appvoid commited on
Commit
a5fdb31
·
verified ·
1 Parent(s): c4b0c71

Qwen90M step 395000; 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 +239 -239
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 390000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
14
- "shard_scheduler.json": "a3cecf4f99b4babbedb1f1e008b99c2bf6e52201ec4ee5a9670c2a4034e567de",
15
- "coverage.json": "eb0c2309e40705779a35e534d2ee49b824e6be0073c4849921791384d1631876",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 395000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15",
14
+ "shard_scheduler.json": "236c080251e2cd0c6338dee63d27f4a3f72745e2b8953e3c3cd894d9a6eaa61a",
15
+ "coverage.json": "e9b422c27949b4fcbed30732a16dcbd7fb3ac93a942ade09b9e407907eebd96b",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 390000,
3
  "stats": {
4
- "climbmix/blocks": 14040000,
5
- "climbmix/targets": 23469768992,
6
- "climbmix/input_tokens": 23471255758,
7
- "ultra_qa/blocks": 5265000,
8
- "ultra_qa/targets": 8589103233,
9
- "ultra_qa/input_tokens": 8589119791,
10
- "cortex/blocks": 5265000,
11
- "cortex/targets": 200266960,
12
- "cortex/input_tokens": 8369485865,
13
- "ultra_style/blocks": 5265000,
14
- "ultra_style/targets": 8328115889,
15
- "ultra_style/input_tokens": 8328552449,
16
- "openmath/blocks": 3510000,
17
- "openmath/targets": 5423887511,
18
- "openmath/input_tokens": 5425694857,
19
- "rewrite6/blocks": 1755000,
20
- "rewrite6/targets": 2877105386,
21
- "rewrite6/input_tokens": 2880259808
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1922,
26
- "row": 3973,
27
  "start": {
28
- "file": 93,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 3,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 6144,
37
- "documents": 7876485
38
  },
39
  "rewrite6": {
40
- "page": 44,
41
- "row": 214,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 180224
47
  }
48
  },
49
- "epoch": 21,
50
  "offset": 0,
51
- "documents": 4493922
52
  },
53
  "openmath": {
54
- "page": 1201,
55
- "row": 3048,
56
  "start": {
57
  "file": 11,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 116,
61
- "row": 146
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 4922344
67
  },
68
  "ultra_qa": {
69
- "page": 519,
70
- "row": 2369,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 0,
76
- "row": 47323
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2128193
82
  },
83
  "ultra_style": {
84
- "page": 793,
85
- "row": 45,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 5,
91
- "row": 82015
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 3248173
97
  },
98
  "cortex": {
99
- "page": 102116,
100
- "row": 47,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 13070895
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 395000,
3
  "stats": {
4
+ "climbmix/blocks": 14220000,
5
+ "climbmix/targets": 23770575825,
6
+ "climbmix/input_tokens": 23772081480,
7
+ "ultra_qa/blocks": 5332500,
8
+ "ultra_qa/targets": 8699064740,
9
+ "ultra_qa/input_tokens": 8699081510,
10
+ "cortex/blocks": 5332500,
11
+ "cortex/targets": 202831075,
12
+ "cortex/input_tokens": 8476791839,
13
+ "ultra_style/blocks": 5332500,
14
+ "ultra_style/targets": 8434525911,
15
+ "ultra_style/input_tokens": 8434968028,
16
+ "openmath/blocks": 3555000,
17
+ "openmath/targets": 5493545787,
18
+ "openmath/input_tokens": 5495376342,
19
+ "rewrite6/blocks": 1777500,
20
+ "rewrite6/targets": 2913987031,
21
+ "rewrite6/input_tokens": 2917181699
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1947,
26
+ "row": 2219,
27
  "start": {
28
+ "file": 94,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 12288,
37
+ "documents": 7977131
38
  },
39
  "rewrite6": {
40
+ "page": 7,
41
+ "row": 3678,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 28672
47
  }
48
  },
49
+ "epoch": 22,
50
  "offset": 0,
51
+ "documents": 4551238
52
  },
53
  "openmath": {
54
+ "page": 1217,
55
+ "row": 561,
56
  "start": {
57
  "file": 11,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 181,
61
+ "row": 682
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 4985393
67
  },
68
  "ultra_qa": {
69
+ "page": 526,
70
+ "row": 988,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 1,
76
+ "row": 11508
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2155484
82
  },
83
  "ultra_style": {
84
+ "page": 803,
85
+ "row": 968,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 6,
91
+ "row": 39519
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 3290056
97
  },
98
  "cortex": {
99
+ "page": 103426,
100
+ "row": 6,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 13238534
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 390000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "a3cecf4f99b4babbedb1f1e008b99c2bf6e52201ec4ee5a9670c2a4034e567de", "coverage.json": "eb0c2309e40705779a35e534d2ee49b824e6be0073c4849921791384d1631876", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000390000"}
 
1
+ {"step": 395000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "e3923fa86352f6d878fd8d37b6b9fc374577f52be522203b0b865badf630da15", "shard_scheduler.json": "236c080251e2cd0c6338dee63d27f4a3f72745e2b8953e3c3cd894d9a6eaa61a", "coverage.json": "e9b422c27949b4fcbed30732a16dcbd7fb3ac93a942ade09b9e407907eebd96b", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000395000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3445e25637ff28b766a77ae05bc59ec15b294ce669d092bc5e88156d9adba6ce
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90923eaf3e7990d5ee70aaacb1ff2a4802811438b0db5687799bf5f234240ca7
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": 75353.21624803833,
26
- "required_pages_per_second_upper_estimate": 0.006531425803399414
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": 9419.15203100479,
40
- "required_pages_per_second_upper_estimate": 0.003757440574040526
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": 18838.30406200958,
54
- "required_pages_per_second_upper_estimate": 0.004357910101918019
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": 28257.456093014374,
68
- "required_pages_per_second_upper_estimate": 0.0017207308206502374
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": 28257.456093014374,
82
- "required_pages_per_second_upper_estimate": 0.00217533719168512
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": 28257.456093014374,
96
- "required_pages_per_second_upper_estimate": 0.35086303305330935
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": 717798,
161
- "producer_batch_seconds": 0.09481285885919988,
162
- "producer_p95_seconds": 0.3795729499979643,
163
- "gpu_update_seconds": 0.9784320254799922,
164
- "producer_batches_per_second": 10.547092578286579,
165
- "gpu_batches_per_second": 1.022043406142013,
166
- "producer_headroom": 10.319613154297931,
167
  "sustainable": true,
168
- "buffer_seconds": 60.662785579759515,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 86,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 285047.90833474684,
174
- "upload_bytes_per_second": 5287078.859485369,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 61.164645552938495,
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": 1.0007863089995226,
184
- "bytes": 6027863,
185
- "pages": 36,
186
- "shards": 2
187
  },
188
  {
189
- "seconds": 0.9187128839985235,
190
- "bytes": 2707013,
191
- "pages": 37,
 
 
 
 
 
 
192
  "shards": 2
193
  },
194
  {
195
- "seconds": 0.9819577199996274,
196
- "bytes": 10903211,
197
- "pages": 33,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 0.860417497002345,
202
- "bytes": 864723,
203
  "pages": 26,
204
  "shards": 1
205
  },
206
  {
207
- "seconds": 1.116861483998946,
208
- "bytes": 7838954,
209
  "pages": 38,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.1656662089990277,
214
- "bytes": 11880694,
215
- "pages": 33,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 1.2683707119977043,
220
- "bytes": 2325676,
221
- "pages": 30,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 1.4917878320011368,
226
- "bytes": 10275338,
227
- "pages": 40,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.1128048520004086,
232
- "bytes": 6259334,
 
 
 
 
 
 
233
  "pages": 36,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.2836710900010075,
238
- "bytes": 2351070,
239
- "pages": 28,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 1.2001916299996083,
244
- "bytes": 10729681,
245
- "pages": 35,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 1.0578519230002712,
250
- "bytes": 8294494,
251
- "pages": 34,
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 1.1691894300020067,
256
- "bytes": 6431968,
257
- "pages": 36,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 0.9079114560008748,
262
- "bytes": 1141999,
263
- "pages": 35,
264
- "shards": 1
265
- },
266
- {
267
- "seconds": 1.1828810689985403,
268
- "bytes": 7222759,
269
- "pages": 29,
270
  "shards": 3
271
  },
272
  {
273
- "seconds": 1.2303334819989686,
274
- "bytes": 10801054,
275
- "pages": 35,
276
- "shards": 3
277
  },
278
  {
279
- "seconds": 0.9937779290012259,
280
- "bytes": 2662341,
281
  "pages": 37,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.1579458060004981,
286
- "bytes": 11572043,
287
- "pages": 34,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 0.8931559880002169,
292
- "bytes": 949478,
293
- "pages": 28,
294
- "shards": 1
295
  },
296
  {
297
- "seconds": 1.1832785620026698,
298
- "bytes": 12385340,
299
- "pages": 36,
300
- "shards": 4
301
  },
302
  {
303
- "seconds": 1.0688240559975384,
304
- "bytes": 6186249,
305
  "pages": 37,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.054510723999556,
310
- "bytes": 1093266,
311
- "pages": 33,
312
- "shards": 1
313
  },
314
  {
315
- "seconds": 1.209129931001371,
316
- "bytes": 7450282,
317
- "pages": 31,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 2.1295198230000096,
322
- "bytes": 324019,
323
  "pages": 10,
324
  "shards": 1
325
  },
326
  {
327
- "seconds": 1.2268210219990578,
328
- "bytes": 15731128,
329
- "pages": 39,
330
- "shards": 4
331
  },
332
  {
333
- "seconds": 0.8699136540017207,
334
- "bytes": 2665528,
335
- "pages": 36,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.3528436140004487,
340
- "bytes": 6109175,
341
- "pages": 36,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.3453164509992348,
346
- "bytes": 11716150,
347
- "pages": 31,
348
- "shards": 4
349
  },
350
  {
351
- "seconds": 0.8494216189974395,
352
- "bytes": 1181935,
353
- "pages": 35,
354
- "shards": 1
355
  },
356
  {
357
- "seconds": 3.5161270839998906,
358
- "bytes": 6159420,
359
- "pages": 30,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 1.2511098670001957,
364
- "bytes": 8306730,
365
- "pages": 34,
366
- "shards": 3
367
  },
368
  {
369
- "seconds": 1.168453648002469,
370
- "bytes": 10300928,
371
- "pages": 38,
372
- "shards": 3
373
  },
374
  {
375
- "seconds": 4.622547731996747,
376
- "bytes": 6328963,
377
  "pages": 33,
378
- "shards": 2
379
  },
380
  {
381
- "seconds": 0.9812373249988013,
382
- "bytes": 2474379,
383
- "pages": 32,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.035560734999308,
388
- "bytes": 6144689,
 
 
 
 
 
 
389
  "pages": 36,
390
- "shards": 2
391
  },
392
  {
393
- "seconds": 1.3160131810000166,
394
- "bytes": 6821427,
395
- "pages": 32,
396
- "shards": 3
397
  },
398
  {
399
- "seconds": 1.3006846259995655,
400
- "bytes": 12006518,
401
- "pages": 37,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 0.8710538570012432,
406
- "bytes": 5900566,
407
- "pages": 28,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 1.0173681279993616,
412
- "bytes": 2612410,
413
- "pages": 37,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.2024259139980131,
418
- "bytes": 11156038,
419
- "pages": 33,
420
  "shards": 3
421
  },
422
  {
423
- "seconds": 0.9720597020022979,
424
- "bytes": 1226766,
425
- "pages": 34,
426
  "shards": 1
427
  },
428
  {
429
- "seconds": 1.4056276539995451,
430
- "bytes": 7428876,
431
- "pages": 30,
432
- "shards": 3
433
  },
434
  {
435
- "seconds": 1.1535106260016619,
436
- "bytes": 12418046,
437
  "pages": 38,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 0.8009545050008455,
442
- "bytes": 2263071,
443
- "pages": 25,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 2.506807755002228,
448
- "bytes": 10223439,
449
- "pages": 45,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 0.890203859002213,
454
- "bytes": 916282,
455
  "pages": 26,
456
  "shards": 1
457
  },
458
  {
459
- "seconds": 1.0841833719969145,
460
- "bytes": 7541614,
 
 
 
 
 
 
461
  "pages": 38,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 0.8043565109983319,
466
- "bytes": 540912,
467
- "pages": 18,
468
  "shards": 1
469
  },
470
  {
471
- "seconds": 1.4004500939990976,
472
- "bytes": 16241287,
473
- "pages": 47,
474
  "shards": 4
475
  },
476
  {
477
- "seconds": 1.0899500639934558,
478
- "bytes": 7493063,
479
- "pages": 31,
480
- "shards": 3
481
- },
482
- {
483
- "seconds": 0.8779368290051934,
484
- "bytes": 1248675,
485
  "pages": 35,
486
- "shards": 1
487
- },
488
- {
489
- "seconds": 1.2020556749994284,
490
- "bytes": 12102958,
491
- "pages": 36,
492
  "shards": 4
493
  },
494
  {
495
- "seconds": 1.048562693002168,
496
- "bytes": 883869,
497
  "pages": 26,
498
  "shards": 1
499
  },
500
  {
501
- "seconds": 1.072776924003847,
502
- "bytes": 6196482,
503
- "pages": 36,
504
- "shards": 2
505
- },
506
- {
507
- "seconds": 0.9420911959969089,
508
- "bytes": 7110283,
509
- "pages": 25,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.256377692996466,
514
- "bytes": 6950847,
515
- "pages": 37,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.3214142249998986,
520
- "bytes": 10934190,
521
  "pages": 36,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.2156976079932065,
526
- "bytes": 2375462,
527
- "pages": 28,
528
- "shards": 2
529
- },
530
- {
531
- "seconds": 0.9770713670004625,
532
- "bytes": 6199366,
533
- "pages": 37,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.0861811289942125,
538
- "bytes": 7695547,
539
  "pages": 38,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.0140557359991362,
544
- "bytes": 5030935,
545
- "pages": 25,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.2526760300024762,
550
- "bytes": 11691054,
551
- "pages": 30,
552
- "shards": 3
553
  },
554
  {
555
- "seconds": 0.9398190409992822,
556
- "bytes": 2819009,
557
  "pages": 37,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.1499773260002257,
562
- "bytes": 6017503,
563
- "pages": 37,
564
  "shards": 2
 
 
 
 
 
 
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": 75571.83458993434,
26
+ "required_pages_per_second_upper_estimate": 0.0065503750871918395
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": 9446.479323741793,
40
+ "required_pages_per_second_upper_estimate": 0.0037683418396927525
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": 18892.958647483585,
54
+ "required_pages_per_second_upper_estimate": 0.004370553478379578
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": 28339.437971225376,
68
+ "required_pages_per_second_upper_estimate": 0.0017257230869076143
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": 28339.437971225376,
82
+ "required_pages_per_second_upper_estimate": 0.0021816483836101666
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": 28339.437971225376,
96
+ "required_pages_per_second_upper_estimate": 0.35188097360499365
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": 717798,
161
+ "producer_batch_seconds": 0.44650785668744675,
162
+ "producer_p95_seconds": 0.43059379299666034,
163
+ "gpu_update_seconds": 0.9756015637315238,
164
+ "producer_batches_per_second": 2.2396022489252525,
165
+ "gpu_batches_per_second": 1.0250086071768438,
166
+ "producer_headroom": 2.1849594561881136,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.48729695135448,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 483988.12176279345,
174
+ "upload_bytes_per_second": 5716853.5664507225,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 103.85258426344731,
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": 1.4056276539995451,
184
+ "bytes": 7428876,
185
+ "pages": 30,
186
+ "shards": 3
187
  },
188
  {
189
+ "seconds": 1.1535106260016619,
190
+ "bytes": 12418046,
191
+ "pages": 38,
192
+ "shards": 3
193
+ },
194
+ {
195
+ "seconds": 0.8009545050008455,
196
+ "bytes": 2263071,
197
+ "pages": 25,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 2.506807755002228,
202
+ "bytes": 10223439,
203
+ "pages": 45,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 0.890203859002213,
208
+ "bytes": 916282,
209
  "pages": 26,
210
  "shards": 1
211
  },
212
  {
213
+ "seconds": 1.0841833719969145,
214
+ "bytes": 7541614,
215
  "pages": 38,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 0.8043565109983319,
220
+ "bytes": 540912,
221
+ "pages": 18,
222
+ "shards": 1
223
  },
224
  {
225
+ "seconds": 1.4004500939990976,
226
+ "bytes": 16241287,
227
+ "pages": 47,
228
+ "shards": 4
229
  },
230
  {
231
+ "seconds": 1.0899500639934558,
232
+ "bytes": 7493063,
233
+ "pages": 31,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 0.8779368290051934,
238
+ "bytes": 1248675,
239
+ "pages": 35,
240
+ "shards": 1
241
+ },
242
+ {
243
+ "seconds": 1.2020556749994284,
244
+ "bytes": 12102958,
245
  "pages": 36,
246
+ "shards": 4
247
  },
248
  {
249
+ "seconds": 1.048562693002168,
250
+ "bytes": 883869,
251
+ "pages": 26,
252
+ "shards": 1
253
  },
254
  {
255
+ "seconds": 1.072776924003847,
256
+ "bytes": 6196482,
257
+ "pages": 36,
258
+ "shards": 2
259
  },
260
  {
261
+ "seconds": 0.9420911959969089,
262
+ "bytes": 7110283,
263
+ "pages": 25,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.256377692996466,
268
+ "bytes": 6950847,
269
+ "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.3214142249998986,
274
+ "bytes": 10934190,
275
+ "pages": 36,
 
 
 
 
 
 
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.2156976079932065,
280
+ "bytes": 2375462,
281
+ "pages": 28,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 0.9770713670004625,
286
+ "bytes": 6199366,
287
  "pages": 37,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0861811289942125,
292
+ "bytes": 7695547,
293
+ "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.0140557359991362,
298
+ "bytes": 5030935,
299
+ "pages": 25,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 1.2526760300024762,
304
+ "bytes": 11691054,
305
+ "pages": 30,
306
+ "shards": 3
307
  },
308
  {
309
+ "seconds": 0.9398190409992822,
310
+ "bytes": 2819009,
311
  "pages": 37,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.1499773260002257,
316
+ "bytes": 6017503,
317
+ "pages": 37,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 1.6606815510021988,
322
+ "bytes": 10233181,
323
+ "pages": 23,
324
  "shards": 3
325
  },
326
  {
327
+ "seconds": 1.0018192189963884,
328
+ "bytes": 366522,
329
  "pages": 10,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 0.8059965610009385,
334
+ "bytes": 2452391,
335
+ "pages": 29,
336
+ "shards": 2
337
  },
338
  {
339
+ "seconds": 0.9446101400026237,
340
+ "bytes": 6258704,
341
+ "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.1053826549978112,
346
+ "bytes": 8176480,
347
+ "pages": 32,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.249701370994444,
352
+ "bytes": 10043413,
353
+ "pages": 32,
354
+ "shards": 3
355
  },
356
  {
357
+ "seconds": 1.0213671140008955,
358
+ "bytes": 6189327,
359
+ "pages": 36,
360
+ "shards": 2
361
  },
362
  {
363
+ "seconds": 1.0084248750063125,
364
+ "bytes": 2332449,
365
+ "pages": 29,
366
+ "shards": 2
367
  },
368
  {
369
+ "seconds": 0.9006169920030516,
370
+ "bytes": 6118351,
371
+ "pages": 36,
372
+ "shards": 2
373
  },
374
  {
375
+ "seconds": 1.270539192002616,
376
+ "bytes": 5502511,
377
+ "pages": 29,
378
+ "shards": 2
379
  },
380
  {
381
+ "seconds": 1.512666299000557,
382
+ "bytes": 13201899,
383
  "pages": 33,
384
+ "shards": 4
385
  },
386
  {
387
+ "seconds": 1.000070736001362,
388
+ "bytes": 6089030,
389
+ "pages": 37,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.0753602199984016,
394
+ "bytes": 871659,
395
+ "pages": 26,
396
+ "shards": 1
397
+ },
398
+ {
399
+ "seconds": 1.484286307997536,
400
+ "bytes": 11527011,
401
  "pages": 36,
402
+ "shards": 4
403
  },
404
  {
405
+ "seconds": 0.9214101749967085,
406
+ "bytes": 1181839,
407
+ "pages": 33,
408
+ "shards": 1
409
  },
410
  {
411
+ "seconds": 1.1708070859967847,
412
+ "bytes": 7634853,
413
+ "pages": 39,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 1.219036098998913,
418
+ "bytes": 6506587,
419
+ "pages": 23,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.234236194002733,
424
+ "bytes": 11430289,
425
+ "pages": 42,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.0073759259976214,
430
+ "bytes": 7448462,
431
+ "pages": 30,
432
  "shards": 3
433
  },
434
  {
435
+ "seconds": 0.7627367220047745,
436
+ "bytes": 1172696,
437
+ "pages": 36,
438
  "shards": 1
439
  },
440
  {
441
+ "seconds": 1.1723295699994196,
442
+ "bytes": 5714080,
443
+ "pages": 28,
444
+ "shards": 2
445
  },
446
  {
447
+ "seconds": 1.1845541919974494,
448
+ "bytes": 7669056,
449
  "pages": 38,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 0.9817348009964917,
454
+ "bytes": 4526983,
455
+ "pages": 22,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.2601451889931923,
460
+ "bytes": 13743089,
461
+ "pages": 49,
462
+ "shards": 4
463
  },
464
  {
465
+ "seconds": 0.8212157870002557,
466
+ "bytes": 890133,
467
  "pages": 26,
468
  "shards": 1
469
  },
470
  {
471
+ "seconds": 1.1462097159965197,
472
+ "bytes": 10417894,
473
+ "pages": 36,
474
+ "shards": 3
475
+ },
476
+ {
477
+ "seconds": 1.1751894160042866,
478
+ "bytes": 7506454,
479
  "pages": 38,
480
  "shards": 3
481
  },
482
  {
483
+ "seconds": 1.0263809830066748,
484
+ "bytes": 924946,
485
+ "pages": 27,
486
  "shards": 1
487
  },
488
  {
489
+ "seconds": 1.1823451020027278,
490
+ "bytes": 13453015,
491
+ "pages": 42,
492
  "shards": 4
493
  },
494
  {
495
+ "seconds": 1.2355715770027018,
496
+ "bytes": 9750066,
 
 
 
 
 
 
497
  "pages": 35,
 
 
 
 
 
 
498
  "shards": 4
499
  },
500
  {
501
+ "seconds": 1.382024664999335,
502
+ "bytes": 840789,
503
  "pages": 26,
504
  "shards": 1
505
  },
506
  {
507
+ "seconds": 1.7087262210043264,
508
+ "bytes": 7732105,
509
+ "pages": 39,
 
 
 
 
 
 
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.1756209270024556,
514
+ "bytes": 879254,
515
+ "pages": 27,
516
+ "shards": 1
517
  },
518
  {
519
+ "seconds": 1.4754508929981967,
520
+ "bytes": 10625968,
521
  "pages": 36,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 0.9398709419983788,
526
+ "bytes": 2450871,
527
+ "pages": 30,
 
 
 
 
 
 
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 1.1338124570029322,
532
+ "bytes": 11954791,
533
  "pages": 38,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.1746257199993124,
538
+ "bytes": 7528355,
539
+ "pages": 38,
540
  "shards": 3
541
  },
542
  {
543
+ "seconds": 1.2720379320016946,
544
+ "bytes": 4790673,
545
+ "pages": 27,
546
+ "shards": 2
547
  },
548
  {
549
+ "seconds": 0.9963719520019367,
550
+ "bytes": 6449131,
551
  "pages": 37,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 0.9008199800009606,
556
+ "bytes": 2360873,
557
+ "pages": 29,
558
  "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.193771635000303,
562
+ "bytes": 12283347,
563
+ "pages": 43,
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 sha256:62981a06b7db55eac80b0820e301dd2522b6f1ca97872b5e04d9d00e4779099f
3
- size 721770367
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44a83aad67663ec446e86253bc28a2f68145bbe9b144e53814d818c154a02730
3
+ size 721765631