appvoid commited on
Commit
f5643b9
·
verified ·
1 Parent(s): f78c9a5

Qwen90M step 785000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +49 -49
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +234 -234
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 780000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "861f90278f68313b09eb4b302450aa4b5f8e1fe53d075de34cb4897b13546b54",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "bb7ceaeda93bfb3b23d8a7a9fc6b379ef73c40abcbb8d58fb638329c569f90e7",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d",
14
- "shard_scheduler.json": "060f57d01a3356f0f8395d272cc24eb06656aeba401c5dcbc708fe202e44878a",
15
- "coverage.json": "f641f693c6260b4b353bf8683e17a82b7b292ad28deab6832a52269599d71b92",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 785000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "9e9efa160bc5d9cdcfdc2ca5084b0841b10448f0013b892e4c614e0c0ff5502f",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "da99d6bb0dd16c05dc0dcdcb4b5f86f1c1ca2eda0fbea8eec1ea8b1c34c0fc81",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d",
14
+ "shard_scheduler.json": "70243b4426905888cedd0ae2bc3df6ea021387e3c4d1a15ff1b998d804d4625b",
15
+ "coverage.json": "2153e069f9b927c191b3d524d2beb884e93a0e0cf00c2acbc8c4eeeff4ac0602",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 780000,
3
  "stats": {
4
- "climbmix/blocks": 28080000,
5
- "climbmix/targets": 46940530276,
6
- "climbmix/input_tokens": 46943505263,
7
- "ultra_qa/blocks": 10530000,
8
- "ultra_qa/targets": 17178444425,
9
- "ultra_qa/input_tokens": 17178477489,
10
- "cortex/blocks": 10530000,
11
- "cortex/targets": 400402885,
12
- "cortex/input_tokens": 16738077716,
13
- "ultra_style/blocks": 10530000,
14
- "ultra_style/targets": 16655737210,
15
- "ultra_style/input_tokens": 16656610869,
16
- "openmath/blocks": 7020000,
17
- "openmath/targets": 10847110567,
18
- "openmath/input_tokens": 10850724850,
19
- "rewrite6/blocks": 3510000,
20
- "rewrite6/targets": 5754125788,
21
- "rewrite6/input_tokens": 5760434426
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3845,
26
- "row": 3683,
27
  "start": {
28
- "file": 186,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 10,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 4096,
37
- "documents": 15752803
38
  },
39
  "rewrite6": {
40
- "page": 37,
41
- "row": 3740,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 151552
47
  }
48
  },
49
- "epoch": 43,
50
- "offset": 2048,
51
- "documents": 8987664
52
  },
53
  "openmath": {
54
- "page": 2403,
55
- "row": 2100,
56
  "start": {
57
  "file": 22,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 236,
61
- "row": 388
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 9844788
67
  },
68
  "ultra_qa": {
69
- "page": 1039,
70
- "row": 799,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 1,
76
- "row": 34212
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 4256543
82
  },
83
  "ultra_style": {
84
- "page": 1586,
85
- "row": 561,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 3,
91
- "row": 77044
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 6496817
97
  },
98
  "cortex": {
99
- "page": 204216,
100
- "row": 101,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 26139749
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 785000,
3
  "stats": {
4
+ "climbmix/blocks": 28260000,
5
+ "climbmix/targets": 47241568234,
6
+ "climbmix/input_tokens": 47244562296,
7
+ "ultra_qa/blocks": 10597500,
8
+ "ultra_qa/targets": 17288525662,
9
+ "ultra_qa/input_tokens": 17288558958,
10
+ "cortex/blocks": 10597500,
11
+ "cortex/targets": 402969786,
12
+ "cortex/input_tokens": 16845487108,
13
+ "ultra_style/blocks": 10597500,
14
+ "ultra_style/targets": 16762084445,
15
+ "ultra_style/input_tokens": 16762964029,
16
+ "openmath/blocks": 7065000,
17
+ "openmath/targets": 10916644933,
18
+ "openmath/input_tokens": 10920282460,
19
+ "rewrite6/blocks": 3532500,
20
+ "rewrite6/targets": 5791040560,
21
+ "rewrite6/input_tokens": 5797389432
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3870,
26
+ "row": 2266,
27
  "start": {
28
+ "file": 187,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 27,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 15853786
38
  },
39
  "rewrite6": {
40
+ "page": 1,
41
+ "row": 3122,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 4096
47
  }
48
  },
49
+ "epoch": 44,
50
+ "offset": 0,
51
+ "documents": 9044994
52
  },
53
  "openmath": {
54
+ "page": 2418,
55
+ "row": 3723,
56
  "start": {
57
  "file": 22,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 297,
61
+ "row": 828
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 9907851
67
  },
68
  "ultra_qa": {
69
+ "page": 1045,
70
+ "row": 3516,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 1,
76
+ "row": 58788
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 4283836
82
  },
83
  "ultra_style": {
84
+ "page": 1596,
85
+ "row": 2099,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 4,
91
+ "row": 32666
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 6539315
97
  },
98
  "cortex": {
99
+ "page": 205528,
100
+ "row": 6,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 26307590
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 780000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "861f90278f68313b09eb4b302450aa4b5f8e1fe53d075de34cb4897b13546b54", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bb7ceaeda93bfb3b23d8a7a9fc6b379ef73c40abcbb8d58fb638329c569f90e7", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "060f57d01a3356f0f8395d272cc24eb06656aeba401c5dcbc708fe202e44878a", "coverage.json": "f641f693c6260b4b353bf8683e17a82b7b292ad28deab6832a52269599d71b92", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000780000"}
 
1
+ {"step": 785000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "9e9efa160bc5d9cdcfdc2ca5084b0841b10448f0013b892e4c614e0c0ff5502f", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "da99d6bb0dd16c05dc0dcdcb4b5f86f1c1ca2eda0fbea8eec1ea8b1c34c0fc81", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "70243b4426905888cedd0ae2bc3df6ea021387e3c4d1a15ff1b998d804d4625b", "coverage.json": "2153e069f9b927c191b3d524d2beb884e93a0e0cf00c2acbc8c4eeeff4ac0602", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000785000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:861f90278f68313b09eb4b302450aa4b5f8e1fe53d075de34cb4897b13546b54
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e9efa160bc5d9cdcfdc2ca5084b0841b10448f0013b892e4c614e0c0ff5502f
3
  size 360622744
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": 77198.30405065721,
26
- "required_pages_per_second_upper_estimate": 0.005340757798968437
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": 9649.788006332152,
40
- "required_pages_per_second_upper_estimate": 0.0040401647277631784
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": 19299.576012664304,
54
- "required_pages_per_second_upper_estimate": 0.004019592719392171
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": 28949.364018996454,
68
- "required_pages_per_second_upper_estimate": 0.001735470169819424
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": 28949.364018996454,
82
- "required_pages_per_second_upper_estimate": 0.0022506857176818108
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": 28949.364018996454,
96
- "required_pages_per_second_upper_estimate": 0.37133135823035174
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.39113858623451847,
162
- "producer_p95_seconds": 0.258143648999976,
163
- "gpu_update_seconds": 0.9550468874500143,
164
- "producer_batches_per_second": 2.5566385807827743,
165
- "gpu_batches_per_second": 1.047069011103749,
166
- "producer_headroom": 2.4417097189112105,
167
  "sustainable": true,
168
- "buffer_seconds": 60.1679539093509,
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": 310654.2999911742,
174
- "upload_bytes_per_second": 4935605.042243135,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 66.65918111611767,
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": 2.3685029429998394,
184
- "bytes": 11410659,
185
- "pages": 34,
186
  "shards": 4
187
  },
188
  {
189
- "seconds": 0.8315555410008528,
190
- "bytes": 1017212,
191
- "pages": 30,
192
- "shards": 1
193
  },
194
  {
195
- "seconds": 1.2100102580006933,
196
- "bytes": 5596380,
197
- "pages": 21,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 1.2646026780003012,
202
- "bytes": 13750693,
203
- "pages": 48,
204
- "shards": 4
205
  },
206
  {
207
- "seconds": 1.3428173240008618,
208
- "bytes": 5558514,
209
- "pages": 31,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.273476110000047,
214
- "bytes": 7479055,
215
- "pages": 38,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 1.084104961999401,
220
- "bytes": 947822,
221
- "pages": 26,
222
- "shards": 1
223
  },
224
  {
225
- "seconds": 1.1121349749992078,
226
- "bytes": 6275461,
227
  "pages": 37,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.2975029250010266,
232
- "bytes": 6264701,
233
- "pages": 26,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.4136529820007127,
238
- "bytes": 11858220,
239
  "pages": 37,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 1.0470604990005086,
244
- "bytes": 6246532,
245
- "pages": 37,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.0089250809996884,
250
- "bytes": 2580280,
251
- "pages": 37,
252
- "shards": 2
253
  },
254
  {
255
- "seconds": 1.1785594149987446,
256
- "bytes": 10228291,
257
- "pages": 34,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 0.9172461540001677,
262
- "bytes": 2354642,
263
- "pages": 28,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.1485085579988663,
268
- "bytes": 12167168,
269
- "pages": 40,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 1.0151017729986052,
274
- "bytes": 6037243,
275
- "pages": 35,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.2390309599995817,
280
- "bytes": 6886865,
281
- "pages": 35,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.3525706070013257,
286
- "bytes": 5675921,
287
  "pages": 28,
288
- "shards": 2
289
  },
290
  {
291
- "seconds": 1.2551854249995813,
292
- "bytes": 7559192,
293
  "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 0.825367109000581,
298
- "bytes": 1141341,
299
- "pages": 33,
300
- "shards": 1
301
  },
302
  {
303
- "seconds": 1.5828043500005151,
304
- "bytes": 15814359,
305
  "pages": 33,
306
- "shards": 4
307
- },
308
- {
309
- "seconds": 1.3585308149995399,
310
- "bytes": 7486305,
311
- "pages": 37,
312
  "shards": 3
313
  },
314
  {
315
- "seconds": 0.8041142309994029,
316
- "bytes": 1148050,
317
- "pages": 34,
318
- "shards": 1
319
- },
320
- {
321
- "seconds": 1.0195367809992604,
322
- "bytes": 6121053,
323
- "pages": 30,
324
- "shards": 2
325
- },
326
- {
327
- "seconds": 2.6576393979994464,
328
- "bytes": 1839915,
329
- "pages": 12,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.339241851999759,
334
- "bytes": 10779600,
335
- "pages": 40,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 4.318171066001014,
340
- "bytes": 1094935,
341
- "pages": 34,
342
- "shards": 1
343
  },
344
  {
345
- "seconds": 1.3115034270012984,
346
- "bytes": 13301880,
347
  "pages": 34,
348
- "shards": 5
349
  },
350
  {
351
- "seconds": 1.1869826189995365,
352
- "bytes": 6445004,
353
  "pages": 37,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.1019184790002328,
358
- "bytes": 6210962,
359
- "pages": 37,
360
- "shards": 3
361
- },
362
- {
363
- "seconds": 1.991184726000938,
364
- "bytes": 5824172,
365
  "pages": 28,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.041803035999692,
370
- "bytes": 1241755,
371
- "pages": 35,
372
  "shards": 1
373
  },
374
  {
375
- "seconds": 1.3904252129996166,
376
- "bytes": 12223668,
377
- "pages": 40,
378
  "shards": 4
379
  },
380
  {
381
- "seconds": 1.4052468339996267,
382
- "bytes": 11659745,
383
  "pages": 33,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.587534140999196,
388
- "bytes": 2763901,
389
- "pages": 37,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.0879035409998323,
394
- "bytes": 6026132,
395
- "pages": 28,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 0.9889250640007958,
400
- "bytes": 952891,
401
- "pages": 29,
402
- "shards": 1
403
  },
404
  {
405
- "seconds": 1.1961971269993228,
406
- "bytes": 12036916,
407
- "pages": 46,
408
- "shards": 4
409
  },
410
  {
411
- "seconds": 1.1770325620018411,
412
- "bytes": 6029556,
413
- "pages": 30,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.1039014870002575,
418
- "bytes": 6886174,
419
- "pages": 32,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 1.8041386129989405,
424
- "bytes": 11898323,
425
- "pages": 40,
426
- "shards": 4
427
  },
428
  {
429
- "seconds": 1.2189035779992992,
430
- "bytes": 6954573,
431
- "pages": 37,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.5599068500014255,
436
- "bytes": 2415324,
437
- "pages": 28,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.996803005997208,
442
- "bytes": 6353091,
443
- "pages": 36,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.149405411000771,
448
- "bytes": 4840926,
449
- "pages": 32,
450
- "shards": 2
451
  },
452
  {
453
- "seconds": 1.5049040789999708,
454
- "bytes": 13497203,
455
- "pages": 34,
456
  "shards": 4
457
  },
458
  {
459
- "seconds": 2.2471259019985155,
460
- "bytes": 6049971,
461
- "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.1292916710008285,
466
- "bytes": 2686677,
467
- "pages": 37,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.3082805390004069,
472
- "bytes": 10490550,
473
- "pages": 32,
474
  "shards": 3
475
  },
476
  {
477
- "seconds": 1.171707494999282,
478
- "bytes": 6154958,
479
  "pages": 37,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.3786574580008164,
484
- "bytes": 2401890,
485
  "pages": 28,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.469864372000302,
490
- "bytes": 11763636,
 
 
 
 
 
 
491
  "pages": 36,
492
- "shards": 3
493
  },
494
  {
495
- "seconds": 1.3772630089988525,
496
- "bytes": 6008251,
497
  "pages": 31,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 1.3044562169998244,
502
- "bytes": 6256652,
503
- "pages": 37,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 1.3504138789976423,
508
- "bytes": 6609036,
 
 
 
 
 
 
509
  "pages": 37,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 0.9880936410008871,
514
- "bytes": 2162409,
515
- "pages": 24,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.2899976480002806,
520
- "bytes": 11585751,
521
- "pages": 45,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.4168642419972457,
526
- "bytes": 6615057,
527
- "pages": 28,
528
  "shards": 3
529
  },
530
  {
531
- "seconds": 1.1043323689991666,
532
- "bytes": 7549149,
533
  "pages": 38,
534
- "shards": 3
535
  },
536
  {
537
- "seconds": 1.1122719240011065,
538
- "bytes": 6287535,
539
- "pages": 37,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.1806445659967721,
544
- "bytes": 6303271,
545
- "pages": 33,
 
 
 
 
 
 
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.2656237250012055,
550
- "bytes": 5785840,
551
- "pages": 28,
552
- "shards": 2
553
  },
554
  {
555
- "seconds": 1.1340223679981136,
556
- "bytes": 1097457,
557
- "pages": 35,
558
  "shards": 1
559
  },
560
  {
561
- "seconds": 3.1717748339979153,
562
- "bytes": 13032899,
563
- "pages": 32,
564
- "shards": 4
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
 
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": 77070.43158003519,
26
+ "required_pages_per_second_upper_estimate": 0.0053319112847457955
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": 9633.803947504399,
40
+ "required_pages_per_second_upper_estimate": 0.004033472536117103
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": 19267.607895008798,
54
+ "required_pages_per_second_upper_estimate": 0.004012934603540484
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": 28901.4118425132,
68
+ "required_pages_per_second_upper_estimate": 0.001732595510057987
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": 28901.4118425132,
82
+ "required_pages_per_second_upper_estimate": 0.0022469576468795716
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": 28901.4118425132,
96
+ "required_pages_per_second_upper_estimate": 0.37071627919746025
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": 0.09094359758591963,
162
+ "producer_p95_seconds": 0.24977909400331555,
163
+ "gpu_update_seconds": 0.9566314666790962,
164
+ "producer_batches_per_second": 10.995826276338393,
165
+ "gpu_batches_per_second": 1.0453346297205295,
166
+ "producer_headroom": 10.518953418082141,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.267782400783055,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 84,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 337241.253587939,
174
+ "upload_bytes_per_second": 4932306.221605968,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 72.36412244537952,
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": 1.8041386129989405,
184
+ "bytes": 11898323,
185
+ "pages": 40,
186
  "shards": 4
187
  },
188
  {
189
+ "seconds": 1.2189035779992992,
190
+ "bytes": 6954573,
191
+ "pages": 37,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.5599068500014255,
196
+ "bytes": 2415324,
197
+ "pages": 28,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 0.996803005997208,
202
+ "bytes": 6353091,
203
+ "pages": 36,
204
+ "shards": 2
205
  },
206
  {
207
+ "seconds": 1.149405411000771,
208
+ "bytes": 4840926,
209
+ "pages": 32,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.5049040789999708,
214
+ "bytes": 13497203,
215
+ "pages": 34,
216
+ "shards": 4
217
  },
218
  {
219
+ "seconds": 2.2471259019985155,
220
+ "bytes": 6049971,
221
+ "pages": 37,
222
+ "shards": 2
223
  },
224
  {
225
+ "seconds": 1.1292916710008285,
226
+ "bytes": 2686677,
227
  "pages": 37,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.3082805390004069,
232
+ "bytes": 10490550,
233
+ "pages": 32,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.171707494999282,
238
+ "bytes": 6154958,
239
  "pages": 37,
240
+ "shards": 2
241
  },
242
  {
243
+ "seconds": 1.3786574580008164,
244
+ "bytes": 2401890,
245
+ "pages": 28,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.469864372000302,
250
+ "bytes": 11763636,
251
+ "pages": 36,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 1.3772630089988525,
256
+ "bytes": 6008251,
257
+ "pages": 31,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 1.3044562169998244,
262
+ "bytes": 6256652,
263
+ "pages": 37,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 1.3504138789976423,
268
+ "bytes": 6609036,
269
+ "pages": 37,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 0.9880936410008871,
274
+ "bytes": 2162409,
275
+ "pages": 24,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.2899976480002806,
280
+ "bytes": 11585751,
281
+ "pages": 45,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.4168642419972457,
286
+ "bytes": 6615057,
287
  "pages": 28,
288
+ "shards": 3
289
  },
290
  {
291
+ "seconds": 1.1043323689991666,
292
+ "bytes": 7549149,
293
  "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.1122719240011065,
298
+ "bytes": 6287535,
299
+ "pages": 37,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 1.1806445659967721,
304
+ "bytes": 6303271,
305
  "pages": 33,
 
 
 
 
 
 
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.2656237250012055,
310
+ "bytes": 5785840,
311
+ "pages": 28,
 
 
 
 
 
 
 
 
 
 
 
 
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.1340223679981136,
316
+ "bytes": 1097457,
317
+ "pages": 35,
318
+ "shards": 1
319
  },
320
  {
321
+ "seconds": 3.1717748339979153,
322
+ "bytes": 13032899,
323
+ "pages": 32,
324
+ "shards": 4
325
  },
326
  {
327
+ "seconds": 1.520787964000192,
328
+ "bytes": 10348218,
329
  "pages": 34,
330
+ "shards": 3
331
  },
332
  {
333
+ "seconds": 1.0010391289979452,
334
+ "bytes": 2627126,
335
  "pages": 37,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.0633560690002923,
340
+ "bytes": 6091672,
 
 
 
 
 
 
341
  "pages": 28,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 0.8510588639983325,
346
+ "bytes": 1121057,
347
+ "pages": 36,
348
  "shards": 1
349
  },
350
  {
351
+ "seconds": 1.2576819179994345,
352
+ "bytes": 11872557,
353
+ "pages": 37,
354
  "shards": 4
355
  },
356
  {
357
+ "seconds": 1.5458234039979288,
358
+ "bytes": 11492771,
359
  "pages": 33,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 1.4007872010006395,
364
+ "bytes": 2588923,
365
+ "pages": 38,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 1.1760680010011129,
370
+ "bytes": 5785195,
371
+ "pages": 27,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.2628126730014628,
376
+ "bytes": 9956766,
377
+ "pages": 33,
378
+ "shards": 3
379
  },
380
  {
381
+ "seconds": 1.1540077350000502,
382
+ "bytes": 2614009,
383
+ "pages": 37,
384
+ "shards": 2
385
  },
386
  {
387
+ "seconds": 1.9888146960001905,
388
+ "bytes": 6170248,
389
+ "pages": 34,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.596614274996682,
394
+ "bytes": 11247081,
395
+ "pages": 38,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 1.2773266180010978,
400
+ "bytes": 7212495,
401
+ "pages": 31,
402
+ "shards": 3
403
  },
404
  {
405
+ "seconds": 1.216828146003536,
406
+ "bytes": 6096262,
407
+ "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.38459085499926,
412
+ "bytes": 2407911,
413
+ "pages": 29,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 1.9793150009973033,
418
+ "bytes": 10539605,
419
+ "pages": 41,
420
+ "shards": 3
421
  },
422
  {
423
+ "seconds": 0.917002105001302,
424
+ "bytes": 895297,
425
+ "pages": 29,
426
+ "shards": 1
427
  },
428
  {
429
+ "seconds": 1.181815096999344,
430
+ "bytes": 13310041,
431
+ "pages": 38,
432
  "shards": 4
433
  },
434
  {
435
+ "seconds": 1.608177831996727,
436
+ "bytes": 5937731,
437
+ "pages": 29,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.4116140810001525,
442
+ "bytes": 4972441,
443
+ "pages": 35,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 2.1865444229988498,
448
+ "bytes": 7544891,
449
+ "pages": 36,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.1021929470007308,
454
+ "bytes": 6736997,
455
  "pages": 37,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.0818150289996993,
460
+ "bytes": 2336300,
461
  "pages": 28,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.2755690529993444,
466
+ "bytes": 16950690,
467
+ "pages": 39,
468
+ "shards": 4
469
+ },
470
+ {
471
+ "seconds": 1.0866290939993632,
472
+ "bytes": 1216935,
473
  "pages": 36,
474
+ "shards": 2
475
  },
476
  {
477
+ "seconds": 1.176057447999483,
478
+ "bytes": 7472528,
479
  "pages": 31,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.2658942909984034,
484
+ "bytes": 6109615,
485
+ "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.894670263001899,
490
+ "bytes": 6431784,
491
+ "pages": 35,
492
+ "shards": 3
493
+ },
494
+ {
495
+ "seconds": 1.2564982580006472,
496
+ "bytes": 6256940,
497
  "pages": 37,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.1927087240001129,
502
+ "bytes": 6721845,
503
+ "pages": 31,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.142808556996897,
508
+ "bytes": 7320475,
509
+ "pages": 31,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.333093295001163,
514
+ "bytes": 10300175,
515
+ "pages": 35,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 1.2003908199985744,
520
+ "bytes": 2637875,
521
  "pages": 38,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 2.519623430998763,
526
+ "bytes": 6025157,
527
+ "pages": 28,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.0435017019990482,
532
+ "bytes": 1045686,
533
+ "pages": 30,
534
+ "shards": 1
535
+ },
536
+ {
537
+ "seconds": 1.3271499310030777,
538
+ "bytes": 10074712,
539
+ "pages": 25,
540
  "shards": 3
541
  },
542
  {
543
+ "seconds": 2.2123866580004687,
544
+ "bytes": 13632293,
545
+ "pages": 49,
546
+ "shards": 4
547
  },
548
  {
549
+ "seconds": 1.2271638120000716,
550
+ "bytes": 936260,
551
+ "pages": 26,
552
  "shards": 1
553
  },
554
  {
555
+ "seconds": 1.1434297949999745,
556
+ "bytes": 7632762,
557
+ "pages": 39,
558
+ "shards": 3
559
+ },
560
+ {
561
+ "seconds": 1.268450359999406,
562
+ "bytes": 10380594,
563
+ "pages": 38,
564
+ "shards": 3
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bb7ceaeda93bfb3b23d8a7a9fc6b379ef73c40abcbb8d58fb638329c569f90e7
3
- size 721774783
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da99d6bb0dd16c05dc0dcdcb4b5f86f1c1ca2eda0fbea8eec1ea8b1c34c0fc81
3
+ size 721781183