appvoid commited on
Commit
25d7698
·
verified ·
1 Parent(s): d6be955

Qwen90M step 255000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +44 -44
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +240 -240
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 250000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
- "shard_scheduler.json": "51e06602336cf2ff9c7fbf2702cb9edfe3a4879775d6e82cb900f0895494a34a",
15
- "coverage.json": "fa6f7ebd74b32c3a670cb83966d87b06345ed8c181b0803237a454541a8aa878",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 255000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "fd97f7dff3c950a4c906fc0d4a7fd3b2954ec1ce69c8fc7fe9fbf0fb58cc2fc1",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "a2ebbe485478b945d252786a3ac085f24afc9d7d83bc13cc75393c2a988fb4dd",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
+ "shard_scheduler.json": "040240fbf545ea1d490b42b841ab997a911ae28bae5ed7a72035f5d7db770fd7",
15
+ "coverage.json": "836ac61bfc32d5771b0a24e2739ea3cc1013fa5a041bce0fbb429b3c8bf253f8",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 250000,
3
  "stats": {
4
- "climbmix/blocks": 9000000,
5
- "climbmix/targets": 15044821319,
6
- "climbmix/input_tokens": 15045774685,
7
- "ultra_qa/blocks": 3375000,
8
- "ultra_qa/targets": 5505900381,
9
- "ultra_qa/input_tokens": 5505910963,
10
- "cortex/blocks": 3375000,
11
- "cortex/targets": 128400391,
12
- "cortex/input_tokens": 5364742698,
13
- "ultra_style/blocks": 3375000,
14
- "ultra_style/targets": 5338964418,
15
- "ultra_style/input_tokens": 5339244172,
16
- "openmath/blocks": 2250000,
17
- "openmath/targets": 3476637803,
18
- "openmath/input_tokens": 3477796696,
19
- "rewrite6/blocks": 1125000,
20
- "rewrite6/targets": 1844270111,
21
- "rewrite6/input_tokens": 1846292063
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1232,
26
- "row": 2049,
27
  "start": {
28
- "file": 59,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 53,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 5048321
38
  },
39
  "rewrite6": {
40
- "page": 1,
41
- "row": 847,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 4096
47
  }
48
  },
49
  "epoch": 14,
50
  "offset": 0,
51
- "documents": 2880599
52
  },
53
  "openmath": {
54
- "page": 770,
55
- "row": 2180,
56
  "start": {
57
  "file": 7,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 97,
61
- "row": 370
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 3156100
67
  },
68
  "ultra_qa": {
69
- "page": 333,
70
- "row": 364,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
- "row": 1742
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 1364332
82
  },
83
  "ultra_style": {
84
- "page": 508,
85
- "row": 1215,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
- "row": 25940
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 2081983
97
  },
98
  "cortex": {
99
- "page": 65453,
100
- "row": 34,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 8378018
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 255000,
3
  "stats": {
4
+ "climbmix/blocks": 9180000,
5
+ "climbmix/targets": 15346078422,
6
+ "climbmix/input_tokens": 15347051076,
7
+ "ultra_qa/blocks": 3442500,
8
+ "ultra_qa/targets": 5615917971,
9
+ "ultra_qa/input_tokens": 5615928771,
10
+ "cortex/blocks": 3442500,
11
+ "cortex/targets": 130960799,
12
+ "cortex/input_tokens": 5472030778,
13
+ "ultra_style/blocks": 3442500,
14
+ "ultra_style/targets": 5445625285,
15
+ "ultra_style/input_tokens": 5445910748,
16
+ "openmath/blocks": 2295000,
17
+ "openmath/targets": 3546217432,
18
+ "openmath/input_tokens": 3547399641,
19
+ "rewrite6/blocks": 1147500,
20
+ "rewrite6/targets": 1881142352,
21
+ "rewrite6/input_tokens": 1883204831
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1257,
26
+ "row": 595,
27
  "start": {
28
+ "file": 60,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 70,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 5149267
38
  },
39
  "rewrite6": {
40
+ "page": 15,
41
+ "row": 1160,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 61440
47
  }
48
  },
49
  "epoch": 14,
50
  "offset": 0,
51
+ "documents": 2938256
52
  },
53
  "openmath": {
54
+ "page": 785,
55
+ "row": 3990,
56
  "start": {
57
  "file": 7,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 158,
61
+ "row": 810
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 3219350
67
  },
68
  "ultra_qa": {
69
+ "page": 339,
70
+ "row": 3105,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 5,
76
+ "row": 26318
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 1391649
82
  },
83
  "ultra_style": {
84
+ "page": 518,
85
+ "row": 2074,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 0,
91
+ "row": 66900
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 2123802
97
  },
98
  "cortex": {
99
+ "page": 66763,
100
+ "row": 12,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 8545676
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 250000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "51e06602336cf2ff9c7fbf2702cb9edfe3a4879775d6e82cb900f0895494a34a", "coverage.json": "fa6f7ebd74b32c3a670cb83966d87b06345ed8c181b0803237a454541a8aa878", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000250000"}
 
1
+ {"step": 255000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "fd97f7dff3c950a4c906fc0d4a7fd3b2954ec1ce69c8fc7fe9fbf0fb58cc2fc1", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a2ebbe485478b945d252786a3ac085f24afc9d7d83bc13cc75393c2a988fb4dd", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "040240fbf545ea1d490b42b841ab997a911ae28bae5ed7a72035f5d7db770fd7", "coverage.json": "836ac61bfc32d5771b0a24e2739ea3cc1013fa5a041bce0fbb429b3c8bf253f8", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000255000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b39daf23cbfc7a4db45dc715dda0257d7a0aceacebd6722e70e52b392a3d914f
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd97f7dff3c950a4c906fc0d4a7fd3b2954ec1ce69c8fc7fe9fbf0fb58cc2fc1
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75570.71051583829,
26
- "required_pages_per_second_upper_estimate": 0.006550277655298134
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9446.338814479786,
40
- "required_pages_per_second_upper_estimate": 0.003768285788447338
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18892.677628959573,
54
- "required_pages_per_second_upper_estimate": 0.004370488469689805
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28339.01644343936,
68
- "required_pages_per_second_upper_estimate": 0.0017256974180770313
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28339.01644343936,
82
- "required_pages_per_second_upper_estimate": 0.0021816159332343353
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28339.01644343936,
96
- "required_pages_per_second_upper_estimate": 0.3518757396406541
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
- "producer_batch_seconds": 0.129124772499992,
162
- "producer_p95_seconds": 0.24966627399953722,
163
- "gpu_update_seconds": 0.9756160752855156,
164
- "producer_batches_per_second": 7.744447332908656,
165
- "gpu_batches_per_second": 1.0249933609461575,
166
- "producer_headroom": 7.555607312187723,
167
  "sustainable": true,
168
- "buffer_seconds": 60.48819666770197,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 382211.29352878686,
174
- "upload_bytes_per_second": 4854486.8576622605,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 82.01364616773323,
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": 0.7338976749997528,
184
- "bytes": 1218142,
185
- "pages": 34,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 1.1807161469996572,
190
- "bytes": 7525394,
191
- "pages": 28,
192
- "shards": 3
193
- },
194
- {
195
- "seconds": 1.1362227980007447,
196
- "bytes": 11027017,
197
- "pages": 35,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 0.8288988310005152,
202
- "bytes": 5801101,
203
- "pages": 35,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.2185628970000835,
208
- "bytes": 7809668,
209
  "pages": 37,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.2807725910006411,
214
- "bytes": 1082933,
215
- "pages": 33,
216
  "shards": 1
217
  },
218
  {
219
- "seconds": 1.2944532310002614,
220
- "bytes": 11460253,
 
 
 
 
 
 
221
  "pages": 33,
222
  "shards": 4
223
  },
224
  {
225
- "seconds": 1.090632098000242,
226
- "bytes": 1122627,
227
  "pages": 33,
228
  "shards": 1
229
  },
230
  {
231
- "seconds": 1.176693084999897,
232
- "bytes": 12183576,
233
- "pages": 36,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.0236794259999442,
238
- "bytes": 7487764,
239
- "pages": 29,
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 1.1006291130006502,
244
- "bytes": 6137509,
245
- "pages": 39,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.0797216399996614,
250
- "bytes": 5987959,
251
- "pages": 27,
252
- "shards": 2
253
  },
254
  {
255
- "seconds": 1.322158891000072,
256
- "bytes": 2479484,
257
  "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 1.3229940370001714,
262
- "bytes": 6465970,
263
- "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.3629516780001723,
268
- "bytes": 11728657,
269
- "pages": 30,
270
  "shards": 4
271
  },
272
  {
273
- "seconds": 1.096408940999936,
274
- "bytes": 6091544,
 
 
 
 
 
 
275
  "pages": 35,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.042788374000338,
280
- "bytes": 6163555,
281
- "pages": 36,
282
- "shards": 2
283
  },
284
  {
285
- "seconds": 1.04696464799963,
286
- "bytes": 2343472,
287
- "pages": 28,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.3791698659997564,
292
- "bytes": 10729274,
293
- "pages": 36,
294
  "shards": 4
295
  },
296
  {
297
- "seconds": 1.2665546050002376,
298
- "bytes": 6293373,
299
- "pages": 36,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 1.4158054800000173,
304
- "bytes": 8082619,
305
- "pages": 30,
306
- "shards": 3
307
  },
308
  {
309
- "seconds": 1.100743914999839,
310
- "bytes": 6375983,
311
  "pages": 35,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.1518813020011294,
316
- "bytes": 6877765,
317
- "pages": 35,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.007327716999498,
322
- "bytes": 296016,
323
- "pages": 9,
324
- "shards": 1
325
  },
326
  {
327
- "seconds": 1.1130705590003345,
328
- "bytes": 6042311,
329
- "pages": 35,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 0.9439528090006206,
334
- "bytes": 7900957,
335
  "pages": 37,
336
  "shards": 3
337
  },
338
  {
339
- "seconds": 0.7852815879996342,
340
- "bytes": 838334,
341
- "pages": 26,
342
  "shards": 1
343
  },
344
  {
345
- "seconds": 4.030324741001095,
346
- "bytes": 15386927,
347
- "pages": 28,
348
- "shards": 4
349
- },
350
- {
351
- "seconds": 0.845626510999864,
352
- "bytes": 2613623,
353
- "pages": 35,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 1.1139741530005267,
358
- "bytes": 5909960,
359
- "pages": 35,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.2427591689993278,
364
- "bytes": 910923,
365
- "pages": 26,
366
  "shards": 1
367
  },
368
  {
369
- "seconds": 1.2945012599993788,
370
- "bytes": 12063808,
371
- "pages": 38,
372
- "shards": 4
373
  },
374
  {
375
- "seconds": 0.8872203939990868,
376
- "bytes": 6084404,
377
- "pages": 35,
378
- "shards": 2
379
  },
380
  {
381
- "seconds": 1.511142709001433,
382
- "bytes": 6655300,
383
- "pages": 29,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.0988184660000115,
388
- "bytes": 7486365,
389
  "pages": 37,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.2685245129996474,
394
- "bytes": 10318034,
395
- "pages": 29,
 
 
 
 
 
 
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 1.292125334000957,
400
- "bytes": 1081552,
401
- "pages": 33,
 
 
 
 
 
 
402
  "shards": 1
403
  },
404
  {
405
- "seconds": 1.197562700999697,
406
- "bytes": 7759288,
407
- "pages": 35,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 0.8769548000000214,
412
- "bytes": 1135557,
413
- "pages": 34,
414
- "shards": 1
415
  },
416
  {
417
- "seconds": 3.838755348999257,
418
- "bytes": 11603941,
419
- "pages": 24,
 
 
 
 
 
 
420
  "shards": 4
421
  },
422
  {
423
- "seconds": 1.57969898599913,
424
- "bytes": 12094782,
425
- "pages": 45,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 0.8754016399998363,
430
- "bytes": 1113722,
431
- "pages": 34,
432
- "shards": 1
433
  },
434
  {
435
- "seconds": 4.401627224999174,
436
- "bytes": 7622267,
437
- "pages": 32,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 0.9423913660011749,
442
- "bytes": 5379592,
443
  "pages": 34,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.2426582910011348,
448
- "bytes": 7714336,
449
- "pages": 37,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 1.0709264350007288,
454
- "bytes": 727110,
455
- "pages": 23,
456
- "shards": 1
457
  },
458
  {
459
- "seconds": 1.4975259919992823,
460
- "bytes": 12239566,
461
- "pages": 44,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 1.4755159030009963,
466
- "bytes": 11647558,
467
- "pages": 33,
468
- "shards": 4
469
  },
470
  {
471
- "seconds": 0.8895111490001,
472
- "bytes": 1197830,
473
- "pages": 33,
474
  "shards": 1
475
  },
476
  {
477
- "seconds": 1.169628391000515,
478
- "bytes": 7496810,
479
  "pages": 38,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.5263510319982743,
484
- "bytes": 5728692,
485
- "pages": 27,
486
- "shards": 3
487
  },
488
  {
489
- "seconds": 1.2046007990011276,
490
- "bytes": 5154137,
491
- "pages": 31,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 2.8337960279986874,
496
- "bytes": 11685666,
497
- "pages": 29,
498
- "shards": 3
499
  },
500
  {
501
- "seconds": 1.282873722999284,
502
- "bytes": 2666361,
503
- "pages": 35,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 1.1408444070002588,
508
- "bytes": 5962047,
509
  "pages": 35,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.150664095999673,
514
- "bytes": 11982825,
515
- "pages": 34,
516
- "shards": 4
517
- },
518
- {
519
- "seconds": 0.9085387359991728,
520
- "bytes": 1084848,
521
- "pages": 33,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.2088771979997546,
526
- "bytes": 6082441,
527
  "pages": 35,
528
  "shards": 2
529
  },
530
  {
531
- "seconds": 1.0392706679995172,
532
- "bytes": 8238622,
533
- "pages": 31,
534
- "shards": 3
535
- },
536
- {
537
- "seconds": 1.0565556260007725,
538
- "bytes": 5902213,
539
- "pages": 31,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.6428124459998799,
544
- "bytes": 11197872,
545
- "pages": 39,
546
  "shards": 4
547
  },
548
  {
549
- "seconds": 1.1795239030016091,
550
- "bytes": 829951,
551
- "pages": 25,
552
- "shards": 1
553
  },
554
  {
555
- "seconds": 1.183389565001562,
556
- "bytes": 6316614,
557
- "pages": 35,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.805858101000922,
562
- "bytes": 7507792,
563
- "pages": 34,
564
  "shards": 3
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75186.44264578017,
26
+ "required_pages_per_second_upper_estimate": 0.006516970290239508
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9398.305330722522,
40
+ "required_pages_per_second_upper_estimate": 0.003749124513611984
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18796.610661445044,
54
+ "required_pages_per_second_upper_estimate": 0.004348265067476201
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28194.915992167564,
68
+ "required_pages_per_second_upper_estimate": 0.0017169224566312173
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28194.915992167564,
82
+ "required_pages_per_second_upper_estimate": 0.002170522681599853
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28194.915992167564,
96
+ "required_pages_per_second_upper_estimate": 0.3500864943090451
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
+ "producer_batch_seconds": 0.08646610254689335,
162
+ "producer_p95_seconds": 0.24833947200022521,
163
+ "gpu_update_seconds": 0.9806023187896891,
164
+ "producer_batches_per_second": 11.565225799991017,
165
+ "gpu_batches_per_second": 1.0197813943926346,
166
+ "producer_headroom": 11.340887236797528,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.79734376496072,
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": 312898.0981319677,
174
+ "upload_bytes_per_second": 4945666.70118425,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 67.14064796418451,
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": 4.401627224999174,
184
+ "bytes": 7622267,
185
+ "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 0.9423913660011749,
190
+ "bytes": 5379592,
191
+ "pages": 34,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.2426582910011348,
196
+ "bytes": 7714336,
197
  "pages": 37,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.0709264350007288,
202
+ "bytes": 727110,
203
+ "pages": 23,
204
  "shards": 1
205
  },
206
  {
207
+ "seconds": 1.4975259919992823,
208
+ "bytes": 12239566,
209
+ "pages": 44,
210
+ "shards": 3
211
+ },
212
+ {
213
+ "seconds": 1.4755159030009963,
214
+ "bytes": 11647558,
215
  "pages": 33,
216
  "shards": 4
217
  },
218
  {
219
+ "seconds": 0.8895111490001,
220
+ "bytes": 1197830,
221
  "pages": 33,
222
  "shards": 1
223
  },
224
  {
225
+ "seconds": 1.169628391000515,
226
+ "bytes": 7496810,
227
+ "pages": 38,
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 1.5263510319982743,
232
+ "bytes": 5728692,
233
+ "pages": 27,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.2046007990011276,
238
+ "bytes": 5154137,
239
+ "pages": 31,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 2.8337960279986874,
244
+ "bytes": 11685666,
245
+ "pages": 29,
246
+ "shards": 3
247
  },
248
  {
249
+ "seconds": 1.282873722999284,
250
+ "bytes": 2666361,
251
  "pages": 35,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 1.1408444070002588,
256
+ "bytes": 5962047,
257
+ "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.150664095999673,
262
+ "bytes": 11982825,
263
+ "pages": 34,
264
  "shards": 4
265
  },
266
  {
267
+ "seconds": 0.9085387359991728,
268
+ "bytes": 1084848,
269
+ "pages": 33,
270
+ "shards": 1
271
+ },
272
+ {
273
+ "seconds": 1.2088771979997546,
274
+ "bytes": 6082441,
275
  "pages": 35,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.0392706679995172,
280
+ "bytes": 8238622,
281
+ "pages": 31,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 1.0565556260007725,
286
+ "bytes": 5902213,
287
+ "pages": 31,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.6428124459998799,
292
+ "bytes": 11197872,
293
+ "pages": 39,
294
  "shards": 4
295
  },
296
  {
297
+ "seconds": 1.1795239030016091,
298
+ "bytes": 829951,
299
+ "pages": 25,
300
+ "shards": 1
 
 
 
 
 
 
301
  },
302
  {
303
+ "seconds": 1.183389565001562,
304
+ "bytes": 6316614,
305
  "pages": 35,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.805858101000922,
310
+ "bytes": 7507792,
311
+ "pages": 34,
312
  "shards": 3
313
  },
314
  {
315
+ "seconds": 2.5084295430006023,
316
+ "bytes": 11998594,
317
+ "pages": 38,
318
+ "shards": 3
319
  },
320
  {
321
+ "seconds": 1.094205030000012,
322
+ "bytes": 275280,
323
+ "pages": 8,
324
+ "shards": 1
325
  },
326
  {
327
+ "seconds": 1.677992199000073,
328
+ "bytes": 7646685,
329
  "pages": 37,
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 0.9418184709993511,
334
+ "bytes": 1092329,
335
+ "pages": 34,
336
  "shards": 1
337
  },
338
  {
339
+ "seconds": 1.2206816289999551,
340
+ "bytes": 9827767,
341
+ "pages": 34,
342
+ "shards": 3
 
 
 
 
 
 
343
  },
344
  {
345
+ "seconds": 1.1708706409990555,
346
+ "bytes": 7164779,
347
+ "pages": 29,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.2709880140009773,
352
+ "bytes": 1216081,
353
+ "pages": 34,
354
  "shards": 1
355
  },
356
  {
357
+ "seconds": 1.5031630230005248,
358
+ "bytes": 11828419,
359
+ "pages": 35,
360
+ "shards": 3
361
  },
362
  {
363
+ "seconds": 1.1863015779999841,
364
+ "bytes": 6891470,
365
+ "pages": 32,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 1.3068818119991192,
370
+ "bytes": 5782255,
371
+ "pages": 36,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.4370965530015383,
376
+ "bytes": 7562097,
377
  "pages": 37,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 0.8747751649989368,
382
+ "bytes": 1118624,
383
+ "pages": 34,
384
+ "shards": 1
385
+ },
386
+ {
387
+ "seconds": 1.1147228019999602,
388
+ "bytes": 10226673,
389
+ "pages": 28,
390
  "shards": 3
391
  },
392
  {
393
+ "seconds": 1.303761271001349,
394
+ "bytes": 13900408,
395
+ "pages": 43,
396
+ "shards": 4
397
+ },
398
+ {
399
+ "seconds": 0.9504027439998026,
400
+ "bytes": 830160,
401
+ "pages": 25,
402
  "shards": 1
403
  },
404
  {
405
+ "seconds": 1.2181768209993606,
406
+ "bytes": 7539545,
407
+ "pages": 38,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.3715665409999929,
412
+ "bytes": 4880519,
413
+ "pages": 35,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.2786054230000445,
418
+ "bytes": 6003136,
419
+ "pages": 35,
420
+ "shards": 2
421
+ },
422
+ {
423
+ "seconds": 1.351524421999784,
424
+ "bytes": 7384349,
425
+ "pages": 30,
426
  "shards": 4
427
  },
428
  {
429
+ "seconds": 1.2958617730000697,
430
+ "bytes": 7155083,
431
+ "pages": 38,
432
+ "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0062387039997702,
436
+ "bytes": 10437075,
437
+ "pages": 30,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 1.1012680910007475,
442
+ "bytes": 7471820,
443
+ "pages": 38,
444
  "shards": 3
445
  },
446
  {
447
+ "seconds": 0.9738275679992512,
448
+ "bytes": 1102759,
449
  "pages": 34,
450
+ "shards": 1
451
  },
452
  {
453
+ "seconds": 1.0251637159999518,
454
+ "bytes": 7585581,
455
+ "pages": 29,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 1.1754312260000006,
460
+ "bytes": 5287839,
461
+ "pages": 37,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.096052540000528,
466
+ "bytes": 12098645,
467
+ "pages": 33,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.8189289659967471,
472
+ "bytes": 7675879,
473
+ "pages": 37,
474
+ "shards": 3
475
  },
476
  {
477
+ "seconds": 1.221604631999071,
478
+ "bytes": 830309,
479
+ "pages": 26,
480
  "shards": 1
481
  },
482
  {
483
+ "seconds": 1.7278612150003028,
484
+ "bytes": 10862160,
485
  "pages": 38,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 0.8819487159998971,
490
+ "bytes": 2668679,
491
+ "pages": 36,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.1101788160012802,
496
+ "bytes": 5690031,
497
+ "pages": 26,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.5474547390003863,
502
+ "bytes": 13326359,
503
+ "pages": 36,
504
+ "shards": 4
505
  },
506
  {
507
+ "seconds": 1.2802494179995847,
508
+ "bytes": 4852642,
509
+ "pages": 37,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.42962039700069,
514
+ "bytes": 6308465,
515
  "pages": 35,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0526814939985343,
520
+ "bytes": 2354967,
521
+ "pages": 27,
522
+ "shards": 2
 
 
 
 
 
 
523
  },
524
  {
525
+ "seconds": 1.0578431400026602,
526
+ "bytes": 6390274,
527
  "pages": 35,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.1717372180028178,
532
+ "bytes": 4762409,
533
+ "pages": 29,
 
 
 
 
 
 
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.406141793999268,
538
+ "bytes": 13345421,
539
+ "pages": 38,
540
  "shards": 4
541
  },
542
  {
543
+ "seconds": 1.0539056280031218,
544
+ "bytes": 6000362,
545
+ "pages": 36,
546
+ "shards": 2
547
  },
548
  {
549
+ "seconds": 1.0857812779977394,
550
+ "bytes": 2115915,
551
+ "pages": 24,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.3246665000006033,
556
+ "bytes": 11374746,
557
+ "pages": 45,
558
  "shards": 3
559
+ },
560
+ {
561
+ "seconds": 1.0599599050001416,
562
+ "bytes": 875478,
563
+ "pages": 25,
564
+ "shards": 1
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e91afb32a91ba214e7473f46d6f99a0a931e405920d457559aea60d50123c7e4
3
- size 721772479
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2ebbe485478b945d252786a3ac085f24afc9d7d83bc13cc75393c2a988fb4dd
3
+ size 721771711