appvoid commited on
Commit
d45a371
·
verified ·
1 Parent(s): 226c919

Qwen90M step 580000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +50 -50
  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": 575000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
- "shard_scheduler.json": "96d62a23720c59296a4eede4e35a800f12a60c2214c596aa38c0fd0f3fb5fd82",
15
- "coverage.json": "a6dd49afdfac1374a27b1c3438e7f2ed72df001cb3fb4b33fb5a8cf83ee869d9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 580000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667",
14
+ "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb",
15
+ "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 575000,
3
  "stats": {
4
- "climbmix/blocks": 20700000,
5
- "climbmix/targets": 34602598775,
6
- "climbmix/input_tokens": 34604791626,
7
- "ultra_qa/blocks": 7762500,
8
- "ultra_qa/targets": 12663846760,
9
- "ultra_qa/input_tokens": 12663871196,
10
- "cortex/blocks": 7762500,
11
- "cortex/targets": 295179734,
12
- "cortex/input_tokens": 12339109967,
13
- "ultra_style/blocks": 7762500,
14
- "ultra_style/targets": 12277961225,
15
- "ultra_style/input_tokens": 12278605070,
16
- "openmath/blocks": 5175000,
17
- "openmath/targets": 7996288244,
18
- "openmath/input_tokens": 7998951398,
19
- "rewrite6/blocks": 2587500,
20
- "rewrite6/targets": 4241858572,
21
- "rewrite6/input_tokens": 4246508964
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2835,
26
- "row": 588,
27
  "start": {
28
- "file": 137,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 16,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 0,
37
- "documents": 11612748
38
  },
39
  "rewrite6": {
40
- "page": 12,
41
- "row": 3116,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 49152
47
  }
48
  },
49
  "epoch": 32,
50
- "offset": 2048,
51
- "documents": 6625196
52
  },
53
  "openmath": {
54
- "page": 1771,
55
- "row": 1892,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 267,
61
- "row": 616
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 7255908
67
  },
68
  "ultra_qa": {
69
- "page": 766,
70
- "row": 264,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 0,
76
- "row": 19779
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 3137800
82
  },
83
  "ultra_style": {
84
- "page": 1169,
85
- "row": 508,
86
  "start": {
87
- "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 7,
91
- "row": 82661
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 4096,
96
- "documents": 4788732
97
  },
98
  "cortex": {
99
- "page": 150548,
100
- "row": 11,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 19270155
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 580000,
3
  "stats": {
4
+ "climbmix/blocks": 20880000,
5
+ "climbmix/targets": 34903766814,
6
+ "climbmix/input_tokens": 34905978789,
7
+ "ultra_qa/blocks": 7830000,
8
+ "ultra_qa/targets": 12773998197,
9
+ "ultra_qa/input_tokens": 12774022841,
10
+ "cortex/blocks": 7830000,
11
+ "cortex/targets": 297744461,
12
+ "cortex/input_tokens": 12446411636,
13
+ "ultra_style/blocks": 7830000,
14
+ "ultra_style/targets": 12384628359,
15
+ "ultra_style/input_tokens": 12385277999,
16
+ "openmath/blocks": 5220000,
17
+ "openmath/targets": 8065835620,
18
+ "openmath/input_tokens": 8068521894,
19
+ "rewrite6/blocks": 2610000,
20
+ "rewrite6/targets": 4278710471,
21
+ "rewrite6/input_tokens": 4283401756
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2859,
26
+ "row": 2966,
27
  "start": {
28
+ "file": 138,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 29,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 10240,
37
+ "documents": 11713430
38
  },
39
  "rewrite6": {
40
+ "page": 26,
41
+ "row": 3974,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 106496
47
  }
48
  },
49
  "epoch": 32,
50
+ "offset": 0,
51
+ "documents": 6683398
52
  },
53
  "openmath": {
54
+ "page": 1786,
55
+ "row": 3533,
56
  "start": {
57
  "file": 16,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 329,
61
+ "row": 56
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 7318989
67
  },
68
  "ultra_qa": {
69
+ "page": 772,
70
+ "row": 2970,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 0,
76
+ "row": 44355
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 3165082
82
  },
83
  "ultra_style": {
84
+ "page": 1179,
85
+ "row": 1648,
86
  "start": {
87
+ "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 0,
91
+ "row": 34590
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 4830832
97
  },
98
  "cortex": {
99
+ "page": 151857,
100
+ "row": 6,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 19437702
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 575000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "96d62a23720c59296a4eede4e35a800f12a60c2214c596aa38c0fd0f3fb5fd82", "coverage.json": "a6dd49afdfac1374a27b1c3438e7f2ed72df001cb3fb4b33fb5a8cf83ee869d9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000575000"}
 
1
+ {"step": 580000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "b7cb16c0913422fe54a766da97356133a598f4da7e691939f51028a8bce75667", "shard_scheduler.json": "4c7cecc17496ee854ce7e9db16f9fc5fe111120cc8b45911284ec118a736e4bb", "coverage.json": "e8919debdd5781a63f54d50ff90e0325881c51f855c15fdd97780f35b3c257bc", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000580000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b5de4855d786eac98b7c718616a9711c0e80f33fa5e572421d5de7b66c81490d
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86da1ad7765ecdf63d4c21ed8d3083e166f543b8c584d4e1ad8585460b37a1c2
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": 76608.2440593413,
26
- "required_pages_per_second_upper_estimate": 0.005299936079641394
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9576.030507417663,
40
- "required_pages_per_second_upper_estimate": 0.004009284003199404
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": 19152.061014835326,
54
- "required_pages_per_second_upper_estimate": 0.003988869235576494
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": 28728.09152225299,
68
- "required_pages_per_second_upper_estimate": 0.001722205221503193
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": 28728.09152225299,
82
- "required_pages_per_second_upper_estimate": 0.002233482754334861
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": 28728.09152225299,
96
- "required_pages_per_second_upper_estimate": 0.36849311222602316
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": 0.6385047882343144,
162
- "producer_p95_seconds": 0.21890624099978595,
163
- "gpu_update_seconds": 0.962402948994494,
164
- "producer_batches_per_second": 1.5661589676803274,
165
- "gpu_batches_per_second": 1.0390658102666737,
166
- "producer_headroom": 1.5072760090897195,
167
  "sustainable": true,
168
- "buffer_seconds": 60.63138578665312,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 310794.11682017095,
174
- "upload_bytes_per_second": 5262266.363159667,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 66.68918255285116,
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": 1.4154219219999504,
184
- "bytes": 7407761,
185
- "pages": 38,
186
  "shards": 3
187
  },
188
  {
189
- "seconds": 1.1865251550007088,
190
- "bytes": 5346992,
191
- "pages": 31,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 1.0492023780007003,
196
- "bytes": 6288876,
197
- "pages": 37,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 1.7196641009995801,
202
- "bytes": 2291853,
203
  "pages": 28,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.0487870670003758,
208
- "bytes": 11950097,
209
- "pages": 41,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.2208401020016026,
214
- "bytes": 11777262,
215
- "pages": 36,
216
- "shards": 4
217
- },
218
- {
219
- "seconds": 0.9635194839993346,
220
- "bytes": 1029009,
221
- "pages": 33,
222
- "shards": 1
223
- },
224
- {
225
- "seconds": 1.0835146409990557,
226
- "bytes": 5946250,
227
- "pages": 30,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 0.9391229799985013,
232
- "bytes": 2632765,
233
- "pages": 34,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.0898778940008924,
238
- "bytes": 10697456,
239
- "pages": 35,
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 1.3169437420001486,
244
- "bytes": 11722554,
245
- "pages": 35,
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 1.2027081380001619,
250
- "bytes": 2601682,
251
- "pages": 37,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.0705139929996221,
256
- "bytes": 6406862,
257
- "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 3.478892785999051,
262
- "bytes": 11471341,
263
- "pages": 38,
264
- "shards": 4
 
 
 
 
 
 
265
  },
266
  {
267
- "seconds": 0.9223496029990201,
268
- "bytes": 897417,
269
  "pages": 26,
270
  "shards": 1
271
  },
272
  {
273
- "seconds": 1.0840730929994606,
274
- "bytes": 6199927,
275
- "pages": 34,
276
  "shards": 2
277
  },
278
  {
279
- "seconds": 1.112097929999436,
280
- "bytes": 8142857,
281
- "pages": 30,
282
- "shards": 3
283
  },
284
  {
285
- "seconds": 1.17854872099997,
286
- "bytes": 10442242,
287
- "pages": 43,
288
- "shards": 4
289
  },
290
  {
291
- "seconds": 1.163782216999607,
292
- "bytes": 7671664,
293
- "pages": 39,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 1.1290235629985546,
298
- "bytes": 1186173,
299
  "pages": 34,
300
- "shards": 1
301
- },
302
- {
303
- "seconds": 1.2037071409995406,
304
- "bytes": 10277758,
305
- "pages": 28,
306
  "shards": 3
307
  },
308
  {
309
- "seconds": 0.9222849990001123,
310
- "bytes": 2428641,
311
- "pages": 31,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.1589227389995358,
316
- "bytes": 11976819,
317
- "pages": 37,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.0021623419997923,
322
- "bytes": 6273488,
323
- "pages": 37,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.0131233649990463,
328
- "bytes": 2204351,
329
- "pages": 24,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.1051343430008274,
334
- "bytes": 10092152,
335
- "pages": 38,
336
- "shards": 3
337
  },
338
  {
339
- "seconds": 0.975858148000043,
340
- "bytes": 861206,
341
- "pages": 26,
342
- "shards": 1
343
  },
344
  {
345
- "seconds": 1.055939247000424,
346
- "bytes": 7659150,
347
  "pages": 38,
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.1426443119999021,
352
- "bytes": 11681956,
353
  "pages": 27,
354
- "shards": 3
355
  },
356
  {
357
- "seconds": 2.1566952029988897,
358
- "bytes": 6925851,
359
- "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 0.9256938060007087,
364
- "bytes": 6110229,
365
- "pages": 34,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 0.7895571030003339,
370
- "bytes": 1089032,
371
- "pages": 34,
 
 
 
 
 
 
372
  "shards": 1
373
  },
374
  {
375
- "seconds": 1.843757688999176,
376
- "bytes": 12452068,
377
- "pages": 39,
378
- "shards": 4
379
  },
380
  {
381
- "seconds": 2.206128002000696,
382
- "bytes": 6045751,
383
- "pages": 29,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.4661580640004104,
388
- "bytes": 8583866,
389
- "pages": 41,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.0245029070010787,
394
- "bytes": 5870358,
395
- "pages": 27,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 0.900488690000202,
400
- "bytes": 1094466,
401
- "pages": 33,
402
- "shards": 1
403
  },
404
  {
405
- "seconds": 1.2211129879997316,
406
- "bytes": 12049529,
407
- "pages": 44,
408
- "shards": 4
409
  },
410
  {
411
- "seconds": 1.1176546409988077,
412
- "bytes": 5812216,
413
- "pages": 27,
 
 
 
 
 
 
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 0.8849807390015485,
418
- "bytes": 1252571,
419
- "pages": 35,
420
- "shards": 1
421
  },
422
  {
423
- "seconds": 1.2397414250008296,
424
- "bytes": 13176621,
425
  "pages": 29,
426
- "shards": 4
427
  },
428
  {
429
- "seconds": 1.2003755249988899,
430
- "bytes": 10273667,
431
- "pages": 44,
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 1.0975800400010485,
436
- "bytes": 2683103,
437
- "pages": 36,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.1853528239989828,
442
- "bytes": 5804023,
443
  "pages": 28,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.138313076000486,
448
- "bytes": 10599364,
449
- "pages": 39,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 0.9668768979991,
454
- "bytes": 2353349,
455
- "pages": 29,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.4940221720007685,
460
- "bytes": 12237361,
461
- "pages": 41,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 0.8858112369998707,
466
- "bytes": 2416252,
467
- "pages": 28,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.2992984630000137,
472
- "bytes": 10017513,
473
- "pages": 40,
474
- "shards": 3
475
  },
476
  {
477
- "seconds": 4.1799396490005165,
478
- "bytes": 5932760,
479
- "pages": 32,
480
- "shards": 2
481
  },
482
  {
483
- "seconds": 0.8289485320001404,
484
- "bytes": 2513176,
485
- "pages": 33,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 0.9106573579992983,
490
- "bytes": 5965787,
491
  "pages": 36,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.575513838999541,
496
- "bytes": 18103820,
497
- "pages": 39,
498
- "shards": 5
499
  },
500
  {
501
- "seconds": 0.8145781550010724,
502
- "bytes": 930729,
503
- "pages": 26,
504
  "shards": 1
505
  },
506
  {
507
- "seconds": 1.3511633730013273,
508
- "bytes": 6255780,
509
  "pages": 36,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.031140414001129,
514
- "bytes": 2762787,
515
- "pages": 38,
516
- "shards": 2
517
  },
518
  {
519
- "seconds": 1.0752210870014096,
520
- "bytes": 5610482,
521
- "pages": 19,
522
- "shards": 2
523
  },
524
  {
525
- "seconds": 1.425927022999531,
526
- "bytes": 10079783,
527
- "pages": 46,
528
  "shards": 3
529
  },
530
  {
531
- "seconds": 1.1401308700005757,
532
- "bytes": 8309975,
533
- "pages": 34,
534
- "shards": 3
535
  },
536
  {
537
- "seconds": 1.2373918849989423,
538
- "bytes": 6097119,
539
- "pages": 33,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 1.1592294109977956,
544
- "bytes": 6491968,
545
- "pages": 35,
546
- "shards": 3
547
  },
548
  {
549
- "seconds": 1.6240387940015353,
550
- "bytes": 5730323,
551
- "pages": 27,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.3926623539991851,
556
- "bytes": 6408754,
557
  "pages": 36,
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.260699045000365,
562
- "bytes": 2670941,
563
- "pages": 37,
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": 76844.82269293365,
26
+ "required_pages_per_second_upper_estimate": 0.0053163031384513715
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": 9605.602836616707,
40
+ "required_pages_per_second_upper_estimate": 0.004021665319894588
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": 19211.205673233413,
54
+ "required_pages_per_second_upper_estimate": 0.004001187508171285
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": 28816.808509850118,
68
+ "required_pages_per_second_upper_estimate": 0.0017275236694466443
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": 28816.808509850118,
82
+ "required_pages_per_second_upper_estimate": 0.0022403801098957493
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": 28816.808509850118,
96
+ "required_pages_per_second_upper_estimate": 0.3696310784860394
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": 0.0661616753671126,
162
+ "producer_p95_seconds": 0.24438537600144628,
163
+ "gpu_update_seconds": 0.9594400431452845,
164
+ "producer_batches_per_second": 15.11449029141539,
165
+ "gpu_batches_per_second": 1.0422746133481668,
166
+ "producer_headroom": 14.501447217314565,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.444722718152924,
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": 406720.83803339687,
174
+ "upload_bytes_per_second": 5366265.038987921,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 87.2728238654273,
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": 1.2003755249988899,
184
+ "bytes": 10273667,
185
+ "pages": 44,
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.0975800400010485,
190
+ "bytes": 2683103,
191
+ "pages": 36,
 
 
 
 
 
 
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.1853528239989828,
196
+ "bytes": 5804023,
197
  "pages": 28,
198
  "shards": 2
199
  },
200
  {
201
+ "seconds": 1.138313076000486,
202
+ "bytes": 10599364,
203
+ "pages": 39,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 0.9668768979991,
208
+ "bytes": 2353349,
209
+ "pages": 29,
 
 
 
 
 
 
 
 
 
 
 
 
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.4940221720007685,
214
+ "bytes": 12237361,
215
+ "pages": 41,
216
+ "shards": 3
217
  },
218
  {
219
+ "seconds": 0.8858112369998707,
220
+ "bytes": 2416252,
221
+ "pages": 28,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.2992984630000137,
226
+ "bytes": 10017513,
227
+ "pages": 40,
228
  "shards": 3
229
  },
230
  {
231
+ "seconds": 4.1799396490005165,
232
+ "bytes": 5932760,
233
+ "pages": 32,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 0.8289485320001404,
238
+ "bytes": 2513176,
239
+ "pages": 33,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 0.9106573579992983,
244
+ "bytes": 5965787,
245
+ "pages": 36,
246
+ "shards": 2
247
+ },
248
+ {
249
+ "seconds": 1.575513838999541,
250
+ "bytes": 18103820,
251
+ "pages": 39,
252
+ "shards": 5
253
  },
254
  {
255
+ "seconds": 0.8145781550010724,
256
+ "bytes": 930729,
257
  "pages": 26,
258
  "shards": 1
259
  },
260
  {
261
+ "seconds": 1.3511633730013273,
262
+ "bytes": 6255780,
263
+ "pages": 36,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 1.031140414001129,
268
+ "bytes": 2762787,
269
+ "pages": 38,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 1.0752210870014096,
274
+ "bytes": 5610482,
275
+ "pages": 19,
276
+ "shards": 2
277
  },
278
  {
279
+ "seconds": 1.425927022999531,
280
+ "bytes": 10079783,
281
+ "pages": 46,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.1401308700005757,
286
+ "bytes": 8309975,
287
  "pages": 34,
 
 
 
 
 
 
288
  "shards": 3
289
  },
290
  {
291
+ "seconds": 1.2373918849989423,
292
+ "bytes": 6097119,
293
+ "pages": 33,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.1592294109977956,
298
+ "bytes": 6491968,
299
+ "pages": 35,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.6240387940015353,
304
+ "bytes": 5730323,
305
+ "pages": 27,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.3926623539991851,
310
+ "bytes": 6408754,
311
+ "pages": 36,
312
  "shards": 2
313
  },
314
  {
315
+ "seconds": 1.260699045000365,
316
+ "bytes": 2670941,
317
+ "pages": 37,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 1.4827766799971869,
322
+ "bytes": 15946512,
323
+ "pages": 34,
324
+ "shards": 4
325
  },
326
  {
327
+ "seconds": 1.33145204000175,
328
+ "bytes": 7515329,
329
  "pages": 38,
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 0.9712299829989206,
334
+ "bytes": 867774,
335
  "pages": 27,
336
+ "shards": 1
337
  },
338
  {
339
+ "seconds": 1.1689961140000378,
340
+ "bytes": 10341916,
341
+ "pages": 38,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 4.051943458998721,
346
+ "bytes": 2586599,
347
+ "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 0.9690040009991208,
352
+ "bytes": 5902816,
353
+ "pages": 29,
354
+ "shards": 2
355
+ },
356
+ {
357
+ "seconds": 0.9907326490028936,
358
+ "bytes": 186236,
359
+ "pages": 7,
360
  "shards": 1
361
  },
362
  {
363
+ "seconds": 1.1790409999994154,
364
+ "bytes": 12018438,
365
+ "pages": 30,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 1.0833680409996305,
370
+ "bytes": 2366981,
371
+ "pages": 28,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.2835563219996402,
376
+ "bytes": 10401632,
377
+ "pages": 38,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 0.9978380930006097,
382
+ "bytes": 912994,
383
+ "pages": 28,
384
+ "shards": 1
385
  },
386
  {
387
+ "seconds": 1.1661826020026638,
388
+ "bytes": 7439707,
389
+ "pages": 37,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 1.0849376400001347,
394
+ "bytes": 10733196,
395
+ "pages": 39,
396
+ "shards": 3
397
  },
398
  {
399
+ "seconds": 1.2397475959987787,
400
+ "bytes": 8159623,
401
+ "pages": 29,
402
+ "shards": 3
403
+ },
404
+ {
405
+ "seconds": 1.2689961300020514,
406
+ "bytes": 6434173,
407
+ "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.3497664780006744,
412
+ "bytes": 7459096,
413
+ "pages": 39,
414
+ "shards": 4
415
  },
416
  {
417
+ "seconds": 1.0805805290001445,
418
+ "bytes": 1076352,
419
  "pages": 29,
420
+ "shards": 1
421
  },
422
  {
423
+ "seconds": 1.1792509940023592,
424
+ "bytes": 9839539,
425
+ "pages": 37,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.0725541200008593,
430
+ "bytes": 2601868,
431
+ "pages": 38,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.330655336001655,
436
+ "bytes": 11531295,
437
  "pages": 28,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 1.5732356919979793,
442
+ "bytes": 10605987,
443
+ "pages": 41,
444
  "shards": 3
445
  },
446
  {
447
+ "seconds": 0.9885955910031043,
448
+ "bytes": 2380823,
449
+ "pages": 28,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 0.9730731119998381,
454
+ "bytes": 6090581,
455
+ "pages": 37,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 1.1632558510027593,
460
+ "bytes": 7851564,
461
+ "pages": 38,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.0146159530013392,
466
+ "bytes": 5636542,
467
+ "pages": 31,
468
+ "shards": 2
469
  },
470
  {
471
+ "seconds": 1.0724632399978873,
472
+ "bytes": 12061908,
473
+ "pages": 38,
474
+ "shards": 3
475
  },
476
  {
477
+ "seconds": 0.9588815639981476,
478
+ "bytes": 2343317,
479
+ "pages": 28,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.1246145429977332,
484
+ "bytes": 6175304,
485
  "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.2072375589996227,
490
+ "bytes": 11138063,
491
+ "pages": 36,
492
+ "shards": 4
493
  },
494
  {
495
+ "seconds": 0.9064097570008016,
496
+ "bytes": 1171302,
497
+ "pages": 35,
498
  "shards": 1
499
  },
500
  {
501
+ "seconds": 0.9584431869989203,
502
+ "bytes": 6298065,
503
  "pages": 36,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 0.972795835998113,
508
+ "bytes": 8254778,
509
+ "pages": 31,
510
+ "shards": 3
511
  },
512
  {
513
+ "seconds": 1.3151332330016885,
514
+ "bytes": 10337557,
515
+ "pages": 35,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 1.029109121998772,
520
+ "bytes": 7643162,
521
+ "pages": 40,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 0.8460896450014843,
526
+ "bytes": 884554,
527
+ "pages": 26,
528
+ "shards": 1
529
  },
530
  {
531
+ "seconds": 1.0631770070031052,
532
+ "bytes": 6064387,
533
+ "pages": 32,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.2800072050013114,
538
+ "bytes": 12000660,
539
+ "pages": 43,
540
+ "shards": 4
541
  },
542
  {
543
+ "seconds": 1.2026952199994412,
544
+ "bytes": 6686271,
545
+ "pages": 28,
546
  "shards": 2
547
  },
548
  {
549
+ "seconds": 1.3011887590000697,
550
+ "bytes": 6177443,
551
  "pages": 36,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.0062806160021864,
556
+ "bytes": 2394266,
557
+ "pages": 29,
558
  "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.1657910530011577,
562
+ "bytes": 10819661,
563
+ "pages": 44,
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 sha256:99765a8124f03fc648201052c5c018c7fdee5c607bfc1d285695d753f55e94fa
3
- size 721756799
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1330a58909395b69dd9e933634473173544669f01ca72451c097b4fe93ebb964
3
+ size 721773055