appvoid commited on
Commit
596994f
·
verified ·
1 Parent(s): 0ae1bfc

Qwen90M step 930000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +48 -48
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +3 -3
  6. shard_scheduler.json +42 -378
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 925000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
- "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343",
14
- "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9",
15
- "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 930000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
+ "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4",
14
+ "shard_scheduler.json": "b3c49f97562f447c57be5ea107f2e5123930591bd4d8ad5d22e1610da58f2a24",
15
+ "coverage.json": "32cef4aca490da53b293a3ee6d9244fadb87af3f5c0e004c874df34f97a8fb94",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 925000,
3
  "stats": {
4
- "climbmix/blocks": 33300000,
5
- "climbmix/targets": 55667089526,
6
- "climbmix/input_tokens": 55670617480,
7
- "ultra_qa/blocks": 12487500,
8
- "ultra_qa/targets": 20372477924,
9
- "ultra_qa/input_tokens": 20372517045,
10
- "cortex/blocks": 12487500,
11
- "cortex/targets": 474833481,
12
- "cortex/input_tokens": 19849928864,
13
- "ultra_style/blocks": 12487500,
14
- "ultra_style/targets": 19750266654,
15
- "ultra_style/input_tokens": 19751301641,
16
- "openmath/blocks": 8325000,
17
- "openmath/targets": 12863641674,
18
- "openmath/input_tokens": 12867927568,
19
- "rewrite6/blocks": 4162500,
20
- "rewrite6/targets": 6823898189,
21
- "rewrite6/input_tokens": 6831379657
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4560,
26
- "row": 2473,
27
  "start": {
28
- "file": 220,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 62,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 18680233
38
  },
39
  "rewrite6": {
40
- "page": 44,
41
- "row": 2622,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 180224
47
  }
48
  },
49
- "epoch": 51,
50
- "offset": 4096,
51
- "documents": 10658450
52
  },
53
  "openmath": {
54
- "page": 2850,
55
- "row": 1200,
56
  "start": {
57
  "file": 26,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 320,
61
- "row": 703
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 11674800
67
  },
68
  "ultra_qa": {
69
- "page": 1232,
70
- "row": 1402,
71
  "start": {
72
  "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 5,
76
- "row": 46700
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 5047674
82
  },
83
  "ultra_style": {
84
- "page": 1880,
85
- "row": 3523,
86
  "start": {
87
  "file": 11,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 1,
91
- "row": 81311
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 0,
96
- "documents": 7704003
97
  },
98
  "cortex": {
99
- "page": 242179,
100
- "row": 111,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 30999023
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 930000,
3
  "stats": {
4
+ "climbmix/blocks": 33480000,
5
+ "climbmix/targets": 55968202162,
6
+ "climbmix/input_tokens": 55971749108,
7
+ "ultra_qa/blocks": 12555000,
8
+ "ultra_qa/targets": 20482474984,
9
+ "ultra_qa/input_tokens": 20482514313,
10
+ "cortex/blocks": 12555000,
11
+ "cortex/targets": 477403369,
12
+ "cortex/input_tokens": 19957266714,
13
+ "ultra_style/blocks": 12555000,
14
+ "ultra_style/targets": 19857263635,
15
+ "ultra_style/input_tokens": 19858304258,
16
+ "openmath/blocks": 8370000,
17
+ "openmath/targets": 12933151733,
18
+ "openmath/input_tokens": 12937461063,
19
+ "rewrite6/blocks": 4185000,
20
+ "rewrite6/targets": 6860771033,
21
+ "rewrite6/input_tokens": 6868292697
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4585,
26
+ "row": 1014,
27
  "start": {
28
+ "file": 221,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 79,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 18781174
38
  },
39
  "rewrite6": {
40
+ "page": 8,
41
+ "row": 1951,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 32768
47
  }
48
  },
49
+ "epoch": 52,
50
+ "offset": 0,
51
+ "documents": 10715727
52
  },
53
  "openmath": {
54
+ "page": 2865,
55
+ "row": 3170,
56
  "start": {
57
  "file": 26,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 382,
61
+ "row": 143
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 11738210
67
  },
68
  "ultra_qa": {
69
+ "page": 1239,
70
+ "row": 96,
71
  "start": {
72
  "file": 9,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 6,
76
+ "row": 10847
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 5075040
82
  },
83
  "ultra_style": {
84
+ "page": 1890,
85
+ "row": 4094,
86
  "start": {
87
  "file": 11,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 2,
91
+ "row": 36445
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 0,
96
+ "documents": 7745534
97
  },
98
  "cortex": {
99
+ "page": 243489,
100
+ "row": 37,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 31166629
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 925000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "49b9042fb538cca213af1e6f33ceb5a89185fc6d2bab488be3f60e812871f343", "shard_scheduler.json": "8d8ed3fb8d4759dd81cc3b414f98721e694b600f7f334cec5e3aac1aacaf22d9", "coverage.json": "881381b0771fcad47a6e5995be5e6e0c4fb037348c5449db000759f8eb314ff5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000925000"}
 
1
+ {"step": 930000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "a7562f2130d78c752d5494512bcbf6d4c62627eee4ff6943c08f7615044b2de4", "shard_scheduler.json": "b3c49f97562f447c57be5ea107f2e5123930591bd4d8ad5d22e1610da58f2a24", "coverage.json": "32cef4aca490da53b293a3ee6d9244fadb87af3f5c0e004c874df34f97a8fb94", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000930000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:776f8e7069a5649f67ac5b264d85a0ad1267fdfc3d127253c1c12d96d6406bb2
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63bcc5339acac95d1469b45dc6f3e0ea8fbbde75e8983e3940d28a7da8869414
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.9382242159999805,
53
  "peak_gb": 89.37790203094482,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9382242159999805,
56
- "input_tokens_per_second": 196456.2381323185
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.9440444560000287,
53
  "peak_gb": 89.37790203094482,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9440444560000287,
56
+ "input_tokens_per_second": 195245.04257031978
57
  },
58
  "measurements": [
59
  {
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 76780.75659017764,
26
- "required_pages_per_second_upper_estimate": 0.005311870896808872
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9597.594573772205,
40
- "required_pages_per_second_upper_estimate": 0.00401831242747314
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19195.18914754441,
54
- "required_pages_per_second_upper_estimate": 0.003997851688254453
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28792.783721316613,
68
- "required_pages_per_second_upper_estimate": 0.0017260834200577887
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28792.783721316613,
82
- "required_pages_per_second_upper_estimate": 0.002238512288261146
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 28792.783721316613,
96
- "required_pages_per_second_upper_estimate": 0.36932291429453973
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,74 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.1442483061252915,
162
- "producer_p95_seconds": 0.33357249200344086,
163
- "gpu_update_seconds": 0.9602406029095034,
164
- "producer_batches_per_second": 6.932490417817578,
165
- "gpu_batches_per_second": 1.041405661216602,
166
- "producer_headroom": 6.656858778469507,
167
  "sustainable": true,
168
- "buffer_seconds": 60.49515798329872,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 247072.63852710513,
174
- "upload_bytes_per_second": 5145160.357073876,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 53.01603667125573,
179
  "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
180
  },
181
  "upload_measurements": [
182
  {
183
- "seconds": 1.046063713998592,
184
- "bytes": 5944086,
185
- "pages": 36,
186
- "shards": 2
187
- },
188
- {
189
- "seconds": 1.323775843000476,
190
- "bytes": 7235009,
191
- "pages": 30,
192
- "shards": 3
193
- },
194
- {
195
- "seconds": 1.370845877998363,
196
- "bytes": 11121979,
197
- "pages": 36,
198
- "shards": 3
199
- },
200
- {
201
- "seconds": 1.6951102359998913,
202
- "bytes": 7495029,
203
- "pages": 40,
204
- "shards": 3
205
- },
206
- {
207
- "seconds": 1.1257126100026653,
208
- "bytes": 946737,
209
- "pages": 27,
210
- "shards": 1
211
- },
212
- {
213
- "seconds": 1.3359714920006809,
214
- "bytes": 10736001,
215
- "pages": 31,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 1.2456738490000134,
220
- "bytes": 7708205,
221
- "pages": 43,
222
- "shards": 3
223
- },
224
- {
225
- "seconds": 1.1626391710015014,
226
- "bytes": 1209462,
227
- "pages": 36,
228
- "shards": 1
229
- },
230
- {
231
- "seconds": 1.3864067120011896,
232
- "bytes": 11713877,
233
- "pages": 34,
234
- "shards": 3
235
- },
236
- {
237
- "seconds": 1.506981171998632,
238
- "bytes": 11206977,
239
- "pages": 38,
240
- "shards": 4
241
- },
242
- {
243
- "seconds": 0.8770163509980193,
244
- "bytes": 891965,
245
- "pages": 27,
246
  "shards": 1
247
  },
248
  {
249
- "seconds": 1.4873648249995313,
250
- "bytes": 7797403,
251
- "pages": 39,
252
- "shards": 3
253
- },
254
- {
255
- "seconds": 1.1096691220009234,
256
- "bytes": 969203,
257
- "pages": 31,
258
- "shards": 1
259
- },
260
- {
261
- "seconds": 1.6109700509987306,
262
- "bytes": 10483727,
263
- "pages": 33,
264
- "shards": 3
265
- },
266
- {
267
- "seconds": 1.5793520259976503,
268
- "bytes": 13424823,
269
- "pages": 37,
270
- "shards": 4
271
- },
272
- {
273
- "seconds": 0.9566230939963134,
274
- "bytes": 1183559,
275
- "pages": 35,
276
- "shards": 1
277
- },
278
- {
279
- "seconds": 1.4667487440019613,
280
- "bytes": 7634603,
281
- "pages": 32,
282
- "shards": 3
283
- },
284
- {
285
- "seconds": 1.2097738270022091,
286
- "bytes": 9815120,
287
  "pages": 37,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 1.1665725029961322,
292
- "bytes": 1196873,
293
- "pages": 37,
294
- "shards": 1
295
- },
296
- {
297
- "seconds": 1.1840190020011505,
298
- "bytes": 7418937,
299
  "pages": 31,
300
- "shards": 3
301
- },
302
- {
303
- "seconds": 0.9838692140037892,
304
- "bytes": 6891211,
305
- "pages": 34,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.3423998879952705,
310
- "bytes": 11990725,
311
- "pages": 41,
312
- "shards": 4
313
- },
314
- {
315
- "seconds": 1.0155430449958658,
316
- "bytes": 5875734,
317
- "pages": 28,
318
- "shards": 2
319
- },
320
- {
321
- "seconds": 2.4478168570058187,
322
- "bytes": 1176227,
323
- "pages": 37,
324
- "shards": 1
325
- },
326
- {
327
- "seconds": 1.676982698001666,
328
- "bytes": 6919911,
329
- "pages": 22,
330
- "shards": 3
331
- },
332
- {
333
- "seconds": 1.0402027899981476,
334
- "bytes": 5156086,
335
- "pages": 30,
336
- "shards": 2
337
- },
338
- {
339
- "seconds": 1.5196284359990386,
340
- "bytes": 12048583,
341
- "pages": 36,
342
- "shards": 3
343
- },
344
- {
345
- "seconds": 1.2240972300060093,
346
- "bytes": 7561561,
347
- "pages": 40,
348
- "shards": 3
349
- },
350
- {
351
- "seconds": 0.9214401089993771,
352
- "bytes": 1231310,
353
  "pages": 36,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.6782647159998305,
358
- "bytes": 12065171,
359
- "pages": 33,
360
- "shards": 4
361
- },
362
- {
363
- "seconds": 1.0347107170018717,
364
- "bytes": 6246238,
365
- "pages": 37,
366
- "shards": 2
367
- },
368
- {
369
- "seconds": 1.040108234003128,
370
- "bytes": 901723,
371
- "pages": 27,
372
- "shards": 1
373
- },
374
- {
375
- "seconds": 1.5148740549993818,
376
- "bytes": 14069210,
377
- "pages": 39,
378
- "shards": 4
379
- },
380
- {
381
- "seconds": 1.4405004430009285,
382
- "bytes": 5051876,
383
  "pages": 30,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.163049670001783,
388
- "bytes": 6248444,
389
  "pages": 39,
390
- "shards": 2
391
- },
392
- {
393
- "seconds": 3.389049104996957,
394
- "bytes": 2610225,
395
- "pages": 33,
396
- "shards": 2
397
- },
398
- {
399
- "seconds": 1.3562926899976446,
400
- "bytes": 6176571,
401
- "pages": 34,
402
- "shards": 2
403
- },
404
- {
405
- "seconds": 1.3079960410032072,
406
- "bytes": 12000675,
407
- "pages": 34,
408
- "shards": 4
409
- },
410
- {
411
- "seconds": 1.1082588089993806,
412
- "bytes": 7050197,
413
- "pages": 37,
414
- "shards": 2
415
- },
416
- {
417
- "seconds": 1.0682790589999058,
418
- "bytes": 6087723,
419
- "pages": 38,
420
- "shards": 2
421
- },
422
- {
423
- "seconds": 1.1598256950019277,
424
- "bytes": 7463538,
425
- "pages": 29,
426
- "shards": 3
427
- },
428
- {
429
- "seconds": 1.2493024350042106,
430
- "bytes": 5521564,
431
- "pages": 41,
432
- "shards": 2
433
- },
434
- {
435
- "seconds": 1.7268630670005223,
436
- "bytes": 7581932,
437
- "pages": 31,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 0.9802855540037854,
442
- "bytes": 1121294,
443
- "pages": 35,
444
- "shards": 1
445
- },
446
- {
447
- "seconds": 2.06611223800428,
448
- "bytes": 11832847,
449
- "pages": 31,
450
- "shards": 3
451
- },
452
- {
453
- "seconds": 1.2940340749992174,
454
- "bytes": 12187860,
455
  "pages": 36,
456
- "shards": 4
457
- },
458
- {
459
- "seconds": 1.0110984060011106,
460
- "bytes": 1179043,
461
- "pages": 36,
462
- "shards": 1
463
- },
464
- {
465
- "seconds": 1.417530385995633,
466
- "bytes": 5826314,
467
- "pages": 31,
468
- "shards": 2
469
- },
470
- {
471
- "seconds": 1.3201328729992383,
472
- "bytes": 7683555,
473
- "pages": 38,
474
  "shards": 3
475
  },
476
  {
477
- "seconds": 1.031134179997025,
478
- "bytes": 5438180,
479
- "pages": 33,
480
- "shards": 2
481
- },
482
- {
483
- "seconds": 1.582061741995858,
484
- "bytes": 13195042,
485
- "pages": 34,
486
- "shards": 4
487
- },
488
- {
489
- "seconds": 0.8400354289988172,
490
- "bytes": 1240516,
491
- "pages": 36,
492
- "shards": 1
493
- },
494
- {
495
- "seconds": 1.1406454610041692,
496
- "bytes": 5969910,
497
- "pages": 37,
498
- "shards": 2
499
- },
500
- {
501
- "seconds": 1.6059057809980004,
502
- "bytes": 11827384,
503
- "pages": 24,
504
- "shards": 4
505
- },
506
- {
507
- "seconds": 1.091920383994875,
508
- "bytes": 1379660,
509
- "pages": 39,
510
- "shards": 1
511
- },
512
- {
513
- "seconds": 1.415137545001926,
514
- "bytes": 7393395,
515
  "pages": 35,
516
- "shards": 3
517
- },
518
- {
519
- "seconds": 1.3102802950015757,
520
- "bytes": 11967731,
521
- "pages": 38,
522
- "shards": 3
523
- },
524
- {
525
- "seconds": 0.9184283650029101,
526
- "bytes": 675909,
527
- "pages": 20,
528
- "shards": 1
529
- },
530
- {
531
- "seconds": 1.4080826390054426,
532
- "bytes": 11404109,
533
- "pages": 41,
534
- "shards": 5
535
- },
536
- {
537
- "seconds": 1.4213992930017412,
538
- "bytes": 1225275,
539
- "pages": 36,
540
  "shards": 1
541
- },
542
- {
543
- "seconds": 1.2654285969983903,
544
- "bytes": 6147750,
545
- "pages": 29,
546
- "shards": 2
547
- },
548
- {
549
- "seconds": 1.2639681400032714,
550
- "bytes": 11134689,
551
- "pages": 41,
552
- "shards": 4
553
- },
554
- {
555
- "seconds": 1.0753672879945952,
556
- "bytes": 6680589,
557
- "pages": 29,
558
- "shards": 2
559
- },
560
- {
561
- "seconds": 1.2491716029981035,
562
- "bytes": 7586487,
563
- "pages": 41,
564
- "shards": 3
565
  }
566
  ],
567
  "cache_key": {
@@ -606,7 +270,7 @@
606
  },
607
  "cache_reused": true,
608
  "resume_probe_seconds": [
609
- 2.975958035000019,
610
- 0.01771132999999736
611
  ]
612
  }
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 77309.27550044844,
26
+ "required_pages_per_second_upper_estimate": 0.005348435061354233
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9663.659437556054,
40
+ "required_pages_per_second_upper_estimate": 0.0040459724063482035
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19327.31887511211,
54
+ "required_pages_per_second_upper_estimate": 0.004025370826011566
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28990.97831266816,
68
+ "required_pages_per_second_upper_estimate": 0.0017379648831836957
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28990.97831266816,
82
+ "required_pages_per_second_upper_estimate": 0.0022539210459728526
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
+ "required_token_positions_per_second": 28990.97831266816,
96
+ "required_pages_per_second_upper_estimate": 0.37186514170762514
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.05302272368743388,
162
+ "producer_p95_seconds": 0.1645179819997793,
163
+ "gpu_update_seconds": 0.9536759919522508,
164
+ "producer_batches_per_second": 18.859838394854,
165
+ "gpu_batches_per_second": 1.048574157720926,
166
+ "producer_headroom": 17.98617508927153,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.0815874929918,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 87,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 433303.45547621214,
174
+ "upload_bytes_per_second": 3085286.3142732033,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 92.97683475699209,
179
  "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
180
  },
181
  "upload_measurements": [
182
  {
183
+ "seconds": 1.8764951709999877,
184
+ "bytes": 52123,
185
+ "pages": 1,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.2044020090006597,
190
+ "bytes": 7888698,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
191
  "pages": 37,
192
  "shards": 3
193
  },
194
  {
195
+ "seconds": 1.2931400809993647,
196
+ "bytes": 6805061,
 
 
 
 
 
 
197
  "pages": 31,
 
 
 
 
 
 
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 1.581821284000398,
202
+ "bytes": 6273764,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
203
  "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 0.9842661719994794,
208
+ "bytes": 2388376,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
209
  "pages": 30,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 4.260054792999654,
214
+ "bytes": 10427634,
215
  "pages": 39,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.4442174710002291,
220
+ "bytes": 7300139,
 
 
 
 
 
 
 
 
 
 
 
 
221
  "pages": 36,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.0513844460001565,
226
+ "bytes": 1119612,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
227
  "pages": 35,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
228
  "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
229
  }
230
  ],
231
  "cache_key": {
 
270
  },
271
  "cache_reused": true,
272
  "resume_probe_seconds": [
273
+ 3.415684295999995,
274
+ 0.013631696000004467
275
  ]
276
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f11cd6ddab4cc1fb62ed242dcbc5a9cf8568c346ec14615dd18e989af69c0520
3
- size 721778943
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c18f555ae584e5131ef117587a3babd2ee418350f9f2a2de452e8445fba3bf56
3
+ size 721765823