appvoid commited on
Commit
5607f27
·
verified ·
1 Parent(s): 58fce14

Qwen90M step 560000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +49 -49
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +4 -4
  6. shard_scheduler.json +98 -332
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 557164,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "9aeb84d48b1a3e7876ab13cf2a9db637ae3eea5a8e14c759a56dfd289a32ce86",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "625680f99c72ef12793fea5968b6c72147fca5bf0986649c268a6a0699e0b367",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
- "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
- "shard_scheduler.json": "21879a49824847fe89e795ae96e1fd2dc476cf1541cf6501892fb7519e8c9036",
15
- "coverage.json": "73df3c9242608e741c41ad8c77453ef20d1273e9c95e5fe1751e37fda2234e71",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 560000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "9fc6ec90f620abfbf95bf8c6ae75a133a76b595b55c08435360d181ac4591f3c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "2e8178c1d4313006341bdad1034631bc9d09a44086d3080d7bd8e8194fb48d2c",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
+ "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
+ "shard_scheduler.json": "cd21f72b325db7b3e64e3a7aac1563607bb2e1d101286d5664c18896e97ea8da",
15
+ "coverage.json": "35cce35db77290dc46d2183e76ef09b89428490347ea9b14cdb657a8652a8114",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 557164,
3
  "stats": {
4
- "climbmix/blocks": 20057904,
5
- "climbmix/targets": 33528559976,
6
- "climbmix/input_tokens": 33530684873,
7
- "ultra_qa/blocks": 7521714,
8
- "ultra_qa/targets": 12270782181,
9
- "ultra_qa/input_tokens": 12270805843,
10
- "cortex/blocks": 7521714,
11
- "cortex/targets": 286034884,
12
- "cortex/input_tokens": 11956373662,
13
- "ultra_style/blocks": 7521714,
14
- "ultra_style/targets": 11896892314,
15
- "ultra_style/input_tokens": 11897516222,
16
- "openmath/blocks": 5014476,
17
- "openmath/targets": 7748188395,
18
- "openmath/input_tokens": 7750768917,
19
- "rewrite6/blocks": 2507238,
20
- "rewrite6/targets": 4110276401,
21
- "rewrite6/input_tokens": 4114782547
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2747,
26
- "row": 2075,
27
  "start": {
28
- "file": 132,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 77,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 11253787
38
  },
39
  "rewrite6": {
40
- "page": 12,
41
- "row": 3016,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 49152
47
  }
48
  },
49
  "epoch": 31,
50
  "offset": 0,
51
- "documents": 6419692
52
  },
53
  "openmath": {
54
- "page": 1716,
55
- "row": 2073,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 42,
61
- "row": 336
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 7030809
67
  },
68
  "ultra_qa": {
69
- "page": 742,
70
- "row": 1162,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 6,
76
- "row": 53994
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 3040394
82
  },
83
  "ultra_style": {
84
- "page": 1132,
85
- "row": 4042,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
- "row": 15848
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 4640714
97
  },
98
  "cortex": {
99
- "page": 145876,
100
- "row": 97,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 18672225
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 560000,
3
  "stats": {
4
+ "climbmix/blocks": 20160000,
5
+ "climbmix/targets": 33699449931,
6
+ "climbmix/input_tokens": 33701585663,
7
+ "ultra_qa/blocks": 7560000,
8
+ "ultra_qa/targets": 12333281836,
9
+ "ultra_qa/input_tokens": 12333305632,
10
+ "cortex/blocks": 7560000,
11
+ "cortex/targets": 287490188,
12
+ "cortex/input_tokens": 12017335871,
13
+ "ultra_style/blocks": 7560000,
14
+ "ultra_style/targets": 11957074205,
15
+ "ultra_style/input_tokens": 11957701334,
16
+ "openmath/blocks": 5040000,
17
+ "openmath/targets": 7787661032,
18
+ "openmath/input_tokens": 7790254729,
19
+ "rewrite6/blocks": 2520000,
20
+ "rewrite6/targets": 4131150735,
21
+ "rewrite6/input_tokens": 4135680080
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2761,
26
+ "row": 1877,
27
  "start": {
28
+ "file": 133,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 50,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 4096,
37
+ "documents": 11310933
38
  },
39
  "rewrite6": {
40
+ "page": 20,
41
+ "row": 3266,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 81920
47
  }
48
  },
49
  "epoch": 31,
50
  "offset": 0,
51
+ "documents": 6452710
52
  },
53
  "openmath": {
54
+ "page": 1725,
55
+ "row": 1041,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 79,
61
+ "row": 200
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 7066641
67
  },
68
  "ultra_qa": {
69
+ "page": 746,
70
+ "row": 300,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 7,
76
+ "row": 5982
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 3055916
82
  },
83
  "ultra_style": {
84
+ "page": 1138,
85
+ "row": 3363,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
+ "row": 40424
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 4096,
96
+ "documents": 4664611
97
  },
98
  "cortex": {
99
+ "page": 146621,
100
+ "row": 50,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 18767538
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 557164, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "9aeb84d48b1a3e7876ab13cf2a9db637ae3eea5a8e14c759a56dfd289a32ce86", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "625680f99c72ef12793fea5968b6c72147fca5bf0986649c268a6a0699e0b367", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "21879a49824847fe89e795ae96e1fd2dc476cf1541cf6501892fb7519e8c9036", "coverage.json": "73df3c9242608e741c41ad8c77453ef20d1273e9c95e5fe1751e37fda2234e71", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000557164"}
 
1
+ {"step": 560000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "9fc6ec90f620abfbf95bf8c6ae75a133a76b595b55c08435360d181ac4591f3c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2e8178c1d4313006341bdad1034631bc9d09a44086d3080d7bd8e8194fb48d2c", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "cd21f72b325db7b3e64e3a7aac1563607bb2e1d101286d5664c18896e97ea8da", "coverage.json": "35cce35db77290dc46d2183e76ef09b89428490347ea9b14cdb657a8652a8114", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000560000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9aeb84d48b1a3e7876ab13cf2a9db637ae3eea5a8e14c759a56dfd289a32ce86
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9fc6ec90f620abfbf95bf8c6ae75a133a76b595b55c08435360d181ac4591f3c
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.9447870520000379,
53
- "peak_gb": 89.37766551971436,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9447870520000379,
56
- "input_tokens_per_second": 195091.58133550777
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 90,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.9371402589999889,
53
+ "peak_gb": 89.37790203094482,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9371402589999889,
56
+ "input_tokens_per_second": 196683.47211620773
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": 76613.12993076043,
26
- "required_pages_per_second_upper_estimate": 0.005300274095562953
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": 9576.641241345054,
40
- "required_pages_per_second_upper_estimate": 0.004009539704741229
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": 19153.28248269011,
54
- "required_pages_per_second_upper_estimate": 0.003989123635118372
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": 28729.923724035165,
68
- "required_pages_per_second_upper_estimate": 0.0017223150592024243
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": 28729.923724035165,
82
- "required_pages_per_second_upper_estimate": 0.002233625199964421
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": 28729.923724035165,
96
- "required_pages_per_second_upper_estimate": 0.36851661374321987
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,176 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.09595589186756115,
162
- "producer_p95_seconds": 0.29021746900252765,
163
- "gpu_update_seconds": 0.9623415733912987,
164
- "producer_batches_per_second": 10.421454905345525,
165
- "gpu_batches_per_second": 1.0391320791390033,
166
- "producer_headroom": 10.028999310636681,
167
  "sustainable": true,
168
- "buffer_seconds": 60.627519123651815,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 370915.98441075865,
174
- "upload_bytes_per_second": 5961453.864962328,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 79.58993577234335,
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": 1.272125319999759,
184
- "bytes": 7866146,
185
- "pages": 39,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.004612441000063,
190
- "bytes": 5431394,
191
- "pages": 26,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 1.1393026930018095,
196
- "bytes": 12031348,
197
- "pages": 42,
198
- "shards": 3
199
- },
200
- {
201
- "seconds": 1.1874672249978175,
202
- "bytes": 7293811,
203
- "pages": 30,
204
- "shards": 3
205
- },
206
- {
207
- "seconds": 0.9514401040032681,
208
- "bytes": 1189191,
209
- "pages": 36,
210
  "shards": 1
211
  },
212
  {
213
- "seconds": 1.3702037240000209,
214
- "bytes": 10453891,
215
- "pages": 36,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 0.9155222249974031,
220
- "bytes": 2430144,
221
- "pages": 31,
222
- "shards": 2
223
- },
224
- {
225
- "seconds": 1.411119290998613,
226
- "bytes": 5478989,
227
- "pages": 19,
228
- "shards": 3
229
- },
230
- {
231
- "seconds": 0.9785845789992891,
232
- "bytes": 1201136,
233
- "pages": 33,
234
  "shards": 1
235
  },
236
  {
237
- "seconds": 1.2566547139977047,
238
- "bytes": 13157311,
239
- "pages": 30,
240
- "shards": 4
241
- },
242
- {
243
- "seconds": 1.2061452220004867,
244
- "bytes": 11040131,
245
- "pages": 39,
246
- "shards": 3
247
- },
248
- {
249
- "seconds": 0.9862021689987159,
250
- "bytes": 2654656,
251
- "pages": 38,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 0.9963805859988497,
256
- "bytes": 5942893,
257
- "pages": 28,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 0.8844175519989221,
262
- "bytes": 1183171,
263
- "pages": 35,
264
- "shards": 1
265
- },
266
- {
267
- "seconds": 1.7914613989996724,
268
- "bytes": 11162167,
269
- "pages": 33,
270
- "shards": 4
271
- },
272
- {
273
- "seconds": 1.1609297110007901,
274
- "bytes": 12171791,
275
- "pages": 35,
276
- "shards": 3
277
- },
278
- {
279
- "seconds": 0.9642307790018094,
280
- "bytes": 2614472,
281
- "pages": 37,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.2208152999992308,
286
- "bytes": 5885163,
287
- "pages": 35,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.0881164640013594,
292
- "bytes": 9917140,
293
- "pages": 33,
294
- "shards": 3
295
- },
296
- {
297
- "seconds": 0.9187301879974257,
298
- "bytes": 2590562,
299
- "pages": 37,
300
  "shards": 2
301
  },
302
  {
303
- "seconds": 1.3539272460002394,
304
- "bytes": 6288510,
305
- "pages": 31,
306
- "shards": 2
307
  },
308
  {
309
- "seconds": 1.2604586979978194,
310
- "bytes": 6693523,
311
- "pages": 28,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.1304304789991875,
316
- "bytes": 12005280,
317
- "pages": 46,
318
- "shards": 4
319
- },
320
- {
321
- "seconds": 1.0543503480002983,
322
- "bytes": 5804640,
323
- "pages": 28,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 0.8439256270030455,
328
- "bytes": 2654364,
329
- "pages": 37,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 0.9141899559981539,
334
- "bytes": 5548542,
335
- "pages": 23,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.0629937340017932,
340
- "bytes": 10277378,
341
- "pages": 44,
342
- "shards": 3
343
- },
344
- {
345
- "seconds": 1.4301319679980224,
346
- "bytes": 8501849,
347
- "pages": 41,
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 0.9612259829991672,
352
- "bytes": 6294173,
353
- "pages": 28,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 1.084874779000529,
358
- "bytes": 7316912,
359
- "pages": 38,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 1.2164157670013083,
364
- "bytes": 5692225,
365
- "pages": 32,
366
- "shards": 2
367
- },
368
- {
369
- "seconds": 1.0190912239995669,
370
- "bytes": 6330148,
371
  "pages": 38,
372
- "shards": 2
373
- },
374
- {
375
- "seconds": 1.0534845100009989,
376
- "bytes": 2332089,
377
- "pages": 28,
378
- "shards": 2
379
- },
380
- {
381
- "seconds": 1.458769362001476,
382
- "bytes": 16198178,
383
- "pages": 39,
384
  "shards": 4
385
  },
386
  {
387
- "seconds": 1.0830437309996341,
388
- "bytes": 7432031,
389
  "pages": 34,
390
- "shards": 3
391
- },
392
- {
393
- "seconds": 1.1578626439986692,
394
- "bytes": 1196629,
395
- "pages": 36,
396
  "shards": 1
397
  },
398
  {
399
- "seconds": 0.9139833829976851,
400
- "bytes": 5815740,
401
- "pages": 28,
402
- "shards": 2
403
- },
404
- {
405
- "seconds": 1.1043080169984023,
406
- "bytes": 6797860,
407
- "pages": 32,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.0245990330004133,
412
- "bytes": 6034884,
413
  "pages": 37,
414
- "shards": 3
415
- },
416
- {
417
- "seconds": 1.1096401040013006,
418
- "bytes": 11998781,
419
- "pages": 38,
420
- "shards": 3
421
- },
422
- {
423
- "seconds": 0.9207302890026767,
424
- "bytes": 2353345,
425
- "pages": 29,
426
- "shards": 2
427
- },
428
- {
429
- "seconds": 0.9312007079970499,
430
- "bytes": 5937086,
431
- "pages": 25,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 0.9517514979997941,
436
- "bytes": 6995458,
437
- "pages": 46,
438
- "shards": 3
439
- },
440
- {
441
- "seconds": 1.261231005002628,
442
- "bytes": 5881548,
443
- "pages": 28,
444
- "shards": 2
445
- },
446
- {
447
- "seconds": 0.9858025629982876,
448
- "bytes": 6530607,
449
  "pages": 37,
450
- "shards": 2
451
- },
452
- {
453
- "seconds": 1.5898211400017317,
454
- "bytes": 12596102,
455
- "pages": 36,
456
- "shards": 4
457
- },
458
- {
459
- "seconds": 0.9551758580018941,
460
- "bytes": 5775284,
461
- "pages": 29,
462
- "shards": 2
463
- },
464
- {
465
- "seconds": 1.2868636950006476,
466
- "bytes": 7532566,
467
- "pages": 40,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 0.9102224750022287,
472
- "bytes": 989956,
473
- "pages": 26,
474
- "shards": 1
475
- },
476
- {
477
- "seconds": 1.172385066998686,
478
- "bytes": 10249964,
479
- "pages": 40,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.0767950209992705,
484
- "bytes": 2405647,
485
- "pages": 29,
486
- "shards": 2
487
- },
488
- {
489
- "seconds": 1.3306630199986103,
490
- "bytes": 11933752,
491
- "pages": 33,
492
- "shards": 3
493
  },
494
  {
495
- "seconds": 1.1605643660004716,
496
- "bytes": 6187685,
497
  "pages": 38,
498
- "shards": 2
499
- },
500
- {
501
- "seconds": 1.1438976530007494,
502
- "bytes": 6974681,
503
- "pages": 33,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 0.9301952720015834,
508
- "bytes": 6018270,
509
- "pages": 37,
510
- "shards": 2
511
- },
512
- {
513
- "seconds": 1.0365287800013903,
514
- "bytes": 2455252,
515
- "pages": 29,
516
- "shards": 2
517
- },
518
- {
519
- "seconds": 1.3521412089976366,
520
- "bytes": 6109192,
521
- "pages": 37,
522
- "shards": 2
523
- },
524
- {
525
- "seconds": 1.5889534400012053,
526
- "bytes": 16077810,
527
- "pages": 39,
528
- "shards": 4
529
- },
530
- {
531
- "seconds": 0.8900750899993,
532
- "bytes": 2373140,
533
- "pages": 29,
534
- "shards": 2
535
- },
536
- {
537
- "seconds": 1.2026867530003074,
538
- "bytes": 6250314,
539
  "pages": 37,
540
- "shards": 2
541
- },
542
- {
543
- "seconds": 1.0975493779988028,
544
- "bytes": 5901759,
545
- "pages": 32,
546
- "shards": 2
547
- },
548
- {
549
- "seconds": 1.0364178640011232,
550
- "bytes": 6898240,
551
- "pages": 40,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.1201825170028314,
556
- "bytes": 6019295,
557
- "pages": 28,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.523898775998532,
562
- "bytes": 8240555,
563
- "pages": 36,
564
- "shards": 3
565
  }
566
  ],
567
  "cache_key": {
@@ -606,7 +372,7 @@
606
  },
607
  "cache_reused": true,
608
  "resume_probe_seconds": [
609
- 1.0657202160000452,
610
- 0.013942628999984663
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": 76406.2405888879,
26
+ "required_pages_per_second_upper_estimate": 0.0052859610108428
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": 9550.780073610988,
40
+ "required_pages_per_second_upper_estimate": 0.003998712173853568
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": 19101.560147221975,
54
+ "required_pages_per_second_upper_estimate": 0.003978351236650072
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": 28652.340220832964,
68
+ "required_pages_per_second_upper_estimate": 0.0017176640466486026
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": 28652.340220832964,
82
+ "required_pages_per_second_upper_estimate": 0.0022275934238442157
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": 28652.340220832964,
96
+ "required_pages_per_second_upper_estimate": 0.36752145586681756
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": 704566,
161
+ "producer_batch_seconds": 0.16272236082813407,
162
+ "producer_p95_seconds": 0.14889653499994893,
163
+ "gpu_update_seconds": 0.9649473581183182,
164
+ "producer_batches_per_second": 6.1454368957699135,
165
+ "gpu_batches_per_second": 1.0363259628484145,
166
+ "producer_headroom": 5.930023097056017,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.79168356145405,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 422396.62886957865,
174
+ "upload_bytes_per_second": 5499815.520343313,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 90.63648366513748,
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": 1.022603193000009,
184
+ "bytes": 47630,
185
+ "pages": 1,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 0.8258607079999933,
190
+ "bytes": 145464,
191
+ "pages": 4,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
192
  "shards": 1
193
  },
194
  {
195
+ "seconds": 1.1176107269999989,
196
+ "bytes": 5541224,
197
+ "pages": 18,
 
 
 
 
 
 
 
 
 
 
 
 
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 1.2054241769999408,
202
+ "bytes": 2385611,
203
+ "pages": 27,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 0.8888624019999725,
208
+ "bytes": 6324210,
209
+ "pages": 34,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.1897545489999857,
214
+ "bytes": 1226596,
215
+ "pages": 34,
 
 
 
 
 
 
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.2994146069999033,
220
+ "bytes": 17142614,
221
+ "pages": 34,
222
+ "shards": 5
223
  },
224
  {
225
+ "seconds": 1.1055204159999903,
226
+ "bytes": 6016330,
227
+ "pages": 35,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 0.9128079020001678,
232
+ "bytes": 2402677,
233
+ "pages": 27,
 
 
 
 
 
 
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.0666157699999985,
238
+ "bytes": 6099291,
239
+ "pages": 35,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.0668149449998054,
244
+ "bytes": 5027383,
245
+ "pages": 32,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.0457134279999991,
250
+ "bytes": 7722680,
251
+ "pages": 38,
 
 
 
 
 
 
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.1198342459999822,
256
+ "bytes": 518226,
257
+ "pages": 18,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.3007597939999869,
262
+ "bytes": 12269409,
263
+ "pages": 44,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.437565898999992,
268
+ "bytes": 12411381,
 
 
 
 
 
 
269
  "pages": 38,
 
 
 
 
 
 
 
 
 
 
 
 
270
  "shards": 4
271
  },
272
  {
273
+ "seconds": 0.765915884000151,
274
+ "bytes": 1219645,
275
  "pages": 34,
 
 
 
 
 
 
276
  "shards": 1
277
  },
278
  {
279
+ "seconds": 1.470795656000064,
280
+ "bytes": 7604342,
281
+ "pages": 30,
 
 
 
 
 
 
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 0.9404902840001341,
286
+ "bytes": 6153110,
287
  "pages": 37,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0696395039999516,
292
+ "bytes": 10812914,
 
 
 
 
 
 
 
 
 
 
 
 
293
  "pages": 37,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.5107661739998548,
298
+ "bytes": 7696973,
299
+ "pages": 34,
 
 
 
 
 
 
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 0.9512165999999525,
304
+ "bytes": 1125687,
305
+ "pages": 30,
306
+ "shards": 1
 
 
 
 
 
 
307
  },
308
  {
309
+ "seconds": 1.0864234489999944,
310
+ "bytes": 7369290,
311
  "pages": 38,
 
 
 
 
 
 
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 1.1581361090002247,
316
+ "bytes": 10067921,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
317
  "pages": 37,
 
 
 
 
 
 
 
 
 
 
 
 
318
  "shards": 3
319
  },
320
  {
321
+ "seconds": 0.7976953110000977,
322
+ "bytes": 1131427,
323
+ "pages": 35,
324
+ "shards": 1
325
  },
326
  {
327
+ "seconds": 1.1826321499997903,
328
+ "bytes": 12996691,
329
+ "pages": 33,
330
+ "shards": 4
331
  }
332
  ],
333
  "cache_key": {
 
372
  },
373
  "cache_reused": true,
374
  "resume_probe_seconds": [
375
+ 2.6133196560000442,
376
+ 0.09247372499999074
377
  ]
378
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:625680f99c72ef12793fea5968b6c72147fca5bf0986649c268a6a0699e0b367
3
- size 721775807
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e8178c1d4313006341bdad1034631bc9d09a44086d3080d7bd8e8194fb48d2c
3
+ size 721753727