appvoid commited on
Commit
174955c
·
verified ·
1 Parent(s): 062b8ec

Qwen90M step 410000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +45 -45
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +244 -244
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 405000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "971a9eeb4c3937f1740c96e56a608fda61840fb0477a14bea2f2a2fb4b6f0d33",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "6dc22034db3d8da70156cf3616a8261ff7bce22ec2f1aaee4adaf8904f0c628d",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
- "shard_scheduler.json": "8466082ba94a8cec381b73bc76af24a887c21c79f8a3381f5bbf4c50f884eaa8",
15
- "coverage.json": "869317a753fc4e9b828a15de2e9ed753b83822bd9b2bdfac8fc380952e35f060",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 410000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
+ "shard_scheduler.json": "dcd8a9cb4cda262678652187e892a303cca733af30618ef17353af63ba53c942",
15
+ "coverage.json": "551c8b6a0d7c9ec77f8c7a79a16a6436443df8715ce046c0113917bd6f5dc3f4",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 405000,
3
  "stats": {
4
- "climbmix/blocks": 14580000,
5
- "climbmix/targets": 24373460669,
6
- "climbmix/input_tokens": 24375004213,
7
- "ultra_qa/blocks": 5467500,
8
- "ultra_qa/targets": 8919175872,
9
- "ultra_qa/input_tokens": 8919193035,
10
- "cortex/blocks": 5467500,
11
- "cortex/targets": 207967668,
12
- "cortex/input_tokens": 8691301141,
13
- "ultra_style/blocks": 5467500,
14
- "ultra_style/targets": 8648596018,
15
- "ultra_style/input_tokens": 8649049645,
16
- "openmath/blocks": 3645000,
17
- "openmath/targets": 5632655091,
18
- "openmath/input_tokens": 5634532171,
19
- "rewrite6/blocks": 1822500,
20
- "rewrite6/targets": 2987706438,
21
- "rewrite6/input_tokens": 2990982166
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1996,
26
- "row": 2155,
27
  "start": {
28
- "file": 96,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 52,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 8177771
38
  },
39
  "rewrite6": {
40
- "page": 36,
41
- "row": 435,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 147456
47
  }
48
  },
49
  "epoch": 22,
50
- "offset": 2048,
51
- "documents": 4666779
52
  },
53
  "openmath": {
54
- "page": 1247,
55
- "row": 3991,
56
  "start": {
57
  "file": 11,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 304,
61
- "row": 562
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 5111703
67
  },
68
  "ultra_qa": {
69
- "page": 539,
70
- "row": 2341,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
- "row": 48
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2210085
82
  },
83
  "ultra_style": {
84
- "page": 823,
85
- "row": 2388,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
- "row": 35437
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 0,
96
- "documents": 3373396
97
  },
98
  "cortex": {
99
- "page": 106043,
100
- "row": 21,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 13573525
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 410000,
3
  "stats": {
4
+ "climbmix/blocks": 14760000,
5
+ "climbmix/targets": 24673977216,
6
+ "climbmix/input_tokens": 24675540036,
7
+ "ultra_qa/blocks": 5535000,
8
+ "ultra_qa/targets": 9029200380,
9
+ "ultra_qa/input_tokens": 9029217755,
10
+ "cortex/blocks": 5535000,
11
+ "cortex/targets": 210523889,
12
+ "cortex/input_tokens": 8798719668,
13
+ "ultra_style/blocks": 5535000,
14
+ "ultra_style/targets": 8755304288,
15
+ "ultra_style/input_tokens": 8755763253,
16
+ "openmath/blocks": 3690000,
17
+ "openmath/targets": 5702199677,
18
+ "openmath/input_tokens": 5704099937,
19
+ "rewrite6/blocks": 1845000,
20
+ "rewrite6/targets": 3024643728,
21
+ "rewrite6/input_tokens": 3027959756
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2021,
26
+ "row": 1375,
27
  "start": {
28
+ "file": 97,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 69,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 8279391
38
  },
39
  "rewrite6": {
40
+ "page": 50,
41
+ "row": 510,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 204800
47
  }
48
  },
49
  "epoch": 22,
50
+ "offset": 0,
51
+ "documents": 4724198
52
  },
53
  "openmath": {
54
+ "page": 1263,
55
+ "row": 1659,
56
  "start": {
57
  "file": 11,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 370,
61
+ "row": 98
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 5174907
67
  },
68
  "ultra_qa": {
69
+ "page": 546,
70
+ "row": 940,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 2,
76
+ "row": 28720
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2237356
82
  },
83
  "ultra_style": {
84
+ "page": 833,
85
+ "row": 2801,
86
  "start": {
87
  "file": 4,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
+ "row": 76397
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 0,
96
+ "documents": 3414769
97
  },
98
  "cortex": {
99
+ "page": 107353,
100
+ "row": 26,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 13741210
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 405000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "971a9eeb4c3937f1740c96e56a608fda61840fb0477a14bea2f2a2fb4b6f0d33", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "6dc22034db3d8da70156cf3616a8261ff7bce22ec2f1aaee4adaf8904f0c628d", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "8466082ba94a8cec381b73bc76af24a887c21c79f8a3381f5bbf4c50f884eaa8", "coverage.json": "869317a753fc4e9b828a15de2e9ed753b83822bd9b2bdfac8fc380952e35f060", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000405000"}
 
1
+ {"step": 410000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "dcd8a9cb4cda262678652187e892a303cca733af30618ef17353af63ba53c942", "coverage.json": "551c8b6a0d7c9ec77f8c7a79a16a6436443df8715ce046c0113917bd6f5dc3f4", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000410000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:971a9eeb4c3937f1740c96e56a608fda61840fb0477a14bea2f2a2fb4b6f0d33
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c6f9da68cb642941532a5ace8e8fec462efe7e94eb3fbc9d86a49411d6d84d5
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": 76765.55632890557,
26
- "required_pages_per_second_upper_estimate": 0.005310819307464605
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": 9595.694541113196,
40
- "required_pages_per_second_upper_estimate": 0.004017516923476006
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": 19191.389082226393,
54
- "required_pages_per_second_upper_estimate": 0.003997060234863171
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": 28787.08362333959,
68
- "required_pages_per_second_upper_estimate": 0.0017257417078876118
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": 28787.08362333959,
82
- "required_pages_per_second_upper_estimate": 0.0022380691307155136
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": 28787.08362333959,
96
- "required_pages_per_second_upper_estimate": 0.3692497995579789
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": 704566,
161
- "producer_batch_seconds": 0.06708011378908907,
162
- "producer_p95_seconds": 0.1769912210002076,
163
- "gpu_update_seconds": 0.9604307390688211,
164
- "producer_batches_per_second": 14.907547759149079,
165
- "gpu_batches_per_second": 1.0411994944784284,
166
- "producer_headroom": 14.3176671120233,
167
  "sustainable": true,
168
- "buffer_seconds": 60.507136561335734,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 432985.9895226683,
174
- "upload_bytes_per_second": 4974423.660555432,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 92.90871395359075,
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": 0.8249086619999844,
184
- "bytes": 105910,
185
- "pages": 3,
186
- "shards": 1
187
  },
188
  {
189
- "seconds": 1.08548185799998,
190
- "bytes": 2013983,
191
- "pages": 18,
192
  "shards": 3
193
  },
194
  {
195
- "seconds": 1.0607301400000324,
196
- "bytes": 873319,
197
- "pages": 26,
198
- "shards": 1
199
- },
200
- {
201
- "seconds": 1.4327509930000133,
202
- "bytes": 9634876,
203
- "pages": 38,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.0578138549999494,
208
- "bytes": 7567895,
209
- "pages": 37,
210
- "shards": 3
211
  },
212
  {
213
- "seconds": 0.8248641469999711,
214
- "bytes": 1159359,
215
- "pages": 34,
216
  "shards": 1
217
  },
218
  {
219
- "seconds": 0.9606663819999994,
220
- "bytes": 10228939,
221
- "pages": 23,
222
- "shards": 3
223
  },
224
  {
225
- "seconds": 1.4356544900001609,
226
- "bytes": 13537180,
227
- "pages": 46,
228
- "shards": 4
229
  },
230
  {
231
- "seconds": 0.9362600690001273,
232
- "bytes": 1025985,
233
- "pages": 34,
234
- "shards": 1
235
  },
236
  {
237
- "seconds": 1.7697618709999006,
238
- "bytes": 7336838,
239
- "pages": 30,
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 1.264844782000182,
244
- "bytes": 9996183,
245
- "pages": 40,
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 0.9787746179999885,
250
- "bytes": 1132423,
251
- "pages": 33,
252
- "shards": 1
253
  },
254
  {
255
- "seconds": 0.8866051049999442,
256
- "bytes": 7112636,
257
- "pages": 23,
258
- "shards": 3
259
  },
260
  {
261
- "seconds": 1.117540433999693,
262
- "bytes": 7055686,
263
- "pages": 40,
264
- "shards": 2
265
  },
266
  {
267
- "seconds": 1.0181373959999291,
268
- "bytes": 10411775,
269
- "pages": 36,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 1.183017940999889,
274
- "bytes": 7545162,
275
- "pages": 32,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 0.9269013600001017,
280
- "bytes": 1114116,
281
  "pages": 34,
282
  "shards": 1
283
  },
284
  {
285
- "seconds": 1.037127201000203,
286
- "bytes": 7612192,
287
- "pages": 37,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 0.9605875500001275,
292
- "bytes": 5330342,
293
- "pages": 25,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.2280645959999674,
298
- "bytes": 12018597,
299
- "pages": 45,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 1.1235430140000062,
304
- "bytes": 7362856,
305
- "pages": 29,
306
- "shards": 3
307
  },
308
  {
309
- "seconds": 0.8067839350001123,
310
- "bytes": 1162221,
311
- "pages": 35,
312
- "shards": 1
313
  },
314
  {
315
- "seconds": 1.3463661700002376,
316
- "bytes": 11675037,
317
- "pages": 32,
318
  "shards": 4
319
  },
320
  {
321
- "seconds": 0.8115306640002018,
322
- "bytes": 1206731,
323
- "pages": 34,
 
 
 
 
 
 
 
 
 
 
 
 
324
  "shards": 1
325
  },
326
  {
327
- "seconds": 1.1671833450000122,
328
- "bytes": 5654442,
329
- "pages": 19,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.1527475369998683,
334
- "bytes": 8325445,
335
- "pages": 38,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 1.049984759000381,
340
- "bytes": 5490081,
341
- "pages": 21,
 
 
 
 
 
 
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 1.3486151890001565,
346
- "bytes": 10508121,
347
- "pages": 37,
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.566957560999981,
352
- "bytes": 2658357,
353
- "pages": 37,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.1752439570000206,
358
- "bytes": 5996670,
359
  "pages": 36,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 1.1166780060002566,
364
- "bytes": 4727222,
365
- "pages": 30,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 2.7728241000004346,
370
- "bytes": 13625242,
371
- "pages": 41,
372
  "shards": 4
373
  },
374
  {
375
- "seconds": 3.518534920000093,
376
- "bytes": 6664316,
377
- "pages": 36,
 
 
 
 
 
 
378
  "shards": 3
379
  },
380
  {
381
- "seconds": 0.9569276100000934,
382
- "bytes": 2410711,
383
- "pages": 28,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.1522871959996337,
388
- "bytes": 11018272,
389
- "pages": 39,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 3.0842820179996124,
394
- "bytes": 6556297,
395
- "pages": 37,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 3.363522853000177,
400
- "bytes": 2328587,
401
- "pages": 27,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 1.091435188000105,
406
- "bytes": 12252514,
407
- "pages": 41,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.0064982160001819,
412
- "bytes": 6233391,
413
- "pages": 31,
414
  "shards": 3
415
  },
416
  {
417
- "seconds": 0.9982179830003588,
418
- "bytes": 6432855,
419
- "pages": 36,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 2.2747711609999897,
424
- "bytes": 6302843,
425
- "pages": 37,
426
- "shards": 2
427
  },
428
  {
429
- "seconds": 1.0353907819999222,
430
- "bytes": 2349590,
431
- "pages": 28,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.3265869809993092,
436
- "bytes": 11233813,
437
- "pages": 40,
438
- "shards": 3
439
- },
440
- {
441
- "seconds": 1.9781427679999979,
442
- "bytes": 8201204,
443
- "pages": 34,
444
- "shards": 3
445
  },
446
  {
447
- "seconds": 0.9253323299999465,
448
- "bytes": 6468324,
449
- "pages": 36,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.0003412309997657,
454
- "bytes": 1020572,
455
- "pages": 29,
456
  "shards": 1
457
  },
458
  {
459
- "seconds": 1.2060352290000083,
460
- "bytes": 11046742,
461
- "pages": 40,
462
- "shards": 4
463
- },
464
- {
465
- "seconds": 1.0752750450001258,
466
- "bytes": 6242403,
467
- "pages": 32,
468
- "shards": 2
469
  },
470
  {
471
- "seconds": 0.8482705019996502,
472
- "bytes": 2583323,
473
- "pages": 36,
474
- "shards": 2
475
  },
476
  {
477
- "seconds": 1.3519573450003008,
478
- "bytes": 11517949,
479
- "pages": 22,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 1.325917474000562,
484
- "bytes": 11191562,
485
- "pages": 43,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.1523795839993909,
490
- "bytes": 2573171,
491
  "pages": 37,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.718842794000011,
496
- "bytes": 5798754,
497
- "pages": 27,
498
  "shards": 2
499
  },
500
  {
501
- "seconds": 1.0535618110006908,
502
- "bytes": 1088694,
503
- "pages": 35,
504
- "shards": 1
505
  },
506
  {
507
- "seconds": 1.3156409709999934,
508
- "bytes": 11593993,
509
  "pages": 35,
510
- "shards": 4
511
  },
512
  {
513
- "seconds": 2.0578567419997853,
514
- "bytes": 11618530,
515
- "pages": 30,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 1.2497664069996972,
520
- "bytes": 1138477,
521
- "pages": 34,
522
- "shards": 1
523
  },
524
  {
525
- "seconds": 1.3184527379999054,
526
- "bytes": 7724848,
527
- "pages": 34,
528
  "shards": 3
529
  },
530
  {
531
- "seconds": 0.9524561609996454,
532
- "bytes": 5747185,
533
- "pages": 38,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.0773643330003324,
538
- "bytes": 7156662,
539
- "pages": 29,
540
- "shards": 3
541
- },
542
- {
543
- "seconds": 1.3278628770003706,
544
- "bytes": 6286061,
545
  "pages": 36,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 0.9090305769996121,
550
- "bytes": 7087614,
551
- "pages": 37,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.3239770070003942,
556
- "bytes": 12583619,
557
  "pages": 38,
558
- "shards": 4
559
  },
560
  {
561
- "seconds": 0.9289323200000581,
562
- "bytes": 883918,
563
- "pages": 25,
564
- "shards": 2
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": 76595.74770061958,
26
+ "required_pages_per_second_upper_estimate": 0.005299071552549478
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": 9574.468462577448,
40
+ "required_pages_per_second_upper_estimate": 0.004008630007644012
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": 19148.936925154896,
54
+ "required_pages_per_second_upper_estimate": 0.0039882185700838456
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": 28723.405387732342,
68
+ "required_pages_per_second_upper_estimate": 0.0017219242948939932
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": 28723.405387732342,
82
+ "required_pages_per_second_upper_estimate": 0.002233118427988008
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": 28723.405387732342,
96
+ "required_pages_per_second_upper_estimate": 0.3684330035239715
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
+ "producer_batch_seconds": 0.09953904679677805,
162
+ "producer_p95_seconds": 0.22590240799945605,
163
+ "gpu_update_seconds": 0.9625599620513087,
164
+ "producer_batches_per_second": 10.046308782136828,
165
+ "gpu_batches_per_second": 1.0388963175539765,
166
+ "producer_headroom": 9.670174600089354,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.641277609232446,
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": 352991.2228158892,
174
+ "upload_bytes_per_second": 5469736.43334308,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 75.74369920117863,
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": 1.0353907819999222,
184
+ "bytes": 2349590,
185
+ "pages": 28,
186
+ "shards": 2
187
  },
188
  {
189
+ "seconds": 1.3265869809993092,
190
+ "bytes": 11233813,
191
+ "pages": 40,
192
  "shards": 3
193
  },
194
  {
195
+ "seconds": 1.9781427679999979,
196
+ "bytes": 8201204,
197
+ "pages": 34,
 
 
 
 
 
 
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 0.9253323299999465,
202
+ "bytes": 6468324,
203
+ "pages": 36,
204
+ "shards": 2
205
  },
206
  {
207
+ "seconds": 1.0003412309997657,
208
+ "bytes": 1020572,
209
+ "pages": 29,
210
  "shards": 1
211
  },
212
  {
213
+ "seconds": 1.2060352290000083,
214
+ "bytes": 11046742,
215
+ "pages": 40,
216
+ "shards": 4
217
  },
218
  {
219
+ "seconds": 1.0752750450001258,
220
+ "bytes": 6242403,
221
+ "pages": 32,
222
+ "shards": 2
223
  },
224
  {
225
+ "seconds": 0.8482705019996502,
226
+ "bytes": 2583323,
227
+ "pages": 36,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 1.3519573450003008,
232
+ "bytes": 11517949,
233
+ "pages": 22,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.325917474000562,
238
+ "bytes": 11191562,
239
+ "pages": 43,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.1523795839993909,
244
+ "bytes": 2573171,
245
+ "pages": 37,
246
+ "shards": 2
247
  },
248
  {
249
+ "seconds": 1.718842794000011,
250
+ "bytes": 5798754,
251
+ "pages": 27,
252
+ "shards": 2
253
  },
254
  {
255
+ "seconds": 1.0535618110006908,
256
+ "bytes": 1088694,
257
+ "pages": 35,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.3156409709999934,
262
+ "bytes": 11593993,
263
+ "pages": 35,
264
+ "shards": 4
265
  },
266
  {
267
+ "seconds": 2.0578567419997853,
268
+ "bytes": 11618530,
269
+ "pages": 30,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.2497664069996972,
274
+ "bytes": 1138477,
275
  "pages": 34,
276
  "shards": 1
277
  },
278
  {
279
+ "seconds": 1.3184527379999054,
280
+ "bytes": 7724848,
281
+ "pages": 34,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 0.9524561609996454,
286
+ "bytes": 5747185,
287
+ "pages": 38,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0773643330003324,
292
+ "bytes": 7156662,
293
+ "pages": 29,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.3278628770003706,
298
+ "bytes": 6286061,
299
+ "pages": 36,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 0.9090305769996121,
304
+ "bytes": 7087614,
305
+ "pages": 37,
306
+ "shards": 2
307
  },
308
  {
309
+ "seconds": 1.3239770070003942,
310
+ "bytes": 12583619,
311
+ "pages": 38,
312
  "shards": 4
313
  },
314
  {
315
+ "seconds": 0.9289323200000581,
316
+ "bytes": 883918,
317
+ "pages": 25,
318
+ "shards": 2
319
+ },
320
+ {
321
+ "seconds": 1.464534527000069,
322
+ "bytes": 7593844,
323
+ "pages": 39,
324
+ "shards": 3
325
+ },
326
+ {
327
+ "seconds": 1.0184807959994941,
328
+ "bytes": 301114,
329
+ "pages": 8,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 1.0933929989987519,
334
+ "bytes": 5793673,
335
+ "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
+ "seconds": 1.0935950399998546,
340
+ "bytes": 4381259,
341
+ "pages": 34,
342
+ "shards": 2
343
  },
344
  {
345
+ "seconds": 1.3099932099994476,
346
+ "bytes": 13324819,
347
+ "pages": 42,
348
+ "shards": 4
349
+ },
350
+ {
351
+ "seconds": 0.9060626449991105,
352
+ "bytes": 6034341,
353
+ "pages": 27,
354
  "shards": 2
355
  },
356
  {
357
+ "seconds": 1.1673664749996533,
358
+ "bytes": 6863070,
359
+ "pages": 38,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 2.7079190219992597,
364
+ "bytes": 5909001,
365
+ "pages": 35,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 1.1004524960007984,
370
+ "bytes": 6145895,
371
  "pages": 36,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 0.9040275779989315,
376
+ "bytes": 2336556,
377
+ "pages": 28,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.34454566200111,
382
+ "bytes": 16421464,
383
+ "pages": 42,
384
  "shards": 4
385
  },
386
  {
387
+ "seconds": 0.9113021489993116,
388
+ "bytes": 908789,
389
+ "pages": 27,
390
+ "shards": 1
391
+ },
392
+ {
393
+ "seconds": 0.963736620000418,
394
+ "bytes": 7684006,
395
+ "pages": 38,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 1.1074004389993206,
400
+ "bytes": 6213956,
401
+ "pages": 36,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.0808812039995246,
406
+ "bytes": 6459806,
407
+ "pages": 30,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.1259296549997089,
412
+ "bytes": 5894764,
413
+ "pages": 28,
414
  "shards": 2
415
  },
416
  {
417
+ "seconds": 0.93207276099929,
418
+ "bytes": 6831941,
419
+ "pages": 31,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 1.0320983010005875,
424
+ "bytes": 7553870,
425
+ "pages": 38,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.168341065000277,
430
+ "bytes": 9926928,
431
+ "pages": 30,
432
  "shards": 3
433
  },
434
  {
435
+ "seconds": 0.8547535799989419,
436
+ "bytes": 1160518,
437
+ "pages": 33,
438
+ "shards": 1
439
  },
440
  {
441
+ "seconds": 0.9846142869992036,
442
+ "bytes": 7709639,
443
+ "pages": 38,
444
+ "shards": 3
445
  },
446
  {
447
+ "seconds": 1.0955427600001713,
448
+ "bytes": 6199869,
449
+ "pages": 37,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 1.6065525880003406,
454
+ "bytes": 12580162,
455
+ "pages": 33,
456
+ "shards": 4
 
 
 
 
 
 
457
  },
458
  {
459
+ "seconds": 0.995714869000949,
460
+ "bytes": 5809499,
461
+ "pages": 30,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 0.7974183579990495,
466
+ "bytes": 1155225,
467
+ "pages": 35,
468
  "shards": 1
469
  },
470
  {
471
+ "seconds": 1.29576844300027,
472
+ "bytes": 7786718,
473
+ "pages": 38,
474
+ "shards": 3
 
 
 
 
 
 
475
  },
476
  {
477
+ "seconds": 1.0071225869996852,
478
+ "bytes": 672738,
479
+ "pages": 21,
480
+ "shards": 1
481
  },
482
  {
483
+ "seconds": 1.3492830719987978,
484
+ "bytes": 12380695,
485
+ "pages": 48,
486
+ "shards": 4
487
  },
488
  {
489
+ "seconds": 1.1991475859995262,
490
+ "bytes": 11620692,
491
+ "pages": 34,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 0.9693775400010054,
496
+ "bytes": 6039517,
497
  "pages": 37,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 0.8906193579987303,
502
+ "bytes": 2228503,
503
+ "pages": 22,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.1276412609986437,
508
+ "bytes": 11020284,
509
+ "pages": 38,
510
+ "shards": 4
511
  },
512
  {
513
+ "seconds": 0.8274413319995801,
514
+ "bytes": 1262070,
515
  "pages": 35,
516
+ "shards": 1
517
  },
518
  {
519
+ "seconds": 1.0886287620014627,
520
+ "bytes": 7651274,
521
+ "pages": 38,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 1.1375053659994592,
526
+ "bytes": 11652818,
527
+ "pages": 36,
528
+ "shards": 3
529
  },
530
  {
531
+ "seconds": 2.7600006199991185,
532
+ "bytes": 6232730,
533
+ "pages": 30,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.0662739279996458,
538
+ "bytes": 6304947,
539
+ "pages": 36,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.1643560489992524,
544
+ "bytes": 6319425,
 
 
 
 
 
 
545
  "pages": 36,
546
  "shards": 2
547
  },
548
  {
549
+ "seconds": 2.530195315999663,
550
+ "bytes": 2235467,
551
+ "pages": 24,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.0548529980005696,
556
+ "bytes": 10543433,
557
  "pages": 38,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 1.3102050620000227,
562
+ "bytes": 8381851,
563
+ "pages": 37,
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 sha256:6dc22034db3d8da70156cf3616a8261ff7bce22ec2f1aaee4adaf8904f0c628d
3
- size 721764287
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2683a5571b60dc0874cc06273234cc7f53125ac1227bf84bc08e5b65eefaa4bc
3
+ size 721789695