM grid: BAR_DENOM is a trained output token and the model reads a supplied denominator as a DENSITY ANNOUNCEMENT, so a forced /16 grid emits 1.69x / 1.46x / 1.30x / 1.21x the authored onsets on easy/normal/hard/oni and erases every triplet (off-/16 onsets 0 % against an authored 7.4-14.5 %). Use the 'deploy_bpm_grid' profile there. DEPLOY_GRID_FIX.md §5.2, §7; 13 songs x 4 courses, one seed.",
103
+ "degenerate_bar_sec": "Across 3 010 bars the authored side places ZERO hits in any bar shorter than 0.375 s (28 such bars, 2 songs); the shortest authored hit-bearing bar is 0.375 s. 0.25 s is the largest round threshold with zero authored counterexamples and a 1.5x margin, and it is the threshold the degenerate_bar_empty gate uses, so fix and gate cannot disagree. Blocks HITS only: authors DO route span geometry through sub-beat scaffolding measures. Source: src/softchart/generate.py DEGENERATE_BAR_SEC comment.",
104
+ "family_bias": "generate.FAMILY_BIAS. Additive logit bias toward the harder sibling's onset POSITIONS when courses are decoded as one nested difficulty family (generate_chart_family). 0 is the independent decode; 'subset' mode is the bias=infinity ablation arm.",
105
+ "family_hand_bias": "generate.FAMILY_HAND_BIAS. Additive logit bias toward the harder sibling's don/ka HAND at an inherited position. Steers type only, never position.",
106
+ "family_mode": "'bias' is the soft steering and the shipped value; 'subset' hard-restricts the easier course to the harder chart's positions and is kept for ablation only.",
107
+ "greedy": "Sampling, not argmax -- the evaluation population was produced at temperature 1.0 and every published number belongs to that decoder.",
108
+ "min_onset_gap_sec": "DEGENERACY GUARD, not a corpus percentile. Two earlier revisions (0.0395 s from a 43 387-pair subset; 0.0300 s, the 1 200-chart / 427 016-pair corpus minimum) were both wrong because a 1 000-song corpus is not the domain. The series record for two adjacent notes is 29.4118 ms (TAIKO-TONGUE-TWISTER oni, BPM 170, 48th notes, 34 hits/s); a 30 ms floor blocks the record itself. 0.020 s clears every documented chart with headroom and still blocks the observed degeneracy (six ka notes 16.7 ms apart rendering as one blob). The model's real defect on this axis is a RATE, not a value -- sub-40 ms gaps are 0.045 % of authored pairs and ~0.35 % of generated ones -- and a mask cannot fix a distribution. Source: src/softchart/generate.py MIN_ONSET_GAP_SEC comment.",
109
+ "motif_constraint": "'auto' = ON iff the checkpoint's own training cache (<cache>/<args.bs_subdir>/manifest.json) records a barscript_md5 equal to the md5 of the serving barscript.py. Only then did the gold sequences the checkpoint trained on satisfy the MOTIF hard constraint the serving FSM would apply, so only then is serving under it train/serve matched. This package records the RESOLVED value and its inputs under train_provenance.motif_constraint. Source: scripts/sc2_generate_eval.py MOTIF_AUTO_RULE (harness contract v2.1).",
110
+ "plan_temperature": "null = follow `temperature`. The four plan steps have an independent diversity knob; it was not exercised in the evaluation, so the package ships it unset rather than tuned.",
111
+ "span_bucket0_profile": "DISABLED (null) -- and this is a KNOWN GAP, not a choice on the merits. Duration bucket 0 is open at the bottom, so nothing bounds how short a span may be beyond the corpus floor and the realized length piles up on the lattice minimum; this checkpoint puts 21.2 % of its spans under 0.25 s against an authored 4.2 %. The mitigation exists in code (generate.SpanBucket0Profile / generate_song_barscript(span_bucket0_profile=...)) but NO PROFILE ARTIFACT HAS BEEN BUILT -- scripts/build_span_bucket0_profile.py has never been run on the train split, and no such file exists on the training host. Absent it the serving path is bit-identical to the shipped one. Building it is the shortest available micro-span fix.",
112
+ "span_min_beats": "Beat-relative companion to span_min_sec (1/3 quarter-note beat), so the floor survives a tempo change. Source: src/softchart/generate.py SPAN_MIN_BEATS.",
113
+ "span_min_sec": "Authored population minimum span duration; steers span closures and masks duration buckets away from lengths no author wrote. Fail-open at the mask level, counted as span_close_floor / span_floor_carry / dur_bucket_floor. Source: src/softchart/generate.py SPAN_MIN_SEC.",
114
+ "temperature": "1.0, as evaluated.",
115
+ "top_p": "0.95, as evaluated."
116
+ },
117
+ "serving_prefix_fix": "sync_token is forced ON at serving. train.py records sync_token=False, but the training prefix carries the SYNC slot, so a loader MUST set model._has_sync = True before decoding or every sync bucket reaches the decoder as the same prefix. sc2_package.load_release does this; see the SYNC note in scripts/sc2_generate_eval.py.",
118
+ "serving_profiles": {
119
+ "default": {
120
+ "overrides": {},
121
+ "when": "the supplied grid comes from a real chart trace (barscript_grid_from_trace) -- i.e. the bar denominators are the authored ones. This is the regime every published metric was measured in."
122
+ },
123
+ "deploy_bpm_grid": {
124
+ "basis": "DEPLOY_GRID_FIX.md §7, verbatim: 'Recommended now: model-emitted BAR_DENOM, masked to divisors of 96.' What it buys: note counts within 8 % of authored at every difficulty instead of 21-69 % over; the DENS plan back on the authored-grid distribution; triplets back from zero; span counts back to authored levels. What it COSTS, and this is not small: nesting coverage 0.635 / 0.706 / 0.778 against 0.835 / 0.845 / 0.866, worse on 12-13 of 13 songs, p <= .003 on all three course pairs -- the family decode's headline benefit is gone. What it does NOT fix: p90 onset offset 7-34 ms, frac_within_jnd 0.34-0.40, micro-spans 26-42 % on hard/oni. The report states the tradeoff as the finding: 'No arm here is acceptable on both axes, and that is the finding. The choice is between a chart whose difficulty is right and a chart whose difficulty ladder is nested, because one token controls both.' Evidence: 13 songs x 4 courses, one seed, this checkpoint. NOTE: DEPLOY_GRID_FIX measured this arm by runtime injection because bar_denoms='model' was then only a proposed interface; it is now implemented in generate_song_barscript, and this profile selects it directly.",
125
+ "overrides": {
126
+ "bar_denom_mask": "div96",
127
+ "bar_denoms": "model"
128
+ },
129
+ "when": "the grid is synthesized from user BPM/offset (barscript_grid_from_bpm), where every bar carries the same denominator and the model would read it as a density announcement."
130
+ }
131
+ },
132
+ "source": {
133
+ "best_val": 0.5769365892960475,
134
+ "checkpoint_file": "best.pt",
135
+ "checkpoint_select": "best",
136
+ "checkpoint_sha256": "44980bf64b1611ea73c1433c41adceff7596cfeece30dfaad5d228eae4a14e6e",
137
+ "mode": "c1",
138
+ "run_dir": "/home/jacob/softchart2/runs/sc2_c1_v13_s1234",
139
+ "seed": 1234,
140
+ "step": 96000
141
+ },
142
+ "train_provenance": {
143
+ "barscript_md5": "634e3dc3ddca8bcdceffd760e663e1b0",
144
+ "bs_subdir": "barscript",
145
+ "cache_manifest": "cache/barscript/manifest.json",
146
+ "dataset": "/home/jacob/softchart2/data/taiko-1000-parsed-clean",
147
+ "labels_version": "barscript_labels_v1",
148
+ "motif_constraint": {
149
+ "requested": "auto",
150
+ "resolved_motif_hard": false,
151
+ "serving_barscript_md5": "9afc715a56ac28e9ec5675d64cd0536c",
152
+ "source": "auto",
153
+ "train_cache_barscript_md5": "634e3dc3ddca8bcdceffd760e663e1b0",
154
+ "train_cache_built_by_serving_barscript": false,
155
+ "train_serve_matched": true
156
+ },
157
+ "revision": "b72da4616d643018e81f372cea06ce51349285e0",
158
+ "serving_barscript_md5": "9afc715a56ac28e9ec5675d64cd0536c",
159
+ "serving_barscript_path": "/home/jacob/softchart2/pkgstage/src/softchart/barscript.py",
160
+ "spec_file": "spec.json",
161
+ "spec_frozen_path": "/home/jacob/softchart2/cache/barscript_spec_frozen.json",
162
+ "spec_frozen_sha1": "cd8767684457915ece0d3d58ab80ade137ad79f5",
163
+ "spec_parameters_match_cache_manifest": true,
164
+ "spec_sha1": "cd8767684457915ece0d3d58ab80ade137ad79f5",
165
+ "spec_sha1_matches_cache_manifest_record": false,
166
+ "spec_sha1_matches_frozen_file": true,
167
+ "spec_sha1_note": "Three sha1s, and they are allowed to differ. spec_sha1 is the file shipped here; spec_frozen_sha1 is the spec artifact currently on the training host; spec_sha1_recorded_in_cache_manifest is what the cache builder recorded when the cache was built. The binding claim is spec_parameters_match_cache_manifest -- the bucket edges, motif tau and deltas shipped here are field-for-field the ones the checkpoint trained under. A sha1 mismatch means a re-serialization, not a different label contract.",
168
+ "spec_sha1_recorded_in_cache_manifest": "0893edb05b6473b983faff2aab166bcc18b4b3d6",
169
+ "train_args": {
170
+ "adapter_rank": 0,
171
+ "adapter_rank_ffn": 0,
172
+ "align": false,
173
+ "aux": true,
174
+ "axis_knobs": true,
175
+ "balance_cap": 2.5,
176
+ "balance_exclude_big": true,
177
+ "balanced": true,
178
+ "batch": 64,
179
+ "beat_all_window": false,
180
+ "beat_head": true,
181
+ "beat_hires": true,
182
+ "beat_weight": 1.0,
183
+ "bs_subdir": "barscript",
184
+ "cache": "/home/jacob/softchart2/cache",
185
+ "cal_axes": [
186
+ "big",
187
+ "span",
188
+ "density"
189
+ ],
190
+ "cal_dens_margin_frac": 0.05,
191
+ "cal_margin": 0.5,
192
+ "cal_rows": 16,
193
+ "cal_weight": 0.05,
194
+ "clean_phase": true,
195
+ "complexity": false,
196
+ "cond_corrupt": 0.1,
197
+ "cond_drop": 0.15,
198
+ "ctx": false,
199
+ "d_model": 256,
200
+ "data_frac": 1.0,
201
+ "dec_layers": 4,
202
+ "dec_share": null,
203
+ "denom_augment": 0.5,
204
+ "denom_augment_mode": "lattice",
205
+ "depth_emb": false,
206
+ "distill": null,
207
+ "distill_tau": 2.0,
208
+ "distill_teacher": null,
209
+ "distill_weight": 0.5,
210
+ "dpo": null,
211
+ "dpo_beta": 0.1,
212
+ "dpo_hard": false,
213
+ "dropout": 0.1,
214
+ "dual": false,
215
+ "dual_slot_p": 0.65,
216
+ "emb_factor": 64,
217
+ "enc_layers": 4,
218
+ "enc_share": null,
219
+ "entropy_floor": 0.0,
220
+ "entropy_floor_weight": 0.1,
221
+ "ffn": 960,
222
+ "func_time": false,
223
+ "global_ctx": false,
224
+ "grammar_mask": true,
225
+ "hierarchical_ctx": true,
226
+ "importance_sampling": true,
227
+ "init_from": null,
228
+ "kd_alpha": 0.5,
229
+ "kd_tau": 2.0,
230
+ "lambda_plan": 1.0,
231
+ "lambda_type": 1.5,
232
+ "lr": 0.0005,
233
+ "lr_shared_core": 1.0,
234
+ "mask_infill": 0.0,
235
+ "max_tgt": 960,
236
+ "mode": "c1",
237
+ "nhead": 8,
238
+ "no_augment": false,
239
+ "out": "/home/jacob/softchart2/runs/sc2_c1_v13_s1234",
240
+ "plan": false,
241
+ "plan_dropout": 0.3,
242
+ "plan_jitter": 0.15,
243
+ "resume": null,
244
+ "rich_section_stats": true,
245
+ "section_ffn": 512,
246
+ "section_layers": 2,
247
+ "seed": 1234,
248
+ "sibling": false,
249
+ "slot": false,
250
+ "span_dur_weight": 1.0,
251
+ "span_duration_head": true,
252
+ "stage_emb": true,
253
+ "steps": 100000,
254
+ "style": false,
255
+ "sync_token": false,
256
+ "tempo_aug": false,
257
+ "tempo_head": true,
258
+ "tempo_weight": 0.2,
259
+ "unique_layernorm": false,
260
+ "unlikelihood_alpha": 1.0,
261
+ "unlikelihood_scope": "hits",
262
+ "unshare_last_dec": false,
263
+ "use_aug": false,
264
+ "val_every": 2000,
265
+ "warmup": 1000,
266
+ "workers": 12
267
+ }
268
+ },
269
+ "v2_flags": {
270
+ "axis_knobs": true,
271
+ "rich_section_stats": true,
272
+ "span_duration_head": true,
273
+ "stage_emb": true,
274
+ "tempo_head": true
275
+ },
276
+ "vocab": {
277
+ "checkpoint_rows": 2071,
278
+ "growth_policy": "append-only. checkpoint.expand_vocab_rows grows an older checkpoint's tok_emb (and a tied out.weight) to the current table with N(0, 0.02) rows under seed 20260725; a checkpoint with MORE rows than the runtime is refused.",
279
+ "mapping_sha256": "f139fbea1a13397015088b1ab2c33a952e97b05bb447bc86ce02f350d50dd679",
280
+ "max_tgt": 960,
281
+ "max_tgt_source": "checkpoint args",
282
+ "size": 2071,
283
+ "slots_per_4_4": 96,
284
+ "window_frames": 1728
285
+ },
286
+ "weights": {
287
+ "cast_cost": {
288
+ "float_elements": 9107291,
289
+ "max_abs_delta": 0.007111072540283203,
290
+ "max_abs_delta_relative_to_tensor_max": 0.0034862233048405743,
291
+ "max_abs_delta_tensor": "frontend.2.weight",
292
+ "note": "bfloat16 keeps 8 mantissa bits, so a cast costs up to 2^-9 relative. Training ran bf16 autocast (train.py) and inference runs bf16 autocast on CUDA (generate._autocast), so fp32 storage carried no information the forward pass could use -- the same argument the v1.5 release made and verified. It is an argument about the COMPUTE dtype, not a proof that the two checkpoints decode identically: sampling is chaotic in the logits, so individual charts can and do differ. fp32 is preserved in the source run directory."
293
+ },
294
+ "cast_from": "float32",
295
+ "dtype": "bfloat16",
296
+ "file": "model.safetensors",
297
+ "parameters": 9107291
298
+ }
299
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce60e24a68903e5c179ab038efe4e0444722298ab9ad6106fefb6f33ccd8488d
3
+ size 18232990
preprocessor_config.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio": {
3
+ "decode_channels": 2,
4
+ "decoded_dtype": "little_endian_float32",
5
+ "decoder": "ffmpeg_f32le_stereo_mean_22050_v1",
6
+ "mono_mix": "arithmetic_mean_float32",
7
+ "sample_rate_hz": 22050
8
+ },
9
+ "cache_dtype": "float16",
10
+ "dataset": "JacobLinCool/taiko-1000-parsed-clean",
11
+ "dataset_revision": "b72da4616d643018e81f372cea06ce51349285e0",
12
+ "density_bucket_nps": 0.75,
13
+ "frames_per_second": 86.1328125,
14
+ "log": {
15
+ "additive_floor": 1e-05,
16
+ "base": "natural"
17
+ },
18
+ "mel": {
19
+ "fmax_hz": 11025.0,
20
+ "fmin_hz": 20.0,
21
+ "htk": false,
22
+ "implementation": "librosa.filters.mel",
23
+ "n_mels": 128,
24
+ "norm": "slaney"
25
+ },
26
+ "meter": {
27
+ "schema": "rational_96_per_4_4_per_course_bounded_safe_v2",
28
+ "slot_safety": "constant_note_bpm_zero_delay_duration_match_event_lattice_v2",
29
+ "terminal_policy": "bounded_intervals_only_v1"
30
+ },
31
+ "model_input_dtype": "float32",
32
+ "schema_version": "softchart_logmel_preprocessing_v1",
33
+ "stft": {
34
+ "center": true,
35
+ "hop_length": 256,
36
+ "implementation": "torch.stft",
37
+ "n_fft": 2048,
38
+ "pad_mode": "reflect",
39
+ "power": 2,
40
+ "window": "torch.hann_window_periodic_true"
41
+ },
42
+ "vocab_mapping_sha256": "f139fbea1a13397015088b1ab2c33a952e97b05bb447bc86ce02f350d50dd679",
43
+ "window_frames": 1728
44
+ }
release_manifest.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config_sha256": "a045e81de4149e0e0650f867f8236fe06b1e65039e134bd16fe413ddfb0da465",
3
+ "created_utc": "2026-07-26T00:17:52+00:00",
4
+ "files": {
5
+ "MODEL_CARD.md": {
6
+ "sha256": "601d0090a0cda9bb670b414338326efa5521cba56ac15e14c9863899da79cdf0",
7
+ "size_bytes": 35953
8
+ },
9
+ "config.json": {
10
+ "sha256": "b5a6a11bd069cb5b3adf33dc50ff194a56cfcc0f9f4a51b78c84d954b89f7d3e",
11
+ "size_bytes": 16554
12
+ },
13
+ "model.safetensors": {
14
+ "sha256": "ce60e24a68903e5c179ab038efe4e0444722298ab9ad6106fefb6f33ccd8488d",
15
+ "size_bytes": 18232990
16
+ },
17
+ "preprocessor_config.json": {
18
+ "sha256": "5788284beb27678e6d3a36054b0319e1f90d947aeab40d786792ce410df92bba",
19
+ "size_bytes": 1291
20
+ },
21
+ "spec.json": {
22
+ "sha256": "7f248652e7fd25649b00de6adf3754fc3dd497cdefce4b1838738676d934f955",
23
+ "size_bytes": 2618
24
+ },
25
+ "vocab.json": {
26
+ "sha256": "3caa9c70d98fd71aac581789abd6899b370397eeacaa00421b8ac13adc349b8e",
27
+ "size_bytes": 35180
28
+ }
29
+ },
30
+ "parameters": 9107291,
31
+ "post_build_edits": [
32
+ "MODEL_CARD.md was edited IN PLACE after the packager ran, to rename the model from 'SoftChart 2.0' to its release name 'BarScript'. Only prose changed: the H1 title, one added paragraph naming the model and its preview status, and one sentence in section 4. No measurement, no identifier and no other file was touched -- the run name sc2_c1_v13_s1234, every md5/sha and config.json are byte-identical to the build. files.MODEL_CARD.md above records the EDITED bytes, which are the bytes that shipped, so manifest verification stays sound; the packager-built card hashed 7afeb0462875a5a928e4f55df5d8220d4e7feaa1801ccaee816ebce6d09c0f67. SOFTCHART2_RELEASE.md section 4 prescribes a package rebuild for card changes; a rebuild was not run because it needs the source checkpoint on the training machine and would re-serialize model.safetensors (section 4.1: header bytes are not order-stable), which would discard the provenance this package already carries."
33
+ ],
34
+ "release_name": "sc2_c1_v13_s1234_best",
35
+ "roundtrip_verified": true,
36
+ "schema_version": "softchart2_release_package_v1",
37
+ "source_checkpoint": "/home/jacob/softchart2/runs/sc2_c1_v13_s1234/best.pt",
38
+ "source_sha256": "44980bf64b1611ea73c1433c41adceff7596cfeece30dfaad5d228eae4a14e6e",
39
+ "upload_note": "This package has NOT been published. Uploading is a human decision; the commands are in SOFTCHART2_RELEASE.md and are not executed by this script, which never contacts the network.",
40
+ "uploaded": false,
41
+ "weights_dtype": "bfloat16"
42
+ }
spec.json ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "labels_version": "barscript_labels_v1",
3
+ "dens_edges": [
4
+ 1,
5
+ 2,
6
+ 3,
7
+ 4,
8
+ 5,
9
+ 6,
10
+ 9
11
+ ],
12
+ "span_dur_edges": [
13
+ 0.9167,
14
+ 1.1192,
15
+ 1.3665,
16
+ 1.6685,
17
+ 2.0372,
18
+ 2.4874,
19
+ 3.037,
20
+ 3.7081,
21
+ 4.5275,
22
+ 5.5279,
23
+ 6.7495,
24
+ 8.2409,
25
+ 10.0619,
26
+ 12.2853,
27
+ 15.0
28
+ ],
29
+ "sync_edges": [
30
+ 1,
31
+ 2,
32
+ 3,
33
+ 4,
34
+ 5
35
+ ],
36
+ "motif_tau": 0.3,
37
+ "motif_deltas": [
38
+ 1,
39
+ 2,
40
+ 4,
41
+ 8
42
+ ],
43
+ "provenance": {
44
+ "frozen": true,
45
+ "ruling_date": "2026-07-25",
46
+ "revision": 2,
47
+ "supersedes": {
48
+ "artifact_md5": "7d5f3c2293d02200bc18b4aa05b12b80",
49
+ "note": "rev 1 was frozen on the pre-normalization population (3,365 courses / 283,234 bars, dens_edges (1,2,3,4,5,7,9), span_dur hi edge 14.9167)"
50
+ },
51
+ "generator_revision": "b72da461",
52
+ "dataset": "taiko-1000-parsed-clean (local parquet), split=train",
53
+ "n_songs": 924,
54
+ "n_courses": 3682,
55
+ "n_bars": 311655,
56
+ "span_dur_anchors_beats": {
57
+ "p50": 2.9166666666666665,
58
+ "p90": 6.916666666666667,
59
+ "p99": 15.0
60
+ },
61
+ "span_normalization": {
62
+ "note": "kusudama/roll inner span-start chars normalized (engine-ignored per BALLOON-arity evidence, 321/328); RollBig-family inner repeats (6-in-6, 6-in-5) stay fail-closed: unobserved in the corpus, no evidence no semantics",
63
+ "charts_rescued": 317,
64
+ "bars_rescued": 28421,
65
+ "span_structure_fail_before_after": [
66
+ 328,
67
+ 3
68
+ ],
69
+ "off_lattice_fail_before_after": [
70
+ 66,
71
+ 74
72
+ ]
73
+ },
74
+ "source_artifact": "logs/barscript_spec_train_normalized.json (md5 ddebdb9dd689bc8b65fe83c870aa7e80, built --tau 0.30)",
75
+ "rulings": {
76
+ "motif_tau": "0.30: REPEAT+VAR .6640 on the normalized population, conservative side of authored reuse .7073",
77
+ "span_budget_cap": "3+ kept",
78
+ "n_span_dur_buckets": "16 kept",
79
+ "max_tgt": 960,
80
+ "max_tgt_note": "960 covers ura max window 912 content tokens (unchanged by normalization) + BOS/condition prefix headroom; vocab.MAX_TGT update is a separate code change, not part of this artifact"
81
+ },
82
+ "generator_file_md5": {
83
+ "src/softchart/barscript.py": "634e3dc3ddca8bcdceffd760e663e1b0",
84
+ "src/softchart/vocab.py": "fda4d05ddcc7055c71e48e7949fe68c7",
85
+ "scripts/build_barscript_spec.py": "0cba67bd3a1bf52cbdcc0c6221a39051"
86
+ },
87
+ "barscript_md5_update": "2026-07-25: 6c8ac0d2 -> 634e3dc3, fsm-mask-only fix (SPAN_START gated on a non-empty legal-position range; never-dead-end guarantee). Encode/label statistics unaffected: local smoke spec re-run is byte-identical."
88
+ }
89
+ }
vocab.json ADDED
@@ -0,0 +1,2080 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "mapping_sha256": "f139fbea1a13397015088b1ab2c33a952e97b05bb447bc86ce02f350d50dd679",
3
+ "position_token_note": "BarScript skeleton POSITION tokens reuse the TIME_* family as BAR-LOCAL lattice indices on the bar's BAR_DENOM grid; they are not audio frames in that stage. See softchart/vocab.py.",
4
+ "schema_version": "softchart_token_vocabulary_v2",
5
+ "size": 2071,
6
+ "supersedes": "softchart_token_vocabulary_v1 (scripts/to_hf.py), which predates every SoftChart 2.0 token",
7
+ "tokens_by_id": [
8
+ "PAD",
9
+ "EOS",
10
+ "BOS",
11
+ "UNK_COND",
12
+ "COURSE_easy",
13
+ "COURSE_normal",
14
+ "COURSE_hard",
15
+ "COURSE_oni",
16
+ "COURSE_ura",
17
+ "LEVEL_1",
18
+ "LEVEL_2",
19
+ "LEVEL_3",
20
+ "LEVEL_4",
21
+ "LEVEL_5",
22
+ "LEVEL_6",
23
+ "LEVEL_7",
24
+ "LEVEL_8",
25
+ "LEVEL_9",
26
+ "LEVEL_10",
27
+ "LEVEL_11",
28
+ "LEVEL_12",
29
+ "DENS_0",
30
+ "DENS_1",
31
+ "DENS_2",
32
+ "DENS_3",
33
+ "DENS_4",
34
+ "DENS_5",
35
+ "DENS_6",
36
+ "DENS_7",
37
+ "DENS_8",
38
+ "DENS_9",
39
+ "DENS_10",
40
+ "DENS_11",
41
+ "DENS_12",
42
+ "DENS_13",
43
+ "DENS_14",
44
+ "DENS_15",
45
+ "NOTE_don",
46
+ "NOTE_ka",
47
+ "NOTE_don_big",
48
+ "NOTE_ka_big",
49
+ "NOTE_roll",
50
+ "NOTE_roll_big",
51
+ "NOTE_balloon",
52
+ "NOTE_end",
53
+ "TIME_0",
54
+ "TIME_1",
55
+ "TIME_2",
56
+ "TIME_3",
57
+ "TIME_4",
58
+ "TIME_5",
59
+ "TIME_6",
60
+ "TIME_7",
61
+ "TIME_8",
62
+ "TIME_9",
63
+ "TIME_10",
64
+ "TIME_11",
65
+ "TIME_12",
66
+ "TIME_13",
67
+ "TIME_14",
68
+ "TIME_15",
69
+ "TIME_16",
70
+ "TIME_17",
71
+ "TIME_18",
72
+ "TIME_19",
73
+ "TIME_20",
74
+ "TIME_21",
75
+ "TIME_22",
76
+ "TIME_23",
77
+ "TIME_24",
78
+ "TIME_25",
79
+ "TIME_26",
80
+ "TIME_27",
81
+ "TIME_28",
82
+ "TIME_29",
83
+ "TIME_30",
84
+ "TIME_31",
85
+ "TIME_32",
86
+ "TIME_33",
87
+ "TIME_34",
88
+ "TIME_35",
89
+ "TIME_36",
90
+ "TIME_37",
91
+ "TIME_38",
92
+ "TIME_39",
93
+ "TIME_40",
94
+ "TIME_41",
95
+ "TIME_42",
96
+ "TIME_43",
97
+ "TIME_44",
98
+ "TIME_45",
99
+ "TIME_46",
100
+ "TIME_47",
101
+ "TIME_48",
102
+ "TIME_49",
103
+ "TIME_50",
104
+ "TIME_51",
105
+ "TIME_52",
106
+ "TIME_53",
107
+ "TIME_54",
108
+ "TIME_55",
109
+ "TIME_56",
110
+ "TIME_57",
111
+ "TIME_58",
112
+ "TIME_59",
113
+ "TIME_60",
114
+ "TIME_61",
115
+ "TIME_62",
116
+ "TIME_63",
117
+ "TIME_64",
118
+ "TIME_65",
119
+ "TIME_66",
120
+ "TIME_67",
121
+ "TIME_68",
122
+ "TIME_69",
123
+ "TIME_70",
124
+ "TIME_71",
125
+ "TIME_72",
126
+ "TIME_73",
127
+ "TIME_74",
128
+ "TIME_75",
129
+ "TIME_76",
130
+ "TIME_77",
131
+ "TIME_78",
132
+ "TIME_79",
133
+ "TIME_80",
134
+ "TIME_81",
135
+ "TIME_82",
136
+ "TIME_83",
137
+ "TIME_84",
138
+ "TIME_85",
139
+ "TIME_86",
140
+ "TIME_87",
141
+ "TIME_88",
142
+ "TIME_89",
143
+ "TIME_90",
144
+ "TIME_91",
145
+ "TIME_92",
146
+ "TIME_93",
147
+ "TIME_94",
148
+ "TIME_95",
149
+ "TIME_96",
150
+ "TIME_97",
151
+ "TIME_98",
152
+ "TIME_99",
153
+ "TIME_100",
154
+ "TIME_101",
155
+ "TIME_102",
156
+ "TIME_103",
157
+ "TIME_104",
158
+ "TIME_105",
159
+ "TIME_106",
160
+ "TIME_107",
161
+ "TIME_108",
162
+ "TIME_109",
163
+ "TIME_110",
164
+ "TIME_111",
165
+ "TIME_112",
166
+ "TIME_113",
167
+ "TIME_114",
168
+ "TIME_115",
169
+ "TIME_116",
170
+ "TIME_117",
171
+ "TIME_118",
172
+ "TIME_119",
173
+ "TIME_120",
174
+ "TIME_121",
175
+ "TIME_122",
176
+ "TIME_123",
177
+ "TIME_124",
178
+ "TIME_125",
179
+ "TIME_126",
180
+ "TIME_127",
181
+ "TIME_128",
182
+ "TIME_129",
183
+ "TIME_130",
184
+ "TIME_131",
185
+ "TIME_132",
186
+ "TIME_133",
187
+ "TIME_134",
188
+ "TIME_135",
189
+ "TIME_136",
190
+ "TIME_137",
191
+ "TIME_138",
192
+ "TIME_139",
193
+ "TIME_140",
194
+ "TIME_141",
195
+ "TIME_142",
196
+ "TIME_143",
197
+ "TIME_144",
198
+ "TIME_145",
199
+ "TIME_146",
200
+ "TIME_147",
201
+ "TIME_148",
202
+ "TIME_149",
203
+ "TIME_150",
204
+ "TIME_151",
205
+ "TIME_152",
206
+ "TIME_153",
207
+ "TIME_154",
208
+ "TIME_155",
209
+ "TIME_156",
210
+ "TIME_157",
211
+ "TIME_158",
212
+ "TIME_159",
213
+ "TIME_160",
214
+ "TIME_161",
215
+ "TIME_162",
216
+ "TIME_163",
217
+ "TIME_164",
218
+ "TIME_165",
219
+ "TIME_166",
220
+ "TIME_167",
221
+ "TIME_168",
222
+ "TIME_169",
223
+ "TIME_170",
224
+ "TIME_171",
225
+ "TIME_172",
226
+ "TIME_173",
227
+ "TIME_174",
228
+ "TIME_175",
229
+ "TIME_176",
230
+ "TIME_177",
231
+ "TIME_178",
232
+ "TIME_179",
233
+ "TIME_180",
234
+ "TIME_181",
235
+ "TIME_182",
236
+ "TIME_183",
237
+ "TIME_184",
238
+ "TIME_185",
239
+ "TIME_186",
240
+ "TIME_187",
241
+ "TIME_188",
242
+ "TIME_189",
243
+ "TIME_190",
244
+ "TIME_191",
245
+ "TIME_192",
246
+ "TIME_193",
247
+ "TIME_194",
248
+ "TIME_195",
249
+ "TIME_196",
250
+ "TIME_197",
251
+ "TIME_198",
252
+ "TIME_199",
253
+ "TIME_200",
254
+ "TIME_201",
255
+ "TIME_202",
256
+ "TIME_203",
257
+ "TIME_204",
258
+ "TIME_205",
259
+ "TIME_206",
260
+ "TIME_207",
261
+ "TIME_208",
262
+ "TIME_209",
263
+ "TIME_210",
264
+ "TIME_211",
265
+ "TIME_212",
266
+ "TIME_213",
267
+ "TIME_214",
268
+ "TIME_215",
269
+ "TIME_216",
270
+ "TIME_217",
271
+ "TIME_218",
272
+ "TIME_219",
273
+ "TIME_220",
274
+ "TIME_221",
275
+ "TIME_222",
276
+ "TIME_223",
277
+ "TIME_224",
278
+ "TIME_225",
279
+ "TIME_226",
280
+ "TIME_227",
281
+ "TIME_228",
282
+ "TIME_229",
283
+ "TIME_230",
284
+ "TIME_231",
285
+ "TIME_232",
286
+ "TIME_233",
287
+ "TIME_234",
288
+ "TIME_235",
289
+ "TIME_236",
290
+ "TIME_237",
291
+ "TIME_238",
292
+ "TIME_239",
293
+ "TIME_240",
294
+ "TIME_241",
295
+ "TIME_242",
296
+ "TIME_243",
297
+ "TIME_244",
298
+ "TIME_245",
299
+ "TIME_246",
300
+ "TIME_247",
301
+ "TIME_248",
302
+ "TIME_249",
303
+ "TIME_250",
304
+ "TIME_251",
305
+ "TIME_252",
306
+ "TIME_253",
307
+ "TIME_254",
308
+ "TIME_255",
309
+ "TIME_256",
310
+ "TIME_257",
311
+ "TIME_258",
312
+ "TIME_259",
313
+ "TIME_260",
314
+ "TIME_261",
315
+ "TIME_262",
316
+ "TIME_263",
317
+ "TIME_264",
318
+ "TIME_265",
319
+ "TIME_266",
320
+ "TIME_267",
321
+ "TIME_268",
322
+ "TIME_269",
323
+ "TIME_270",
324
+ "TIME_271",
325
+ "TIME_272",
326
+ "TIME_273",
327
+ "TIME_274",
328
+ "TIME_275",
329
+ "TIME_276",
330
+ "TIME_277",
331
+ "TIME_278",
332
+ "TIME_279",
333
+ "TIME_280",
334
+ "TIME_281",
335
+ "TIME_282",
336
+ "TIME_283",
337
+ "TIME_284",
338
+ "TIME_285",
339
+ "TIME_286",
340
+ "TIME_287",
341
+ "TIME_288",
342
+ "TIME_289",
343
+ "TIME_290",
344
+ "TIME_291",
345
+ "TIME_292",
346
+ "TIME_293",
347
+ "TIME_294",
348
+ "TIME_295",
349
+ "TIME_296",
350
+ "TIME_297",
351
+ "TIME_298",
352
+ "TIME_299",
353
+ "TIME_300",
354
+ "TIME_301",
355
+ "TIME_302",
356
+ "TIME_303",
357
+ "TIME_304",
358
+ "TIME_305",
359
+ "TIME_306",
360
+ "TIME_307",
361
+ "TIME_308",
362
+ "TIME_309",
363
+ "TIME_310",
364
+ "TIME_311",
365
+ "TIME_312",
366
+ "TIME_313",
367
+ "TIME_314",
368
+ "TIME_315",
369
+ "TIME_316",
370
+ "TIME_317",
371
+ "TIME_318",
372
+ "TIME_319",
373
+ "TIME_320",
374
+ "TIME_321",
375
+ "TIME_322",
376
+ "TIME_323",
377
+ "TIME_324",
378
+ "TIME_325",
379
+ "TIME_326",
380
+ "TIME_327",
381
+ "TIME_328",
382
+ "TIME_329",
383
+ "TIME_330",
384
+ "TIME_331",
385
+ "TIME_332",
386
+ "TIME_333",
387
+ "TIME_334",
388
+ "TIME_335",
389
+ "TIME_336",
390
+ "TIME_337",
391
+ "TIME_338",
392
+ "TIME_339",
393
+ "TIME_340",
394
+ "TIME_341",
395
+ "TIME_342",
396
+ "TIME_343",
397
+ "TIME_344",
398
+ "TIME_345",
399
+ "TIME_346",
400
+ "TIME_347",
401
+ "TIME_348",
402
+ "TIME_349",
403
+ "TIME_350",
404
+ "TIME_351",
405
+ "TIME_352",
406
+ "TIME_353",
407
+ "TIME_354",
408
+ "TIME_355",
409
+ "TIME_356",
410
+ "TIME_357",
411
+ "TIME_358",
412
+ "TIME_359",
413
+ "TIME_360",
414
+ "TIME_361",
415
+ "TIME_362",
416
+ "TIME_363",
417
+ "TIME_364",
418
+ "TIME_365",
419
+ "TIME_366",
420
+ "TIME_367",
421
+ "TIME_368",
422
+ "TIME_369",
423
+ "TIME_370",
424
+ "TIME_371",
425
+ "TIME_372",
426
+ "TIME_373",
427
+ "TIME_374",
428
+ "TIME_375",
429
+ "TIME_376",
430
+ "TIME_377",
431
+ "TIME_378",
432
+ "TIME_379",
433
+ "TIME_380",
434
+ "TIME_381",
435
+ "TIME_382",
436
+ "TIME_383",
437
+ "TIME_384",
438
+ "TIME_385",
439
+ "TIME_386",
440
+ "TIME_387",
441
+ "TIME_388",
442
+ "TIME_389",
443
+ "TIME_390",
444
+ "TIME_391",
445
+ "TIME_392",
446
+ "TIME_393",
447
+ "TIME_394",
448
+ "TIME_395",
449
+ "TIME_396",
450
+ "TIME_397",
451
+ "TIME_398",
452
+ "TIME_399",
453
+ "TIME_400",
454
+ "TIME_401",
455
+ "TIME_402",
456
+ "TIME_403",
457
+ "TIME_404",
458
+ "TIME_405",
459
+ "TIME_406",
460
+ "TIME_407",
461
+ "TIME_408",
462
+ "TIME_409",
463
+ "TIME_410",
464
+ "TIME_411",
465
+ "TIME_412",
466
+ "TIME_413",
467
+ "TIME_414",
468
+ "TIME_415",
469
+ "TIME_416",
470
+ "TIME_417",
471
+ "TIME_418",
472
+ "TIME_419",
473
+ "TIME_420",
474
+ "TIME_421",
475
+ "TIME_422",
476
+ "TIME_423",
477
+ "TIME_424",
478
+ "TIME_425",
479
+ "TIME_426",
480
+ "TIME_427",
481
+ "TIME_428",
482
+ "TIME_429",
483
+ "TIME_430",
484
+ "TIME_431",
485
+ "TIME_432",
486
+ "TIME_433",
487
+ "TIME_434",
488
+ "TIME_435",
489
+ "TIME_436",
490
+ "TIME_437",
491
+ "TIME_438",
492
+ "TIME_439",
493
+ "TIME_440",
494
+ "TIME_441",
495
+ "TIME_442",
496
+ "TIME_443",
497
+ "TIME_444",
498
+ "TIME_445",
499
+ "TIME_446",
500
+ "TIME_447",
501
+ "TIME_448",
502
+ "TIME_449",
503
+ "TIME_450",
504
+ "TIME_451",
505
+ "TIME_452",
506
+ "TIME_453",
507
+ "TIME_454",
508
+ "TIME_455",
509
+ "TIME_456",
510
+ "TIME_457",
511
+ "TIME_458",
512
+ "TIME_459",
513
+ "TIME_460",
514
+ "TIME_461",
515
+ "TIME_462",
516
+ "TIME_463",
517
+ "TIME_464",
518
+ "TIME_465",
519
+ "TIME_466",
520
+ "TIME_467",
521
+ "TIME_468",
522
+ "TIME_469",
523
+ "TIME_470",
524
+ "TIME_471",
525
+ "TIME_472",
526
+ "TIME_473",
527
+ "TIME_474",
528
+ "TIME_475",
529
+ "TIME_476",
530
+ "TIME_477",
531
+ "TIME_478",
532
+ "TIME_479",
533
+ "TIME_480",
534
+ "TIME_481",
535
+ "TIME_482",
536
+ "TIME_483",
537
+ "TIME_484",
538
+ "TIME_485",
539
+ "TIME_486",
540
+ "TIME_487",
541
+ "TIME_488",
542
+ "TIME_489",
543
+ "TIME_490",
544
+ "TIME_491",
545
+ "TIME_492",
546
+ "TIME_493",
547
+ "TIME_494",
548
+ "TIME_495",
549
+ "TIME_496",
550
+ "TIME_497",
551
+ "TIME_498",
552
+ "TIME_499",
553
+ "TIME_500",
554
+ "TIME_501",
555
+ "TIME_502",
556
+ "TIME_503",
557
+ "TIME_504",
558
+ "TIME_505",
559
+ "TIME_506",
560
+ "TIME_507",
561
+ "TIME_508",
562
+ "TIME_509",
563
+ "TIME_510",
564
+ "TIME_511",
565
+ "TIME_512",
566
+ "TIME_513",
567
+ "TIME_514",
568
+ "TIME_515",
569
+ "TIME_516",
570
+ "TIME_517",
571
+ "TIME_518",
572
+ "TIME_519",
573
+ "TIME_520",
574
+ "TIME_521",
575
+ "TIME_522",
576
+ "TIME_523",
577
+ "TIME_524",
578
+ "TIME_525",
579
+ "TIME_526",
580
+ "TIME_527",
581
+ "TIME_528",
582
+ "TIME_529",
583
+ "TIME_530",
584
+ "TIME_531",
585
+ "TIME_532",
586
+ "TIME_533",
587
+ "TIME_534",
588
+ "TIME_535",
589
+ "TIME_536",
590
+ "TIME_537",
591
+ "TIME_538",
592
+ "TIME_539",
593
+ "TIME_540",
594
+ "TIME_541",
595
+ "TIME_542",
596
+ "TIME_543",
597
+ "TIME_544",
598
+ "TIME_545",
599
+ "TIME_546",
600
+ "TIME_547",
601
+ "TIME_548",
602
+ "TIME_549",
603
+ "TIME_550",
604
+ "TIME_551",
605
+ "TIME_552",
606
+ "TIME_553",
607
+ "TIME_554",
608
+ "TIME_555",
609
+ "TIME_556",
610
+ "TIME_557",
611
+ "TIME_558",
612
+ "TIME_559",
613
+ "TIME_560",
614
+ "TIME_561",
615
+ "TIME_562",
616
+ "TIME_563",
617
+ "TIME_564",
618
+ "TIME_565",
619
+ "TIME_566",
620
+ "TIME_567",
621
+ "TIME_568",
622
+ "TIME_569",
623
+ "TIME_570",
624
+ "TIME_571",
625
+ "TIME_572",
626
+ "TIME_573",
627
+ "TIME_574",
628
+ "TIME_575",
629
+ "TIME_576",
630
+ "TIME_577",
631
+ "TIME_578",
632
+ "TIME_579",
633
+ "TIME_580",
634
+ "TIME_581",
635
+ "TIME_582",
636
+ "TIME_583",
637
+ "TIME_584",
638
+ "TIME_585",
639
+ "TIME_586",
640
+ "TIME_587",
641
+ "TIME_588",
642
+ "TIME_589",
643
+ "TIME_590",
644
+ "TIME_591",
645
+ "TIME_592",
646
+ "TIME_593",
647
+ "TIME_594",
648
+ "TIME_595",
649
+ "TIME_596",
650
+ "TIME_597",
651
+ "TIME_598",
652
+ "TIME_599",
653
+ "TIME_600",
654
+ "TIME_601",
655
+ "TIME_602",
656
+ "TIME_603",
657
+ "TIME_604",
658
+ "TIME_605",
659
+ "TIME_606",
660
+ "TIME_607",
661
+ "TIME_608",
662
+ "TIME_609",
663
+ "TIME_610",
664
+ "TIME_611",
665
+ "TIME_612",
666
+ "TIME_613",
667
+ "TIME_614",
668
+ "TIME_615",
669
+ "TIME_616",
670
+ "TIME_617",
671
+ "TIME_618",
672
+ "TIME_619",
673
+ "TIME_620",
674
+ "TIME_621",
675
+ "TIME_622",
676
+ "TIME_623",
677
+ "TIME_624",
678
+ "TIME_625",
679
+ "TIME_626",
680
+ "TIME_627",
681
+ "TIME_628",
682
+ "TIME_629",
683
+ "TIME_630",
684
+ "TIME_631",
685
+ "TIME_632",
686
+ "TIME_633",
687
+ "TIME_634",
688
+ "TIME_635",
689
+ "TIME_636",
690
+ "TIME_637",
691
+ "TIME_638",
692
+ "TIME_639",
693
+ "TIME_640",
694
+ "TIME_641",
695
+ "TIME_642",
696
+ "TIME_643",
697
+ "TIME_644",
698
+ "TIME_645",
699
+ "TIME_646",
700
+ "TIME_647",
701
+ "TIME_648",
702
+ "TIME_649",
703
+ "TIME_650",
704
+ "TIME_651",
705
+ "TIME_652",
706
+ "TIME_653",
707
+ "TIME_654",
708
+ "TIME_655",
709
+ "TIME_656",
710
+ "TIME_657",
711
+ "TIME_658",
712
+ "TIME_659",
713
+ "TIME_660",
714
+ "TIME_661",
715
+ "TIME_662",
716
+ "TIME_663",
717
+ "TIME_664",
718
+ "TIME_665",
719
+ "TIME_666",
720
+ "TIME_667",
721
+ "TIME_668",
722
+ "TIME_669",
723
+ "TIME_670",
724
+ "TIME_671",
725
+ "TIME_672",
726
+ "TIME_673",
727
+ "TIME_674",
728
+ "TIME_675",
729
+ "TIME_676",
730
+ "TIME_677",
731
+ "TIME_678",
732
+ "TIME_679",
733
+ "TIME_680",
734
+ "TIME_681",
735
+ "TIME_682",
736
+ "TIME_683",
737
+ "TIME_684",
738
+ "TIME_685",
739
+ "TIME_686",
740
+ "TIME_687",
741
+ "TIME_688",
742
+ "TIME_689",
743
+ "TIME_690",
744
+ "TIME_691",
745
+ "TIME_692",
746
+ "TIME_693",
747
+ "TIME_694",
748
+ "TIME_695",
749
+ "TIME_696",
750
+ "TIME_697",
751
+ "TIME_698",
752
+ "TIME_699",
753
+ "TIME_700",
754
+ "TIME_701",
755
+ "TIME_702",
756
+ "TIME_703",
757
+ "TIME_704",
758
+ "TIME_705",
759
+ "TIME_706",
760
+ "TIME_707",
761
+ "TIME_708",
762
+ "TIME_709",
763
+ "TIME_710",
764
+ "TIME_711",
765
+ "TIME_712",
766
+ "TIME_713",
767
+ "TIME_714",
768
+ "TIME_715",
769
+ "TIME_716",
770
+ "TIME_717",
771
+ "TIME_718",
772
+ "TIME_719",
773
+ "TIME_720",
774
+ "TIME_721",
775
+ "TIME_722",
776
+ "TIME_723",
777
+ "TIME_724",
778
+ "TIME_725",
779
+ "TIME_726",
780
+ "TIME_727",
781
+ "TIME_728",
782
+ "TIME_729",
783
+ "TIME_730",
784
+ "TIME_731",
785
+ "TIME_732",
786
+ "TIME_733",
787
+ "TIME_734",
788
+ "TIME_735",
789
+ "TIME_736",
790
+ "TIME_737",
791
+ "TIME_738",
792
+ "TIME_739",
793
+ "TIME_740",
794
+ "TIME_741",
795
+ "TIME_742",
796
+ "TIME_743",
797
+ "TIME_744",
798
+ "TIME_745",
799
+ "TIME_746",
800
+ "TIME_747",
801
+ "TIME_748",
802
+ "TIME_749",
803
+ "TIME_750",
804
+ "TIME_751",
805
+ "TIME_752",
806
+ "TIME_753",
807
+ "TIME_754",
808
+ "TIME_755",
809
+ "TIME_756",
810
+ "TIME_757",
811
+ "TIME_758",
812
+ "TIME_759",
813
+ "TIME_760",
814
+ "TIME_761",
815
+ "TIME_762",
816
+ "TIME_763",
817
+ "TIME_764",
818
+ "TIME_765",
819
+ "TIME_766",
820
+ "TIME_767",
821
+ "TIME_768",
822
+ "TIME_769",
823
+ "TIME_770",
824
+ "TIME_771",
825
+ "TIME_772",
826
+ "TIME_773",
827
+ "TIME_774",
828
+ "TIME_775",
829
+ "TIME_776",
830
+ "TIME_777",
831
+ "TIME_778",
832
+ "TIME_779",
833
+ "TIME_780",
834
+ "TIME_781",
835
+ "TIME_782",
836
+ "TIME_783",
837
+ "TIME_784",
838
+ "TIME_785",
839
+ "TIME_786",
840
+ "TIME_787",
841
+ "TIME_788",
842
+ "TIME_789",
843
+ "TIME_790",
844
+ "TIME_791",
845
+ "TIME_792",
846
+ "TIME_793",
847
+ "TIME_794",
848
+ "TIME_795",
849
+ "TIME_796",
850
+ "TIME_797",
851
+ "TIME_798",
852
+ "TIME_799",
853
+ "TIME_800",
854
+ "TIME_801",
855
+ "TIME_802",
856
+ "TIME_803",
857
+ "TIME_804",
858
+ "TIME_805",
859
+ "TIME_806",
860
+ "TIME_807",
861
+ "TIME_808",
862
+ "TIME_809",
863
+ "TIME_810",
864
+ "TIME_811",
865
+ "TIME_812",
866
+ "TIME_813",
867
+ "TIME_814",
868
+ "TIME_815",
869
+ "TIME_816",
870
+ "TIME_817",
871
+ "TIME_818",
872
+ "TIME_819",
873
+ "TIME_820",
874
+ "TIME_821",
875
+ "TIME_822",
876
+ "TIME_823",
877
+ "TIME_824",
878
+ "TIME_825",
879
+ "TIME_826",
880
+ "TIME_827",
881
+ "TIME_828",
882
+ "TIME_829",
883
+ "TIME_830",
884
+ "TIME_831",
885
+ "TIME_832",
886
+ "TIME_833",
887
+ "TIME_834",
888
+ "TIME_835",
889
+ "TIME_836",
890
+ "TIME_837",
891
+ "TIME_838",
892
+ "TIME_839",
893
+ "TIME_840",
894
+ "TIME_841",
895
+ "TIME_842",
896
+ "TIME_843",
897
+ "TIME_844",
898
+ "TIME_845",
899
+ "TIME_846",
900
+ "TIME_847",
901
+ "TIME_848",
902
+ "TIME_849",
903
+ "TIME_850",
904
+ "TIME_851",
905
+ "TIME_852",
906
+ "TIME_853",
907
+ "TIME_854",
908
+ "TIME_855",
909
+ "TIME_856",
910
+ "TIME_857",
911
+ "TIME_858",
912
+ "TIME_859",
913
+ "TIME_860",
914
+ "TIME_861",
915
+ "TIME_862",
916
+ "TIME_863",
917
+ "TIME_864",
918
+ "TIME_865",
919
+ "TIME_866",
920
+ "TIME_867",
921
+ "TIME_868",
922
+ "TIME_869",
923
+ "TIME_870",
924
+ "TIME_871",
925
+ "TIME_872",
926
+ "TIME_873",
927
+ "TIME_874",
928
+ "TIME_875",
929
+ "TIME_876",
930
+ "TIME_877",
931
+ "TIME_878",
932
+ "TIME_879",
933
+ "TIME_880",
934
+ "TIME_881",
935
+ "TIME_882",
936
+ "TIME_883",
937
+ "TIME_884",
938
+ "TIME_885",
939
+ "TIME_886",
940
+ "TIME_887",
941
+ "TIME_888",
942
+ "TIME_889",
943
+ "TIME_890",
944
+ "TIME_891",
945
+ "TIME_892",
946
+ "TIME_893",
947
+ "TIME_894",
948
+ "TIME_895",
949
+ "TIME_896",
950
+ "TIME_897",
951
+ "TIME_898",
952
+ "TIME_899",
953
+ "TIME_900",
954
+ "TIME_901",
955
+ "TIME_902",
956
+ "TIME_903",
957
+ "TIME_904",
958
+ "TIME_905",
959
+ "TIME_906",
960
+ "TIME_907",
961
+ "TIME_908",
962
+ "TIME_909",
963
+ "TIME_910",
964
+ "TIME_911",
965
+ "TIME_912",
966
+ "TIME_913",
967
+ "TIME_914",
968
+ "TIME_915",
969
+ "TIME_916",
970
+ "TIME_917",
971
+ "TIME_918",
972
+ "TIME_919",
973
+ "TIME_920",
974
+ "TIME_921",
975
+ "TIME_922",
976
+ "TIME_923",
977
+ "TIME_924",
978
+ "TIME_925",
979
+ "TIME_926",
980
+ "TIME_927",
981
+ "TIME_928",
982
+ "TIME_929",
983
+ "TIME_930",
984
+ "TIME_931",
985
+ "TIME_932",
986
+ "TIME_933",
987
+ "TIME_934",
988
+ "TIME_935",
989
+ "TIME_936",
990
+ "TIME_937",
991
+ "TIME_938",
992
+ "TIME_939",
993
+ "TIME_940",
994
+ "TIME_941",
995
+ "TIME_942",
996
+ "TIME_943",
997
+ "TIME_944",
998
+ "TIME_945",
999
+ "TIME_946",
1000
+ "TIME_947",
1001
+ "TIME_948",
1002
+ "TIME_949",
1003
+ "TIME_950",
1004
+ "TIME_951",
1005
+ "TIME_952",
1006
+ "TIME_953",
1007
+ "TIME_954",
1008
+ "TIME_955",
1009
+ "TIME_956",
1010
+ "TIME_957",
1011
+ "TIME_958",
1012
+ "TIME_959",
1013
+ "TIME_960",
1014
+ "TIME_961",
1015
+ "TIME_962",
1016
+ "TIME_963",
1017
+ "TIME_964",
1018
+ "TIME_965",
1019
+ "TIME_966",
1020
+ "TIME_967",
1021
+ "TIME_968",
1022
+ "TIME_969",
1023
+ "TIME_970",
1024
+ "TIME_971",
1025
+ "TIME_972",
1026
+ "TIME_973",
1027
+ "TIME_974",
1028
+ "TIME_975",
1029
+ "TIME_976",
1030
+ "TIME_977",
1031
+ "TIME_978",
1032
+ "TIME_979",
1033
+ "TIME_980",
1034
+ "TIME_981",
1035
+ "TIME_982",
1036
+ "TIME_983",
1037
+ "TIME_984",
1038
+ "TIME_985",
1039
+ "TIME_986",
1040
+ "TIME_987",
1041
+ "TIME_988",
1042
+ "TIME_989",
1043
+ "TIME_990",
1044
+ "TIME_991",
1045
+ "TIME_992",
1046
+ "TIME_993",
1047
+ "TIME_994",
1048
+ "TIME_995",
1049
+ "TIME_996",
1050
+ "TIME_997",
1051
+ "TIME_998",
1052
+ "TIME_999",
1053
+ "TIME_1000",
1054
+ "TIME_1001",
1055
+ "TIME_1002",
1056
+ "TIME_1003",
1057
+ "TIME_1004",
1058
+ "TIME_1005",
1059
+ "TIME_1006",
1060
+ "TIME_1007",
1061
+ "TIME_1008",
1062
+ "TIME_1009",
1063
+ "TIME_1010",
1064
+ "TIME_1011",
1065
+ "TIME_1012",
1066
+ "TIME_1013",
1067
+ "TIME_1014",
1068
+ "TIME_1015",
1069
+ "TIME_1016",
1070
+ "TIME_1017",
1071
+ "TIME_1018",
1072
+ "TIME_1019",
1073
+ "TIME_1020",
1074
+ "TIME_1021",
1075
+ "TIME_1022",
1076
+ "TIME_1023",
1077
+ "TIME_1024",
1078
+ "TIME_1025",
1079
+ "TIME_1026",
1080
+ "TIME_1027",
1081
+ "TIME_1028",
1082
+ "TIME_1029",
1083
+ "TIME_1030",
1084
+ "TIME_1031",
1085
+ "TIME_1032",
1086
+ "TIME_1033",
1087
+ "TIME_1034",
1088
+ "TIME_1035",
1089
+ "TIME_1036",
1090
+ "TIME_1037",
1091
+ "TIME_1038",
1092
+ "TIME_1039",
1093
+ "TIME_1040",
1094
+ "TIME_1041",
1095
+ "TIME_1042",
1096
+ "TIME_1043",
1097
+ "TIME_1044",
1098
+ "TIME_1045",
1099
+ "TIME_1046",
1100
+ "TIME_1047",
1101
+ "TIME_1048",
1102
+ "TIME_1049",
1103
+ "TIME_1050",
1104
+ "TIME_1051",
1105
+ "TIME_1052",
1106
+ "TIME_1053",
1107
+ "TIME_1054",
1108
+ "TIME_1055",
1109
+ "TIME_1056",
1110
+ "TIME_1057",
1111
+ "TIME_1058",
1112
+ "TIME_1059",
1113
+ "TIME_1060",
1114
+ "TIME_1061",
1115
+ "TIME_1062",
1116
+ "TIME_1063",
1117
+ "TIME_1064",
1118
+ "TIME_1065",
1119
+ "TIME_1066",
1120
+ "TIME_1067",
1121
+ "TIME_1068",
1122
+ "TIME_1069",
1123
+ "TIME_1070",
1124
+ "TIME_1071",
1125
+ "TIME_1072",
1126
+ "TIME_1073",
1127
+ "TIME_1074",
1128
+ "TIME_1075",
1129
+ "TIME_1076",
1130
+ "TIME_1077",
1131
+ "TIME_1078",
1132
+ "TIME_1079",
1133
+ "TIME_1080",
1134
+ "TIME_1081",
1135
+ "TIME_1082",
1136
+ "TIME_1083",
1137
+ "TIME_1084",
1138
+ "TIME_1085",
1139
+ "TIME_1086",
1140
+ "TIME_1087",
1141
+ "TIME_1088",
1142
+ "TIME_1089",
1143
+ "TIME_1090",
1144
+ "TIME_1091",
1145
+ "TIME_1092",
1146
+ "TIME_1093",
1147
+ "TIME_1094",
1148
+ "TIME_1095",
1149
+ "TIME_1096",
1150
+ "TIME_1097",
1151
+ "TIME_1098",
1152
+ "TIME_1099",
1153
+ "TIME_1100",
1154
+ "TIME_1101",
1155
+ "TIME_1102",
1156
+ "TIME_1103",
1157
+ "TIME_1104",
1158
+ "TIME_1105",
1159
+ "TIME_1106",
1160
+ "TIME_1107",
1161
+ "TIME_1108",
1162
+ "TIME_1109",
1163
+ "TIME_1110",
1164
+ "TIME_1111",
1165
+ "TIME_1112",
1166
+ "TIME_1113",
1167
+ "TIME_1114",
1168
+ "TIME_1115",
1169
+ "TIME_1116",
1170
+ "TIME_1117",
1171
+ "TIME_1118",
1172
+ "TIME_1119",
1173
+ "TIME_1120",
1174
+ "TIME_1121",
1175
+ "TIME_1122",
1176
+ "TIME_1123",
1177
+ "TIME_1124",
1178
+ "TIME_1125",
1179
+ "TIME_1126",
1180
+ "TIME_1127",
1181
+ "TIME_1128",
1182
+ "TIME_1129",
1183
+ "TIME_1130",
1184
+ "TIME_1131",
1185
+ "TIME_1132",
1186
+ "TIME_1133",
1187
+ "TIME_1134",
1188
+ "TIME_1135",
1189
+ "TIME_1136",
1190
+ "TIME_1137",
1191
+ "TIME_1138",
1192
+ "TIME_1139",
1193
+ "TIME_1140",
1194
+ "TIME_1141",
1195
+ "TIME_1142",
1196
+ "TIME_1143",
1197
+ "TIME_1144",
1198
+ "TIME_1145",
1199
+ "TIME_1146",
1200
+ "TIME_1147",
1201
+ "TIME_1148",
1202
+ "TIME_1149",
1203
+ "TIME_1150",
1204
+ "TIME_1151",
1205
+ "TIME_1152",
1206
+ "TIME_1153",
1207
+ "TIME_1154",
1208
+ "TIME_1155",
1209
+ "TIME_1156",
1210
+ "TIME_1157",
1211
+ "TIME_1158",
1212
+ "TIME_1159",
1213
+ "TIME_1160",
1214
+ "TIME_1161",
1215
+ "TIME_1162",
1216
+ "TIME_1163",
1217
+ "TIME_1164",
1218
+ "TIME_1165",
1219
+ "TIME_1166",
1220
+ "TIME_1167",
1221
+ "TIME_1168",
1222
+ "TIME_1169",
1223
+ "TIME_1170",
1224
+ "TIME_1171",
1225
+ "TIME_1172",
1226
+ "TIME_1173",
1227
+ "TIME_1174",
1228
+ "TIME_1175",
1229
+ "TIME_1176",
1230
+ "TIME_1177",
1231
+ "TIME_1178",
1232
+ "TIME_1179",
1233
+ "TIME_1180",
1234
+ "TIME_1181",
1235
+ "TIME_1182",
1236
+ "TIME_1183",
1237
+ "TIME_1184",
1238
+ "TIME_1185",
1239
+ "TIME_1186",
1240
+ "TIME_1187",
1241
+ "TIME_1188",
1242
+ "TIME_1189",
1243
+ "TIME_1190",
1244
+ "TIME_1191",
1245
+ "TIME_1192",
1246
+ "TIME_1193",
1247
+ "TIME_1194",
1248
+ "TIME_1195",
1249
+ "TIME_1196",
1250
+ "TIME_1197",
1251
+ "TIME_1198",
1252
+ "TIME_1199",
1253
+ "TIME_1200",
1254
+ "TIME_1201",
1255
+ "TIME_1202",
1256
+ "TIME_1203",
1257
+ "TIME_1204",
1258
+ "TIME_1205",
1259
+ "TIME_1206",
1260
+ "TIME_1207",
1261
+ "TIME_1208",
1262
+ "TIME_1209",
1263
+ "TIME_1210",
1264
+ "TIME_1211",
1265
+ "TIME_1212",
1266
+ "TIME_1213",
1267
+ "TIME_1214",
1268
+ "TIME_1215",
1269
+ "TIME_1216",
1270
+ "TIME_1217",
1271
+ "TIME_1218",
1272
+ "TIME_1219",
1273
+ "TIME_1220",
1274
+ "TIME_1221",
1275
+ "TIME_1222",
1276
+ "TIME_1223",
1277
+ "TIME_1224",
1278
+ "TIME_1225",
1279
+ "TIME_1226",
1280
+ "TIME_1227",
1281
+ "TIME_1228",
1282
+ "TIME_1229",
1283
+ "TIME_1230",
1284
+ "TIME_1231",
1285
+ "TIME_1232",
1286
+ "TIME_1233",
1287
+ "TIME_1234",
1288
+ "TIME_1235",
1289
+ "TIME_1236",
1290
+ "TIME_1237",
1291
+ "TIME_1238",
1292
+ "TIME_1239",
1293
+ "TIME_1240",
1294
+ "TIME_1241",
1295
+ "TIME_1242",
1296
+ "TIME_1243",
1297
+ "TIME_1244",
1298
+ "TIME_1245",
1299
+ "TIME_1246",
1300
+ "TIME_1247",
1301
+ "TIME_1248",
1302
+ "TIME_1249",
1303
+ "TIME_1250",
1304
+ "TIME_1251",
1305
+ "TIME_1252",
1306
+ "TIME_1253",
1307
+ "TIME_1254",
1308
+ "TIME_1255",
1309
+ "TIME_1256",
1310
+ "TIME_1257",
1311
+ "TIME_1258",
1312
+ "TIME_1259",
1313
+ "TIME_1260",
1314
+ "TIME_1261",
1315
+ "TIME_1262",
1316
+ "TIME_1263",
1317
+ "TIME_1264",
1318
+ "TIME_1265",
1319
+ "TIME_1266",
1320
+ "TIME_1267",
1321
+ "TIME_1268",
1322
+ "TIME_1269",
1323
+ "TIME_1270",
1324
+ "TIME_1271",
1325
+ "TIME_1272",
1326
+ "TIME_1273",
1327
+ "TIME_1274",
1328
+ "TIME_1275",
1329
+ "TIME_1276",
1330
+ "TIME_1277",
1331
+ "TIME_1278",
1332
+ "TIME_1279",
1333
+ "TIME_1280",
1334
+ "TIME_1281",
1335
+ "TIME_1282",
1336
+ "TIME_1283",
1337
+ "TIME_1284",
1338
+ "TIME_1285",
1339
+ "TIME_1286",
1340
+ "TIME_1287",
1341
+ "TIME_1288",
1342
+ "TIME_1289",
1343
+ "TIME_1290",
1344
+ "TIME_1291",
1345
+ "TIME_1292",
1346
+ "TIME_1293",
1347
+ "TIME_1294",
1348
+ "TIME_1295",
1349
+ "TIME_1296",
1350
+ "TIME_1297",
1351
+ "TIME_1298",
1352
+ "TIME_1299",
1353
+ "TIME_1300",
1354
+ "TIME_1301",
1355
+ "TIME_1302",
1356
+ "TIME_1303",
1357
+ "TIME_1304",
1358
+ "TIME_1305",
1359
+ "TIME_1306",
1360
+ "TIME_1307",
1361
+ "TIME_1308",
1362
+ "TIME_1309",
1363
+ "TIME_1310",
1364
+ "TIME_1311",
1365
+ "TIME_1312",
1366
+ "TIME_1313",
1367
+ "TIME_1314",
1368
+ "TIME_1315",
1369
+ "TIME_1316",
1370
+ "TIME_1317",
1371
+ "TIME_1318",
1372
+ "TIME_1319",
1373
+ "TIME_1320",
1374
+ "TIME_1321",
1375
+ "TIME_1322",
1376
+ "TIME_1323",
1377
+ "TIME_1324",
1378
+ "TIME_1325",
1379
+ "TIME_1326",
1380
+ "TIME_1327",
1381
+ "TIME_1328",
1382
+ "TIME_1329",
1383
+ "TIME_1330",
1384
+ "TIME_1331",
1385
+ "TIME_1332",
1386
+ "TIME_1333",
1387
+ "TIME_1334",
1388
+ "TIME_1335",
1389
+ "TIME_1336",
1390
+ "TIME_1337",
1391
+ "TIME_1338",
1392
+ "TIME_1339",
1393
+ "TIME_1340",
1394
+ "TIME_1341",
1395
+ "TIME_1342",
1396
+ "TIME_1343",
1397
+ "TIME_1344",
1398
+ "TIME_1345",
1399
+ "TIME_1346",
1400
+ "TIME_1347",
1401
+ "TIME_1348",
1402
+ "TIME_1349",
1403
+ "TIME_1350",
1404
+ "TIME_1351",
1405
+ "TIME_1352",
1406
+ "TIME_1353",
1407
+ "TIME_1354",
1408
+ "TIME_1355",
1409
+ "TIME_1356",
1410
+ "TIME_1357",
1411
+ "TIME_1358",
1412
+ "TIME_1359",
1413
+ "TIME_1360",
1414
+ "TIME_1361",
1415
+ "TIME_1362",
1416
+ "TIME_1363",
1417
+ "TIME_1364",
1418
+ "TIME_1365",
1419
+ "TIME_1366",
1420
+ "TIME_1367",
1421
+ "TIME_1368",
1422
+ "TIME_1369",
1423
+ "TIME_1370",
1424
+ "TIME_1371",
1425
+ "TIME_1372",
1426
+ "TIME_1373",
1427
+ "TIME_1374",
1428
+ "TIME_1375",
1429
+ "TIME_1376",
1430
+ "TIME_1377",
1431
+ "TIME_1378",
1432
+ "TIME_1379",
1433
+ "TIME_1380",
1434
+ "TIME_1381",
1435
+ "TIME_1382",
1436
+ "TIME_1383",
1437
+ "TIME_1384",
1438
+ "TIME_1385",
1439
+ "TIME_1386",
1440
+ "TIME_1387",
1441
+ "TIME_1388",
1442
+ "TIME_1389",
1443
+ "TIME_1390",
1444
+ "TIME_1391",
1445
+ "TIME_1392",
1446
+ "TIME_1393",
1447
+ "TIME_1394",
1448
+ "TIME_1395",
1449
+ "TIME_1396",
1450
+ "TIME_1397",
1451
+ "TIME_1398",
1452
+ "TIME_1399",
1453
+ "TIME_1400",
1454
+ "TIME_1401",
1455
+ "TIME_1402",
1456
+ "TIME_1403",
1457
+ "TIME_1404",
1458
+ "TIME_1405",
1459
+ "TIME_1406",
1460
+ "TIME_1407",
1461
+ "TIME_1408",
1462
+ "TIME_1409",
1463
+ "TIME_1410",
1464
+ "TIME_1411",
1465
+ "TIME_1412",
1466
+ "TIME_1413",
1467
+ "TIME_1414",
1468
+ "TIME_1415",
1469
+ "TIME_1416",
1470
+ "TIME_1417",
1471
+ "TIME_1418",
1472
+ "TIME_1419",
1473
+ "TIME_1420",
1474
+ "TIME_1421",
1475
+ "TIME_1422",
1476
+ "TIME_1423",
1477
+ "TIME_1424",
1478
+ "TIME_1425",
1479
+ "TIME_1426",
1480
+ "TIME_1427",
1481
+ "TIME_1428",
1482
+ "TIME_1429",
1483
+ "TIME_1430",
1484
+ "TIME_1431",
1485
+ "TIME_1432",
1486
+ "TIME_1433",
1487
+ "TIME_1434",
1488
+ "TIME_1435",
1489
+ "TIME_1436",
1490
+ "TIME_1437",
1491
+ "TIME_1438",
1492
+ "TIME_1439",
1493
+ "TIME_1440",
1494
+ "TIME_1441",
1495
+ "TIME_1442",
1496
+ "TIME_1443",
1497
+ "TIME_1444",
1498
+ "TIME_1445",
1499
+ "TIME_1446",
1500
+ "TIME_1447",
1501
+ "TIME_1448",
1502
+ "TIME_1449",
1503
+ "TIME_1450",
1504
+ "TIME_1451",
1505
+ "TIME_1452",
1506
+ "TIME_1453",
1507
+ "TIME_1454",
1508
+ "TIME_1455",
1509
+ "TIME_1456",
1510
+ "TIME_1457",
1511
+ "TIME_1458",
1512
+ "TIME_1459",
1513
+ "TIME_1460",
1514
+ "TIME_1461",
1515
+ "TIME_1462",
1516
+ "TIME_1463",
1517
+ "TIME_1464",
1518
+ "TIME_1465",
1519
+ "TIME_1466",
1520
+ "TIME_1467",
1521
+ "TIME_1468",
1522
+ "TIME_1469",
1523
+ "TIME_1470",
1524
+ "TIME_1471",
1525
+ "TIME_1472",
1526
+ "TIME_1473",
1527
+ "TIME_1474",
1528
+ "TIME_1475",
1529
+ "TIME_1476",
1530
+ "TIME_1477",
1531
+ "TIME_1478",
1532
+ "TIME_1479",
1533
+ "TIME_1480",
1534
+ "TIME_1481",
1535
+ "TIME_1482",
1536
+ "TIME_1483",
1537
+ "TIME_1484",
1538
+ "TIME_1485",
1539
+ "TIME_1486",
1540
+ "TIME_1487",
1541
+ "TIME_1488",
1542
+ "TIME_1489",
1543
+ "TIME_1490",
1544
+ "TIME_1491",
1545
+ "TIME_1492",
1546
+ "TIME_1493",
1547
+ "TIME_1494",
1548
+ "TIME_1495",
1549
+ "TIME_1496",
1550
+ "TIME_1497",
1551
+ "TIME_1498",
1552
+ "TIME_1499",
1553
+ "TIME_1500",
1554
+ "TIME_1501",
1555
+ "TIME_1502",
1556
+ "TIME_1503",
1557
+ "TIME_1504",
1558
+ "TIME_1505",
1559
+ "TIME_1506",
1560
+ "TIME_1507",
1561
+ "TIME_1508",
1562
+ "TIME_1509",
1563
+ "TIME_1510",
1564
+ "TIME_1511",
1565
+ "TIME_1512",
1566
+ "TIME_1513",
1567
+ "TIME_1514",
1568
+ "TIME_1515",
1569
+ "TIME_1516",
1570
+ "TIME_1517",
1571
+ "TIME_1518",
1572
+ "TIME_1519",
1573
+ "TIME_1520",
1574
+ "TIME_1521",
1575
+ "TIME_1522",
1576
+ "TIME_1523",
1577
+ "TIME_1524",
1578
+ "TIME_1525",
1579
+ "TIME_1526",
1580
+ "TIME_1527",
1581
+ "TIME_1528",
1582
+ "TIME_1529",
1583
+ "TIME_1530",
1584
+ "TIME_1531",
1585
+ "TIME_1532",
1586
+ "TIME_1533",
1587
+ "TIME_1534",
1588
+ "TIME_1535",
1589
+ "TIME_1536",
1590
+ "TIME_1537",
1591
+ "TIME_1538",
1592
+ "TIME_1539",
1593
+ "TIME_1540",
1594
+ "TIME_1541",
1595
+ "TIME_1542",
1596
+ "TIME_1543",
1597
+ "TIME_1544",
1598
+ "TIME_1545",
1599
+ "TIME_1546",
1600
+ "TIME_1547",
1601
+ "TIME_1548",
1602
+ "TIME_1549",
1603
+ "TIME_1550",
1604
+ "TIME_1551",
1605
+ "TIME_1552",
1606
+ "TIME_1553",
1607
+ "TIME_1554",
1608
+ "TIME_1555",
1609
+ "TIME_1556",
1610
+ "TIME_1557",
1611
+ "TIME_1558",
1612
+ "TIME_1559",
1613
+ "TIME_1560",
1614
+ "TIME_1561",
1615
+ "TIME_1562",
1616
+ "TIME_1563",
1617
+ "TIME_1564",
1618
+ "TIME_1565",
1619
+ "TIME_1566",
1620
+ "TIME_1567",
1621
+ "TIME_1568",
1622
+ "TIME_1569",
1623
+ "TIME_1570",
1624
+ "TIME_1571",
1625
+ "TIME_1572",
1626
+ "TIME_1573",
1627
+ "TIME_1574",
1628
+ "TIME_1575",
1629
+ "TIME_1576",
1630
+ "TIME_1577",
1631
+ "TIME_1578",
1632
+ "TIME_1579",
1633
+ "TIME_1580",
1634
+ "TIME_1581",
1635
+ "TIME_1582",
1636
+ "TIME_1583",
1637
+ "TIME_1584",
1638
+ "TIME_1585",
1639
+ "TIME_1586",
1640
+ "TIME_1587",
1641
+ "TIME_1588",
1642
+ "TIME_1589",
1643
+ "TIME_1590",
1644
+ "TIME_1591",
1645
+ "TIME_1592",
1646
+ "TIME_1593",
1647
+ "TIME_1594",
1648
+ "TIME_1595",
1649
+ "TIME_1596",
1650
+ "TIME_1597",
1651
+ "TIME_1598",
1652
+ "TIME_1599",
1653
+ "TIME_1600",
1654
+ "TIME_1601",
1655
+ "TIME_1602",
1656
+ "TIME_1603",
1657
+ "TIME_1604",
1658
+ "TIME_1605",
1659
+ "TIME_1606",
1660
+ "TIME_1607",
1661
+ "TIME_1608",
1662
+ "TIME_1609",
1663
+ "TIME_1610",
1664
+ "TIME_1611",
1665
+ "TIME_1612",
1666
+ "TIME_1613",
1667
+ "TIME_1614",
1668
+ "TIME_1615",
1669
+ "TIME_1616",
1670
+ "TIME_1617",
1671
+ "TIME_1618",
1672
+ "TIME_1619",
1673
+ "TIME_1620",
1674
+ "TIME_1621",
1675
+ "TIME_1622",
1676
+ "TIME_1623",
1677
+ "TIME_1624",
1678
+ "TIME_1625",
1679
+ "TIME_1626",
1680
+ "TIME_1627",
1681
+ "TIME_1628",
1682
+ "TIME_1629",
1683
+ "TIME_1630",
1684
+ "TIME_1631",
1685
+ "TIME_1632",
1686
+ "TIME_1633",
1687
+ "TIME_1634",
1688
+ "TIME_1635",
1689
+ "TIME_1636",
1690
+ "TIME_1637",
1691
+ "TIME_1638",
1692
+ "TIME_1639",
1693
+ "TIME_1640",
1694
+ "TIME_1641",
1695
+ "TIME_1642",
1696
+ "TIME_1643",
1697
+ "TIME_1644",
1698
+ "TIME_1645",
1699
+ "TIME_1646",
1700
+ "TIME_1647",
1701
+ "TIME_1648",
1702
+ "TIME_1649",
1703
+ "TIME_1650",
1704
+ "TIME_1651",
1705
+ "TIME_1652",
1706
+ "TIME_1653",
1707
+ "TIME_1654",
1708
+ "TIME_1655",
1709
+ "TIME_1656",
1710
+ "TIME_1657",
1711
+ "TIME_1658",
1712
+ "TIME_1659",
1713
+ "TIME_1660",
1714
+ "TIME_1661",
1715
+ "TIME_1662",
1716
+ "TIME_1663",
1717
+ "TIME_1664",
1718
+ "TIME_1665",
1719
+ "TIME_1666",
1720
+ "TIME_1667",
1721
+ "TIME_1668",
1722
+ "TIME_1669",
1723
+ "TIME_1670",
1724
+ "TIME_1671",
1725
+ "TIME_1672",
1726
+ "TIME_1673",
1727
+ "TIME_1674",
1728
+ "TIME_1675",
1729
+ "TIME_1676",
1730
+ "TIME_1677",
1731
+ "TIME_1678",
1732
+ "TIME_1679",
1733
+ "TIME_1680",
1734
+ "TIME_1681",
1735
+ "TIME_1682",
1736
+ "TIME_1683",
1737
+ "TIME_1684",
1738
+ "TIME_1685",
1739
+ "TIME_1686",
1740
+ "TIME_1687",
1741
+ "TIME_1688",
1742
+ "TIME_1689",
1743
+ "TIME_1690",
1744
+ "TIME_1691",
1745
+ "TIME_1692",
1746
+ "TIME_1693",
1747
+ "TIME_1694",
1748
+ "TIME_1695",
1749
+ "TIME_1696",
1750
+ "TIME_1697",
1751
+ "TIME_1698",
1752
+ "TIME_1699",
1753
+ "TIME_1700",
1754
+ "TIME_1701",
1755
+ "TIME_1702",
1756
+ "TIME_1703",
1757
+ "TIME_1704",
1758
+ "TIME_1705",
1759
+ "TIME_1706",
1760
+ "TIME_1707",
1761
+ "TIME_1708",
1762
+ "TIME_1709",
1763
+ "TIME_1710",
1764
+ "TIME_1711",
1765
+ "TIME_1712",
1766
+ "TIME_1713",
1767
+ "TIME_1714",
1768
+ "TIME_1715",
1769
+ "TIME_1716",
1770
+ "TIME_1717",
1771
+ "TIME_1718",
1772
+ "TIME_1719",
1773
+ "TIME_1720",
1774
+ "TIME_1721",
1775
+ "TIME_1722",
1776
+ "TIME_1723",
1777
+ "TIME_1724",
1778
+ "TIME_1725",
1779
+ "TIME_1726",
1780
+ "TIME_1727",
1781
+ "SEP",
1782
+ "SIB",
1783
+ "STYLE_0",
1784
+ "STYLE_1",
1785
+ "STYLE_2",
1786
+ "STYLE_3",
1787
+ "STYLE_4",
1788
+ "STYLE_5",
1789
+ "STYLE_6",
1790
+ "STYLE_7",
1791
+ "SYNC_0",
1792
+ "SYNC_1",
1793
+ "SYNC_2",
1794
+ "SYNC_3",
1795
+ "SYNC_4",
1796
+ "SYNC_5",
1797
+ "PLAN",
1798
+ "PDENS_0",
1799
+ "PDENS_1",
1800
+ "PDENS_2",
1801
+ "PDENS_3",
1802
+ "PDENS_4",
1803
+ "PDENS_5",
1804
+ "PDENS_6",
1805
+ "PDENS_7",
1806
+ "PFLAG_0",
1807
+ "PFLAG_1",
1808
+ "PFLAG_2",
1809
+ "MASK",
1810
+ "CPLX_0",
1811
+ "CPLX_1",
1812
+ "CPLX_2",
1813
+ "CPLX_3",
1814
+ "CPLX_4",
1815
+ "CPLX_5",
1816
+ "MODE_slot",
1817
+ "MODE_time",
1818
+ "BAR",
1819
+ "REAL",
1820
+ "DENS_D8_0",
1821
+ "DENS_D8_1",
1822
+ "DENS_D8_2",
1823
+ "DENS_D8_3",
1824
+ "DENS_D8_4",
1825
+ "DENS_D8_5",
1826
+ "DENS_D8_6",
1827
+ "DENS_D8_7",
1828
+ "SPAN_BUDGET_0",
1829
+ "SPAN_BUDGET_1",
1830
+ "SPAN_BUDGET_2",
1831
+ "SPAN_BUDGET_3",
1832
+ "BAR_SYNC_0",
1833
+ "BAR_SYNC_1",
1834
+ "BAR_SYNC_2",
1835
+ "BAR_SYNC_3",
1836
+ "BAR_SYNC_4",
1837
+ "BAR_SYNC_5",
1838
+ "MOTIF_NEW",
1839
+ "MOTIF_REPEAT_1",
1840
+ "MOTIF_REPEAT_2",
1841
+ "MOTIF_REPEAT_4",
1842
+ "MOTIF_REPEAT_8",
1843
+ "MOTIF_VAR_1",
1844
+ "MOTIF_VAR_2",
1845
+ "MOTIF_VAR_4",
1846
+ "MOTIF_VAR_8",
1847
+ "BIG_RATE_0",
1848
+ "BIG_RATE_1",
1849
+ "BIG_RATE_2",
1850
+ "BIG_RATE_3",
1851
+ "BIG_RATE_4",
1852
+ "BIG_RATE_5",
1853
+ "BIG_RATE_6",
1854
+ "BIG_RATE_7",
1855
+ "SPAN_RATE_0",
1856
+ "SPAN_RATE_1",
1857
+ "SPAN_RATE_2",
1858
+ "SPAN_RATE_3",
1859
+ "SPAN_RATE_4",
1860
+ "SPAN_RATE_5",
1861
+ "SPAN_RATE_6",
1862
+ "SPAN_RATE_7",
1863
+ "STREAM_0",
1864
+ "STREAM_1",
1865
+ "STREAM_2",
1866
+ "STREAM_3",
1867
+ "STREAM_4",
1868
+ "STREAM_5",
1869
+ "SPAN_DUR_0",
1870
+ "SPAN_DUR_1",
1871
+ "SPAN_DUR_2",
1872
+ "SPAN_DUR_3",
1873
+ "SPAN_DUR_4",
1874
+ "SPAN_DUR_5",
1875
+ "SPAN_DUR_6",
1876
+ "SPAN_DUR_7",
1877
+ "SPAN_DUR_8",
1878
+ "SPAN_DUR_9",
1879
+ "SPAN_DUR_10",
1880
+ "SPAN_DUR_11",
1881
+ "SPAN_DUR_12",
1882
+ "SPAN_DUR_13",
1883
+ "SPAN_DUR_14",
1884
+ "SPAN_DUR_15",
1885
+ "SPAN_START",
1886
+ "SPAN_END",
1887
+ "BAR_DENOM_1",
1888
+ "BAR_DENOM_2",
1889
+ "BAR_DENOM_3",
1890
+ "BAR_DENOM_4",
1891
+ "BAR_DENOM_5",
1892
+ "BAR_DENOM_6",
1893
+ "BAR_DENOM_7",
1894
+ "BAR_DENOM_8",
1895
+ "BAR_DENOM_9",
1896
+ "BAR_DENOM_10",
1897
+ "BAR_DENOM_11",
1898
+ "BAR_DENOM_12",
1899
+ "BAR_DENOM_13",
1900
+ "BAR_DENOM_14",
1901
+ "BAR_DENOM_15",
1902
+ "BAR_DENOM_16",
1903
+ "BAR_DENOM_17",
1904
+ "BAR_DENOM_18",
1905
+ "BAR_DENOM_19",
1906
+ "BAR_DENOM_20",
1907
+ "BAR_DENOM_21",
1908
+ "BAR_DENOM_22",
1909
+ "BAR_DENOM_23",
1910
+ "BAR_DENOM_24",
1911
+ "BAR_DENOM_25",
1912
+ "BAR_DENOM_26",
1913
+ "BAR_DENOM_27",
1914
+ "BAR_DENOM_28",
1915
+ "BAR_DENOM_29",
1916
+ "BAR_DENOM_30",
1917
+ "BAR_DENOM_31",
1918
+ "BAR_DENOM_32",
1919
+ "BAR_DENOM_33",
1920
+ "BAR_DENOM_34",
1921
+ "BAR_DENOM_35",
1922
+ "BAR_DENOM_36",
1923
+ "BAR_DENOM_37",
1924
+ "BAR_DENOM_38",
1925
+ "BAR_DENOM_39",
1926
+ "BAR_DENOM_40",
1927
+ "BAR_DENOM_41",
1928
+ "BAR_DENOM_42",
1929
+ "BAR_DENOM_43",
1930
+ "BAR_DENOM_44",
1931
+ "BAR_DENOM_45",
1932
+ "BAR_DENOM_46",
1933
+ "BAR_DENOM_47",
1934
+ "BAR_DENOM_48",
1935
+ "BAR_DENOM_49",
1936
+ "BAR_DENOM_50",
1937
+ "BAR_DENOM_51",
1938
+ "BAR_DENOM_52",
1939
+ "BAR_DENOM_53",
1940
+ "BAR_DENOM_54",
1941
+ "BAR_DENOM_55",
1942
+ "BAR_DENOM_56",
1943
+ "BAR_DENOM_57",
1944
+ "BAR_DENOM_58",
1945
+ "BAR_DENOM_59",
1946
+ "BAR_DENOM_60",
1947
+ "BAR_DENOM_61",
1948
+ "BAR_DENOM_62",
1949
+ "BAR_DENOM_63",
1950
+ "BAR_DENOM_64",
1951
+ "BAR_DENOM_65",
1952
+ "BAR_DENOM_66",
1953
+ "BAR_DENOM_67",
1954
+ "BAR_DENOM_68",
1955
+ "BAR_DENOM_69",
1956
+ "BAR_DENOM_70",
1957
+ "BAR_DENOM_71",
1958
+ "BAR_DENOM_72",
1959
+ "BAR_DENOM_73",
1960
+ "BAR_DENOM_74",
1961
+ "BAR_DENOM_75",
1962
+ "BAR_DENOM_76",
1963
+ "BAR_DENOM_77",
1964
+ "BAR_DENOM_78",
1965
+ "BAR_DENOM_79",
1966
+ "BAR_DENOM_80",
1967
+ "BAR_DENOM_81",
1968
+ "BAR_DENOM_82",
1969
+ "BAR_DENOM_83",
1970
+ "BAR_DENOM_84",
1971
+ "BAR_DENOM_85",
1972
+ "BAR_DENOM_86",
1973
+ "BAR_DENOM_87",
1974
+ "BAR_DENOM_88",
1975
+ "BAR_DENOM_89",
1976
+ "BAR_DENOM_90",
1977
+ "BAR_DENOM_91",
1978
+ "BAR_DENOM_92",
1979
+ "BAR_DENOM_93",
1980
+ "BAR_DENOM_94",
1981
+ "BAR_DENOM_95",
1982
+ "BAR_DENOM_96",
1983
+ "BAR_DENOM_97",
1984
+ "BAR_DENOM_98",
1985
+ "BAR_DENOM_99",
1986
+ "BAR_DENOM_100",
1987
+ "BAR_DENOM_101",
1988
+ "BAR_DENOM_102",
1989
+ "BAR_DENOM_103",
1990
+ "BAR_DENOM_104",
1991
+ "BAR_DENOM_105",
1992
+ "BAR_DENOM_106",
1993
+ "BAR_DENOM_107",
1994
+ "BAR_DENOM_108",
1995
+ "BAR_DENOM_109",
1996
+ "BAR_DENOM_110",
1997
+ "BAR_DENOM_111",
1998
+ "BAR_DENOM_112",
1999
+ "BAR_DENOM_113",
2000
+ "BAR_DENOM_114",
2001
+ "BAR_DENOM_115",
2002
+ "BAR_DENOM_116",
2003
+ "BAR_DENOM_117",
2004
+ "BAR_DENOM_118",
2005
+ "BAR_DENOM_119",
2006
+ "BAR_DENOM_120",
2007
+ "BAR_DENOM_121",
2008
+ "BAR_DENOM_122",
2009
+ "BAR_DENOM_123",
2010
+ "BAR_DENOM_124",
2011
+ "BAR_DENOM_125",
2012
+ "BAR_DENOM_126",
2013
+ "BAR_DENOM_127",
2014
+ "BAR_DENOM_128",
2015
+ "BAR_DENOM_129",
2016
+ "BAR_DENOM_130",
2017
+ "BAR_DENOM_131",
2018
+ "BAR_DENOM_132",
2019
+ "BAR_DENOM_133",
2020
+ "BAR_DENOM_134",
2021
+ "BAR_DENOM_135",
2022
+ "BAR_DENOM_136",
2023
+ "BAR_DENOM_137",
2024
+ "BAR_DENOM_138",
2025
+ "BAR_DENOM_139",
2026
+ "BAR_DENOM_140",
2027
+ "BAR_DENOM_141",
2028
+ "BAR_DENOM_142",
2029
+ "BAR_DENOM_143",
2030
+ "BAR_DENOM_144",
2031
+ "BAR_DENOM_145",
2032
+ "BAR_DENOM_146",
2033
+ "BAR_DENOM_147",
2034
+ "BAR_DENOM_148",
2035
+ "BAR_DENOM_149",
2036
+ "BAR_DENOM_150",
2037
+ "BAR_DENOM_151",
2038
+ "BAR_DENOM_152",
2039
+ "BAR_DENOM_153",
2040
+ "BAR_DENOM_154",
2041
+ "BAR_DENOM_155",
2042
+ "BAR_DENOM_156",
2043
+ "BAR_DENOM_157",
2044
+ "BAR_DENOM_158",
2045
+ "BAR_DENOM_159",
2046
+ "BAR_DENOM_160",
2047
+ "BAR_DENOM_161",
2048
+ "BAR_DENOM_162",
2049
+ "BAR_DENOM_163",
2050
+ "BAR_DENOM_164",
2051
+ "BAR_DENOM_165",
2052
+ "BAR_DENOM_166",
2053
+ "BAR_DENOM_167",
2054
+ "BAR_DENOM_168",
2055
+ "BAR_DENOM_169",
2056
+ "BAR_DENOM_170",
2057
+ "BAR_DENOM_171",
2058
+ "BAR_DENOM_172",
2059
+ "BAR_DENOM_173",
2060
+ "BAR_DENOM_174",
2061
+ "BAR_DENOM_175",
2062
+ "BAR_DENOM_176",
2063
+ "BAR_DENOM_177",
2064
+ "BAR_DENOM_178",
2065
+ "BAR_DENOM_179",
2066
+ "BAR_DENOM_180",
2067
+ "BAR_DENOM_181",
2068
+ "BAR_DENOM_182",
2069
+ "BAR_DENOM_183",
2070
+ "BAR_DENOM_184",
2071
+ "BAR_DENOM_185",
2072
+ "BAR_DENOM_186",
2073
+ "BAR_DENOM_187",
2074
+ "BAR_DENOM_188",
2075
+ "BAR_DENOM_189",
2076
+ "BAR_DENOM_190",
2077
+ "BAR_DENOM_191",
2078
+ "BAR_DENOM_192"
2079
+ ]
2080
+ }