chironYu commited on
Commit
d7d056e
·
verified ·
1 Parent(s): 3213d06

The view-size sweep of Table A6: nine checkpoints; the plan and the card follow the appendix's table numbers

Browse files

I-CE with each review view drawn until it holds at least 8, 32 or 64 sentences (the recipe stops at 16), folds 0 to 2 on the 512-sentence packs, each at the epoch the validation rank rule selects. rank_ckpt_plan.json lists them (166 runs). The appendix's new Table A6 moved every later table one place, so the card's table numbers follow: the old Tables A6 to A18 are A7 to A19.

README.md CHANGED
@@ -24,11 +24,11 @@ reviews (Qwen3-Embedding-0.6B, 1,024-d) and an MLP maps the pooled result to one
24
 
25
  ## Contents
26
 
27
- 266 checkpoints, about 580 MiB, and four JSON files that list every one of them:
28
 
29
  | Listed in | Checkpoints | What they are |
30
  |---|---|---|
31
- | `rank_ckpt_plan.json` | 157 | the five-fold runs behind the paper's numbers, each at the epoch the validation rank rule of Section 5 selects. One is the DINO-style control of Table A1, trained on fold 0 only; five, CE capped at 2,048 sentences, are the partners of I-CE runs the paper prints and are not printed themselves |
32
  | `i2bce_sel_pool.json` | 5 | I + SimCLR on the five folds (Figure 8, Appendix L), selected by the same rule |
33
  | `fixed_split_ckpt_plan.json` | 69 | the fixed-split runs: 15 runs at their selected epochs, two of them with an earlier checkpoint kept beside the selected one, and 52 more checkpoints of two runs that the paper reads across their epochs (field `trajectory`) |
34
  | `rank_ckpt_plan.json`, field `init_ckpt` | 5 | the BYOL checkpoints the five two-stage runs started from |
@@ -43,8 +43,8 @@ Every tower that a number or a figure of the paper reads is here, at the checkpo
43
  at. A run saved a checkpoint every 50 epochs and the rank rule chose among them; the ones it did
44
  not choose are not here, except for the two runs listed under `trajectory`. The other passages
45
  that follow runs over training quote what the workers logged at each checkpoint: the learnable
46
- temperature and the view-count sweep in Appendix C, and the decline of the views-first anchors
47
- after epoch 50 in Supplement A.
48
 
49
  ## File names
50
 
@@ -57,33 +57,36 @@ after epoch 50 in Supplement A.
57
  whole 2,048-sentence review pool (a median of 2,068 sentences). The paper calls `_g2048` and
58
  `_gpool` alike its 2,048-sentence budget; the main five-fold towers are the `_gpool` runs.
59
  - `<suffix>`: `_wllm` for LLM-rewritten documents as the document view, `_bw` for a run
60
- initialized from a BYOL tower, `_d512` for a run on the current 512-sentence packs (see below).
 
 
61
  - `_fp`: the strong views were drawn from the full review pool. `ep<N>` is the epoch.
62
 
63
  ## The five-fold runs
64
 
65
- Table and figure numbers are those of the paper, its appendices (Tables A1 to A18) and its
66
  supplement (Tables S1 to S4). The last column comes from the provenance table of the code
67
  repository, `contrast_experiment/w9/Pod/w9_profiles.py`, which cites by cell: a table named for a
68
  row reads at least one of the row's runs. That file also names the notebook that trains each cell.
69
 
70
  | Runs | What they are | Pack | Folds | Read by |
71
  |---|---|---|---|---|
72
- | `ai2ce_icetf_d512` | CE + anchor-in-I | 512 | 5 | Table 2, Section 6.2, Table A14 |
73
- | `ce_cetf_d512`, `i2ce_icetf_d512` | CE; I-CE, the recipe | 512 | 5 | Table 2, Table A2, Table A3, Table A4, Table A5, Table A6, Table A10, Table A14, Table A18, Table S1 |
74
  | `d1r2_i2ce_icetf_d512`, `d1r4_i2ce_icetf_d512`, `d1r5_i2ce_icetf_d512`, `d1r5_i3ce_icetf_d512`, `d1r6_i2ce_icetf_d512` | I-CE, V = 3; I-CE, V = 5; I-CE, V = 6; I-CE, V = 6 with λ = 3; I-CE, V = 7 | 512 | 0 to 2 | Table A5 |
75
- | `i2ce_icetf_wllm_d512`, `nodoc_i2ce_icetf_d512` | I-CE, LLM-rewritten documents; I-CE without the document view | 512 | 5 | Table A18 |
 
76
  | `i2cet05_icetf_d512`, `i2cet10_icetf_d512` | I-CE, τ = 0.05; I-CE, τ = 0.10 | 512 | 5 | Table A3 |
77
  | `i3ce_icetf_d512`, `i4ce_icetf_d512`, `ice_icetf_d512` | I-CE, λ = 3; I-CE, λ = 4; I-CE, λ = 1 | 512 | 0 to 2 | Table A4 |
78
- | `slot16i2cemean_icetf_d512`, `slot1i2cemean_icetf_d512`, `slot2i2cemean_icetf_d512`, `slot8i2cemean_icetf_d512` | I-CE, 16 latent queries; I-CE, 1 latent query; I-CE, 2 latent queries; I-CE, 8 latent queries | 512 | 0 to 2 | Table A6 |
79
  | `vfai2ce_icetf_d512` | I-CE, views-first anchors | 512 | 5 | Table S1 |
80
- | `ce_cetf_g1024`, `i2ce_icetf_g1024` | CE; I-CE, the recipe | 1,024 | 5 | Table A2, Table A10, Table A14 |
81
- | `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | 5 | Table A3, Table A10 |
82
  | `i2ce_icetl_g2048`, `i2cet05_icetf_g2048`, `i2cet10_icetf_g2048` | I-CE, learnable τ; I-CE, τ = 0.05; I-CE, τ = 0.10 | 2,048 | 5 | Table A3 |
83
- | `bce_cetf_gpool`, `byol_bytf_gpool`, `epd_v20i10c20_cetf_gpool`, `mq3072i2ce_icetf_gpool` | SimCLR-style; BYOL; VICReg (batch 192); I-CE with an EMA teacher and a memory bank | pool | 5 | Figure 7, Figure 8, Table 2, Section 3.2, Table A1, Table A12, Table A15, Table A16, Table A17, Figure A2, Table S2, Table S3, Table S4 |
84
- | `ce_cetf_gpool`, `i2ce_icetf_gpool` | CE; I-CE, the recipe | pool | 5 | Figure 7, Figure 8, Table 2, Section 6.1, Table A1, Table A2, Table A6, Table A12, Table A13, Table A14, Table A15, Table A16, Table A17, Figure A2, Table S2, Table S3, Table S4 |
85
- | `slot8i2cemean_icetf_gpool` | I-CE, 8 latent queries | pool | 5 | Table A6 |
86
- | `swin168step84loop2i2ce_icetf_gpool`, `swin168step84loop2i2ce_icetf_gpool_bw` | I-CE with the windowed teacher; I-CE, two-stage: BYOL warm start, then the windowed teacher | pool | 5 | Table A12, Table S2, Table S3, Table S4 |
87
  | `dino_dntf_gpool` | DINO at its published settings, the control of Section 4.3 | pool | 0 only | Table A1 |
88
 
89
  I + SimCLR (`i2bce_icetf`, files without a pack or a suffix, five folds) backs Figure 8 and the
@@ -97,10 +100,10 @@ it carries the two prototype heads (see Format).
97
  | Runs | What they are | Pack | Folds | Read by |
98
  |---|---|---|---|---|
99
  | `ai25auni2_icetf_d512`, `ai2auni25_icetf_d512`, `ai4auni2_icetf_d512`, `ai6auni2_icetf_d512`, `i4uni2_icetf` | alignment + uniformity aimed at the anchor gallery, the four settings of Section 6.2 (`ai25auni2` is Table 3's a+u @ gallery row); alignment + uniformity aimed at the batch views (Table 3's a+u @ views row) | 512 | fixed split | Table 3, Section 6.2, Appendix L |
100
- | `ai2ce_icetf_d512`, `ce_cetf_d512`, `i2ce_icetf_d512` | CE + anchor-in-I; CE; I-CE, the recipe | 512 | fixed split | Table 3, Section 6.2, Table A7 |
101
  | `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | fixed split | Figure 9, Figure A1, Figure A3, Appendix L |
102
- | `bce_cetf`, `i2bce_icetf`, `epd_v20i10c20_cetf` | SimCLR-style; I + SimCLR; VICReg, v/i/c = 20/10/20 (the recipe weights) | 512 | fixed split | Figure 9, Figure A3, Appendix L, Table A7 |
103
- | `epd_v25i25c1_cetf`, `epd_v20i10c15_cetf` | VICReg, v/i/c = 25/25/1; VICReg, v/i/c = 20/10/15 | 512 | fixed split | Table A7 |
104
 
105
  Figure 9, Figure A3 and the neighborhood statistics of Appendix L read six encoders on the fixed
106
  split: the frozen embedder, which has no checkpoint, and five towers, namely `i2ce_icetf_g2048`
@@ -162,7 +165,7 @@ python contrast_experiment/w9/eval/k10_eval.py score cells.json draws.npz score
162
  python contrast_experiment/w9/eval/k10_eval.py table scores.json
163
  ```
164
 
165
- `cells` takes the 162 five-fold checkpoints of `rank_ckpt_plan.json` and `i2bce_sel_pool.json`;
166
  `--only <regex>` keeps a subset, for example `--only "_d512$"` for the 512-sentence runs. In its
167
  table the five families of the main comparison go by the paper's names (`I-CE`, `CE`, `SimCLR`,
168
  `VICReg`, `BYOL`) and every other family by its plan key without the fold; a family with one
@@ -176,7 +179,7 @@ python contrast_experiment/w9/eval/fixed_split_eval.py --ckpt-dir checkpoints
176
 
177
  It reads every run of `fixed_split_ckpt_plan.json` at its selected checkpoint and prints each
178
  reading beside the number the paper prints: Table 3, the gallery-aimed settings of Section 6.2
179
- and Table A7 on the one 4,096-sentence anchor draw of the fixed split, the SimCLR-style test
180
  curve of Appendix L, the latent queries of Figure A1, and, through `k10_eval.py`, the towers of
181
  Figure 9 against the ten draws. A near tie can move a cell by one query of 204 (0.005), and the
182
  collapsed a+u @ gallery row by about 0.01.
@@ -230,6 +233,11 @@ dataset keeps the first packs as `wscan_gal_rev_rawprefix.npz`; the workers read
230
 
231
  ## Changes
232
 
 
 
 
 
 
233
  - 2026-11-12. The code repository gained `figures/neighborhoods/`, the scripts that drew
234
  Figures 8, 9 and A3.
235
  - 2026-11-11, at night. The code repository gained `fixed_split_eval.py`, which reads the
 
24
 
25
  ## Contents
26
 
27
+ 275 checkpoints, about 593 MiB, and four JSON files that list every one of them:
28
 
29
  | Listed in | Checkpoints | What they are |
30
  |---|---|---|
31
+ | `rank_ckpt_plan.json` | 166 | the five-fold runs behind the paper's numbers, each at the epoch the validation rank rule of Section 5 selects. One is the DINO-style control of Table A1, trained on fold 0 only; five, CE capped at 2,048 sentences, are the partners of I-CE runs the paper prints and are not printed themselves |
32
  | `i2bce_sel_pool.json` | 5 | I + SimCLR on the five folds (Figure 8, Appendix L), selected by the same rule |
33
  | `fixed_split_ckpt_plan.json` | 69 | the fixed-split runs: 15 runs at their selected epochs, two of them with an earlier checkpoint kept beside the selected one, and 52 more checkpoints of two runs that the paper reads across their epochs (field `trajectory`) |
34
  | `rank_ckpt_plan.json`, field `init_ckpt` | 5 | the BYOL checkpoints the five two-stage runs started from |
 
43
  at. A run saved a checkpoint every 50 epochs and the rank rule chose among them; the ones it did
44
  not choose are not here, except for the two runs listed under `trajectory`. The other passages
45
  that follow runs over training quote what the workers logged at each checkpoint: the learnable
46
+ temperature and the view-count and view-size sweeps in Appendix C, and the decline of the
47
+ views-first anchors after epoch 50 in Supplement A.
48
 
49
  ## File names
50
 
 
57
  whole 2,048-sentence review pool (a median of 2,068 sentences). The paper calls `_g2048` and
58
  `_gpool` alike its 2,048-sentence budget; the main five-fold towers are the `_gpool` runs.
59
  - `<suffix>`: `_wllm` for LLM-rewritten documents as the document view, `_bw` for a run
60
+ initialized from a BYOL tower, `_w<N>` for review views drawn until they hold at least N
61
+ sentences (the recipe stops at 16), `_d512` for a run on the current 512-sentence packs (see
62
+ below).
63
  - `_fp`: the strong views were drawn from the full review pool. `ep<N>` is the epoch.
64
 
65
  ## The five-fold runs
66
 
67
+ Table and figure numbers are those of the paper, its appendices (Tables A1 to A19) and its
68
  supplement (Tables S1 to S4). The last column comes from the provenance table of the code
69
  repository, `contrast_experiment/w9/Pod/w9_profiles.py`, which cites by cell: a table named for a
70
  row reads at least one of the row's runs. That file also names the notebook that trains each cell.
71
 
72
  | Runs | What they are | Pack | Folds | Read by |
73
  |---|---|---|---|---|
74
+ | `ai2ce_icetf_d512` | CE + anchor-in-I | 512 | 5 | Table 2, Section 6.2, Table A15 |
75
+ | `ce_cetf_d512`, `i2ce_icetf_d512` | CE; I-CE, the recipe | 512 | 5 | Table 2, Table A2, Table A3, Table A4, Table A5, Table A6, Table A7, Table A11, Table A15, Table A19, Table S1 |
76
  | `d1r2_i2ce_icetf_d512`, `d1r4_i2ce_icetf_d512`, `d1r5_i2ce_icetf_d512`, `d1r5_i3ce_icetf_d512`, `d1r6_i2ce_icetf_d512` | I-CE, V = 3; I-CE, V = 5; I-CE, V = 6; I-CE, V = 6 with λ = 3; I-CE, V = 7 | 512 | 0 to 2 | Table A5 |
77
+ | `i2ce_icetf_w8_d512`, `i2ce_icetf_w32_d512`, `i2ce_icetf_w64_d512` | I-CE, review views of at least 8 sentences; of at least 32; of at least 64 | 512 | 0 to 2 | Table A6 |
78
+ | `i2ce_icetf_wllm_d512`, `nodoc_i2ce_icetf_d512` | I-CE, LLM-rewritten documents; I-CE without the document view | 512 | 5 | Table A19 |
79
  | `i2cet05_icetf_d512`, `i2cet10_icetf_d512` | I-CE, τ = 0.05; I-CE, τ = 0.10 | 512 | 5 | Table A3 |
80
  | `i3ce_icetf_d512`, `i4ce_icetf_d512`, `ice_icetf_d512` | I-CE, λ = 3; I-CE, λ = 4; I-CE, λ = 1 | 512 | 0 to 2 | Table A4 |
81
+ | `slot16i2cemean_icetf_d512`, `slot1i2cemean_icetf_d512`, `slot2i2cemean_icetf_d512`, `slot8i2cemean_icetf_d512` | I-CE, 16 latent queries; I-CE, 1 latent query; I-CE, 2 latent queries; I-CE, 8 latent queries | 512 | 0 to 2 | Table A7 |
82
  | `vfai2ce_icetf_d512` | I-CE, views-first anchors | 512 | 5 | Table S1 |
83
+ | `ce_cetf_g1024`, `i2ce_icetf_g1024` | CE; I-CE, the recipe | 1,024 | 5 | Table A2, Table A11, Table A15 |
84
+ | `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | 5 | Table A3, Table A11 |
85
  | `i2ce_icetl_g2048`, `i2cet05_icetf_g2048`, `i2cet10_icetf_g2048` | I-CE, learnable τ; I-CE, τ = 0.05; I-CE, τ = 0.10 | 2,048 | 5 | Table A3 |
86
+ | `bce_cetf_gpool`, `byol_bytf_gpool`, `epd_v20i10c20_cetf_gpool`, `mq3072i2ce_icetf_gpool` | SimCLR-style; BYOL; VICReg (batch 192); I-CE with an EMA teacher and a memory bank | pool | 5 | Figure 7, Figure 8, Table 2, Section 3.2, Table A1, Table A13, Table A16, Table A17, Table A18, Figure A2, Table S2, Table S3, Table S4 |
87
+ | `ce_cetf_gpool`, `i2ce_icetf_gpool` | CE; I-CE, the recipe | pool | 5 | Figure 7, Figure 8, Table 2, Section 6.1, Table A1, Table A2, Table A7, Table A13, Table A14, Table A15, Table A16, Table A17, Table A18, Figure A2, Table S2, Table S3, Table S4 |
88
+ | `slot8i2cemean_icetf_gpool` | I-CE, 8 latent queries | pool | 5 | Table A7 |
89
+ | `swin168step84loop2i2ce_icetf_gpool`, `swin168step84loop2i2ce_icetf_gpool_bw` | I-CE with the windowed teacher; I-CE, two-stage: BYOL warm start, then the windowed teacher | pool | 5 | Table A13, Table S2, Table S3, Table S4 |
90
  | `dino_dntf_gpool` | DINO at its published settings, the control of Section 4.3 | pool | 0 only | Table A1 |
91
 
92
  I + SimCLR (`i2bce_icetf`, files without a pack or a suffix, five folds) backs Figure 8 and the
 
100
  | Runs | What they are | Pack | Folds | Read by |
101
  |---|---|---|---|---|
102
  | `ai25auni2_icetf_d512`, `ai2auni25_icetf_d512`, `ai4auni2_icetf_d512`, `ai6auni2_icetf_d512`, `i4uni2_icetf` | alignment + uniformity aimed at the anchor gallery, the four settings of Section 6.2 (`ai25auni2` is Table 3's a+u @ gallery row); alignment + uniformity aimed at the batch views (Table 3's a+u @ views row) | 512 | fixed split | Table 3, Section 6.2, Appendix L |
103
+ | `ai2ce_icetf_d512`, `ce_cetf_d512`, `i2ce_icetf_d512` | CE + anchor-in-I; CE; I-CE, the recipe | 512 | fixed split | Table 3, Section 6.2, Table A8 |
104
  | `ce_cetf_g2048`, `i2ce_icetf_g2048` | CE; I-CE, the recipe | 2,048 | fixed split | Figure 9, Figure A1, Figure A3, Appendix L |
105
+ | `bce_cetf`, `i2bce_icetf`, `epd_v20i10c20_cetf` | SimCLR-style; I + SimCLR; VICReg, v/i/c = 20/10/20 (the recipe weights) | 512 | fixed split | Figure 9, Figure A3, Appendix L, Table A8 |
106
+ | `epd_v25i25c1_cetf`, `epd_v20i10c15_cetf` | VICReg, v/i/c = 25/25/1; VICReg, v/i/c = 20/10/15 | 512 | fixed split | Table A8 |
107
 
108
  Figure 9, Figure A3 and the neighborhood statistics of Appendix L read six encoders on the fixed
109
  split: the frozen embedder, which has no checkpoint, and five towers, namely `i2ce_icetf_g2048`
 
165
  python contrast_experiment/w9/eval/k10_eval.py table scores.json
166
  ```
167
 
168
+ `cells` takes the 171 five-fold checkpoints of `rank_ckpt_plan.json` and `i2bce_sel_pool.json`;
169
  `--only <regex>` keeps a subset, for example `--only "_d512$"` for the 512-sentence runs. In its
170
  table the five families of the main comparison go by the paper's names (`I-CE`, `CE`, `SimCLR`,
171
  `VICReg`, `BYOL`) and every other family by its plan key without the fold; a family with one
 
179
 
180
  It reads every run of `fixed_split_ckpt_plan.json` at its selected checkpoint and prints each
181
  reading beside the number the paper prints: Table 3, the gallery-aimed settings of Section 6.2
182
+ and Table A8 on the one 4,096-sentence anchor draw of the fixed split, the SimCLR-style test
183
  curve of Appendix L, the latent queries of Figure A1, and, through `k10_eval.py`, the towers of
184
  Figure 9 against the ten draws. A near tie can move a cell by one query of 204 (0.005), and the
185
  collapsed a+u @ gallery row by about 0.01.
 
233
 
234
  ## Changes
235
 
236
+ - 2026-11-12, later that day. Nine checkpoints were added, the view-size sweep of the appendix's
237
+ new Table A6: I-CE with each review view drawn until it holds at least 8, 32 or 64 sentences
238
+ (`i2ce_icetf_fold<k>_w<N>_d512`, folds 0 to 2). `rank_ckpt_plan.json` lists them. The new table
239
+ moved every later appendix table one place, so the old Tables A6 to A18 are Tables A7 to A19
240
+ in the tables above; no earlier entry changed its checkpoint.
241
  - 2026-11-12. The code repository gained `figures/neighborhoods/`, the scripts that drew
242
  Figures 8, 9 and A3.
243
  - 2026-11-11, at night. The code repository gained `fixed_split_eval.py`, which reads the
ckpt_w9cv_wcle_i2ce_icetf_fold0_w32_d512_fp_ep1700.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfafb905c3989ae34474a24f6d1841bc1efb1e1b44f823e91aca0d2b887eee4e
3
+ size 1452509
ckpt_w9cv_wcle_i2ce_icetf_fold0_w64_d512_fp_ep2000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16c295bd3ca867035e98aeeb6bf6d1bebb3baadf94a62787a6648384cd308754
3
+ size 1452509
ckpt_w9cv_wcle_i2ce_icetf_fold0_w8_d512_fp_ep1850.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:709f72276caec20a72b9fc10667b8bb26c75d40058f1f8ed7a5bf2eba1f2f438
3
+ size 1452492
ckpt_w9cv_wcle_i2ce_icetf_fold1_w32_d512_fp_ep1300.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18c4155b1414c7783ee710b7306a9fde9a65f423da64747a29b9507d68254eb1
3
+ size 1452509
ckpt_w9cv_wcle_i2ce_icetf_fold1_w64_d512_fp_ep950.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a5aebdeaed6d016b0f4b85dc53e369cf3363991a874fc34a9deb82c29694bba
3
+ size 1452492
ckpt_w9cv_wcle_i2ce_icetf_fold1_w8_d512_fp_ep1800.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bca75f8340c9786ebc021db8a21d662fc6b3d6cdc3a34ad50c7c8d22a709c58f
3
+ size 1452492
ckpt_w9cv_wcle_i2ce_icetf_fold2_w32_d512_fp_ep1550.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b24af1da78dc89c11d49ba62f450d67f8d2d07be8ec89d5d1582ecda6695365
3
+ size 1452509
ckpt_w9cv_wcle_i2ce_icetf_fold2_w64_d512_fp_ep1750.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb19e080fb423c8d032e62f9f0b6e527dc3909498f3c8b0bb033f6fdbff07708
3
+ size 1452509
ckpt_w9cv_wcle_i2ce_icetf_fold2_w8_d512_fp_ep1850.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cffc1869a3b2bab8beec44f48a501e0dc522cb79026f72aff3a438697997e6a4
3
+ size 1452492
rank_ckpt_plan.json CHANGED
@@ -339,6 +339,96 @@
339
  "sha256": "1f565f77cecf8a283d8323ecfd7793fc1c155823187518f4e15e35e6cc85a73d",
340
  "ranksel": 103.3357064881622
341
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
342
  "i2cet05_icetf_fold0_d512": {
343
  "arm": "i2cet05_icetf",
344
  "fold": 0,
 
339
  "sha256": "1f565f77cecf8a283d8323ecfd7793fc1c155823187518f4e15e35e6cc85a73d",
340
  "ranksel": 103.3357064881622
341
  },
342
+ "i2ce_icetf_fold0_w8_d512": {
343
+ "arm": "i2ce_icetf",
344
+ "fold": 0,
345
+ "cap": "",
346
+ "sfx": "_w8_d512",
347
+ "ep": 1850,
348
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold0_w8_d512_fp_ep1850.pt",
349
+ "sha256": "709f72276caec20a72b9fc10667b8bb26c75d40058f1f8ed7a5bf2eba1f2f438",
350
+ "ranksel": 102.15828396294594
351
+ },
352
+ "i2ce_icetf_fold1_w8_d512": {
353
+ "arm": "i2ce_icetf",
354
+ "fold": 1,
355
+ "cap": "",
356
+ "sfx": "_w8_d512",
357
+ "ep": 1800,
358
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold1_w8_d512_fp_ep1800.pt",
359
+ "sha256": "bca75f8340c9786ebc021db8a21d662fc6b3d6cdc3a34ad50c7c8d22a709c58f",
360
+ "ranksel": 100.92692735919152
361
+ },
362
+ "i2ce_icetf_fold2_w8_d512": {
363
+ "arm": "i2ce_icetf",
364
+ "fold": 2,
365
+ "cap": "",
366
+ "sfx": "_w8_d512",
367
+ "ep": 1850,
368
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold2_w8_d512_fp_ep1850.pt",
369
+ "sha256": "cffc1869a3b2bab8beec44f48a501e0dc522cb79026f72aff3a438697997e6a4",
370
+ "ranksel": 104.10870524765727
371
+ },
372
+ "i2ce_icetf_fold0_w32_d512": {
373
+ "arm": "i2ce_icetf",
374
+ "fold": 0,
375
+ "cap": "",
376
+ "sfx": "_w32_d512",
377
+ "ep": 1700,
378
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold0_w32_d512_fp_ep1700.pt",
379
+ "sha256": "bfafb905c3989ae34474a24f6d1841bc1efb1e1b44f823e91aca0d2b887eee4e",
380
+ "ranksel": 102.08490009321653
381
+ },
382
+ "i2ce_icetf_fold1_w32_d512": {
383
+ "arm": "i2ce_icetf",
384
+ "fold": 1,
385
+ "cap": "",
386
+ "sfx": "_w32_d512",
387
+ "ep": 1300,
388
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold1_w32_d512_fp_ep1300.pt",
389
+ "sha256": "18c4155b1414c7783ee710b7306a9fde9a65f423da64747a29b9507d68254eb1",
390
+ "ranksel": 100.07076937461179
391
+ },
392
+ "i2ce_icetf_fold2_w32_d512": {
393
+ "arm": "i2ce_icetf",
394
+ "fold": 2,
395
+ "cap": "",
396
+ "sfx": "_w32_d512",
397
+ "ep": 1550,
398
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold2_w32_d512_fp_ep1550.pt",
399
+ "sha256": "1b24af1da78dc89c11d49ba62f450d67f8d2d07be8ec89d5d1582ecda6695365",
400
+ "ranksel": 102.88761013766694
401
+ },
402
+ "i2ce_icetf_fold0_w64_d512": {
403
+ "arm": "i2ce_icetf",
404
+ "fold": 0,
405
+ "cap": "",
406
+ "sfx": "_w64_d512",
407
+ "ep": 2000,
408
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold0_w64_d512_fp_ep2000.pt",
409
+ "sha256": "16c295bd3ca867035e98aeeb6bf6d1bebb3baadf94a62787a6648384cd308754",
410
+ "ranksel": 98.52861143847147
411
+ },
412
+ "i2ce_icetf_fold1_w64_d512": {
413
+ "arm": "i2ce_icetf",
414
+ "fold": 1,
415
+ "cap": "",
416
+ "sfx": "_w64_d512",
417
+ "ep": 950,
418
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold1_w64_d512_fp_ep950.pt",
419
+ "sha256": "9a5aebdeaed6d016b0f4b85dc53e369cf3363991a874fc34a9deb82c29694bba",
420
+ "ranksel": 96.14252478109898
421
+ },
422
+ "i2ce_icetf_fold2_w64_d512": {
423
+ "arm": "i2ce_icetf",
424
+ "fold": 2,
425
+ "cap": "",
426
+ "sfx": "_w64_d512",
427
+ "ep": 1750,
428
+ "ckpt": "ckpt_w9cv_wcle_i2ce_icetf_fold2_w64_d512_fp_ep1750.pt",
429
+ "sha256": "fb19e080fb423c8d032e62f9f0b6e527dc3909498f3c8b0bb033f6fdbff07708",
430
+ "ranksel": 102.6732653888046
431
+ },
432
  "i2cet05_icetf_fold0_d512": {
433
  "arm": "i2cet05_icetf",
434
  "fold": 0,