appvoid commited on
Commit
ed98763
·
verified ·
1 Parent(s): 1d5433c

Qwen90M step 425000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +46 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +241 -241
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 420000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "cfce39253aa3ae4c325817f7ed5dea73b4ff79ab53ddb17407c5d79b88e8a80a",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "9af8f113fcd330718ea1f4faa81d21b284817f06e6050316baaf832fa2400fa0",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
- "shard_scheduler.json": "b660820a9781b36ec51f7d77448e3fea7f7e2d6ad47c4a9335f882c29c9b5655",
15
- "coverage.json": "c74982ad4eb4d14007b9313e674f56558ace0621e80565c1e8d7fd9ccd3565be",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 425000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e6a8119feebeb9b4f48e3d9e2dc5415b553b928301f4bcae69173e0b92d62b00",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "b280b0d64c4eb0d14a1021d0c9cdbedc4c0bf8eb245a63c69257180b96af6158",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
+ "shard_scheduler.json": "53c18fdc60e1935cc354da354050e76668d361282678ac8d9bee2b8f0811a722",
15
+ "coverage.json": "3b07084bb276dfdd5fdda54e8af8ef1840ed097f7c487c84187259f480a569a9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 420000,
3
  "stats": {
4
- "climbmix/blocks": 15120000,
5
- "climbmix/targets": 25276172773,
6
- "climbmix/input_tokens": 25277773647,
7
- "ultra_qa/blocks": 5670000,
8
- "ultra_qa/targets": 9249409599,
9
- "ultra_qa/input_tokens": 9249427358,
10
- "cortex/blocks": 5670000,
11
- "cortex/targets": 215646829,
12
- "cortex/input_tokens": 9013142666,
13
- "ultra_style/blocks": 5670000,
14
- "ultra_style/targets": 8968862759,
15
- "ultra_style/input_tokens": 8969333094,
16
- "openmath/blocks": 3780000,
17
- "openmath/targets": 5841040378,
18
- "openmath/input_tokens": 5842986747,
19
- "rewrite6/blocks": 1890000,
20
- "rewrite6/targets": 3098343100,
21
- "rewrite6/input_tokens": 3101740258
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2070,
26
- "row": 2319,
27
  "start": {
28
- "file": 100,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 17,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 8481039
38
  },
39
  "rewrite6": {
40
- "page": 28,
41
- "row": 741,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 114688
47
  }
48
  },
49
  "epoch": 23,
50
- "offset": 2048,
51
- "documents": 4839721
52
  },
53
  "openmath": {
54
- "page": 1294,
55
- "row": 601,
56
  "start": {
57
  "file": 12,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 60,
61
- "row": 424
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 5300825
67
  },
68
  "ultra_qa": {
69
- "page": 559,
70
- "row": 2275,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
- "row": 17426
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 2291939
82
  },
83
  "ultra_style": {
84
- "page": 854,
85
- "row": 414,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 0,
91
- "row": 73272
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 3498398
97
  },
98
  "cortex": {
99
- "page": 109970,
100
- "row": 5,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 14076165
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 425000,
3
  "stats": {
4
+ "climbmix/blocks": 15300000,
5
+ "climbmix/targets": 25576792778,
6
+ "climbmix/input_tokens": 25578412852,
7
+ "ultra_qa/blocks": 5737500,
8
+ "ultra_qa/targets": 9359436204,
9
+ "ultra_qa/input_tokens": 9359454150,
10
+ "cortex/blocks": 5737500,
11
+ "cortex/targets": 218211819,
12
+ "cortex/input_tokens": 9120392161,
13
+ "ultra_style/blocks": 5737500,
14
+ "ultra_style/targets": 9075454296,
15
+ "ultra_style/input_tokens": 9075929780,
16
+ "openmath/blocks": 3825000,
17
+ "openmath/targets": 5910673230,
18
+ "openmath/input_tokens": 5912642746,
19
+ "rewrite6/blocks": 1912500,
20
+ "rewrite6/targets": 3135273649,
21
+ "rewrite6/input_tokens": 3138711114
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2095,
26
+ "row": 1309,
27
  "start": {
28
+ "file": 101,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 34,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 8582429
38
  },
39
  "rewrite6": {
40
+ "page": 42,
41
+ "row": 888,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 172032
47
  }
48
  },
49
  "epoch": 23,
50
+ "offset": 0,
51
+ "documents": 4897212
52
  },
53
  "openmath": {
54
+ "page": 1309,
55
+ "row": 2292,
56
  "start": {
57
  "file": 12,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 121,
61
+ "row": 864
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 5363956
67
  },
68
  "ultra_qa": {
69
+ "page": 566,
70
+ "row": 874,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
+ "row": 46098
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 2319210
82
  },
83
  "ultra_style": {
84
+ "page": 864,
85
+ "row": 304,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 1,
91
+ "row": 29158
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 3539248
97
  },
98
  "cortex": {
99
+ "page": 111278,
100
+ "row": 123,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 14243707
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 420000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "cfce39253aa3ae4c325817f7ed5dea73b4ff79ab53ddb17407c5d79b88e8a80a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9af8f113fcd330718ea1f4faa81d21b284817f06e6050316baaf832fa2400fa0", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "b660820a9781b36ec51f7d77448e3fea7f7e2d6ad47c4a9335f882c29c9b5655", "coverage.json": "c74982ad4eb4d14007b9313e674f56558ace0621e80565c1e8d7fd9ccd3565be", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000420000"}
 
1
+ {"step": 425000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e6a8119feebeb9b4f48e3d9e2dc5415b553b928301f4bcae69173e0b92d62b00", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "b280b0d64c4eb0d14a1021d0c9cdbedc4c0bf8eb245a63c69257180b96af6158", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "53c18fdc60e1935cc354da354050e76668d361282678ac8d9bee2b8f0811a722", "coverage.json": "3b07084bb276dfdd5fdda54e8af8ef1840ed097f7c487c84187259f480a569a9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000425000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cfce39253aa3ae4c325817f7ed5dea73b4ff79ab53ddb17407c5d79b88e8a80a
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6a8119feebeb9b4f48e3d9e2dc5415b553b928301f4bcae69173e0b92d62b00
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": 77052.14963302753,
26
- "required_pages_per_second_upper_estimate": 0.005330646497231844
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": 9631.51870412844,
40
- "required_pages_per_second_upper_estimate": 0.004032515752436897
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": 19263.03740825688,
54
- "required_pages_per_second_upper_estimate": 0.004011982691681902
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": 28894.556112385326,
68
- "required_pages_per_second_upper_estimate": 0.0017321845195049147
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": 28894.556112385326,
82
- "required_pages_per_second_upper_estimate": 0.002246424644017299
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": 28894.556112385326,
96
- "required_pages_per_second_upper_estimate": 0.3706283412524894
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": 0.09326356839858363,
162
- "producer_p95_seconds": 0.27388932800022303,
163
- "gpu_update_seconds": 0.956858443938822,
164
- "producer_batches_per_second": 10.722300434895077,
165
- "gpu_batches_per_second": 1.0450866649444923,
166
- "producer_headroom": 10.259723709578259,
167
  "sustainable": true,
168
- "buffer_seconds": 60.28208196814579,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 376352.5842703375,
174
- "upload_bytes_per_second": 5020670.556623916,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 80.75650354464143,
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": 1.2489348130002327,
184
- "bytes": 8291174,
185
- "pages": 32,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.122991837999507,
190
- "bytes": 10369739,
191
- "pages": 39,
192
  "shards": 3
193
  },
194
  {
195
- "seconds": 1.367829029000859,
196
- "bytes": 6101351,
197
- "pages": 37,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 0.9580737389987917,
202
- "bytes": 2350217,
203
- "pages": 28,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.1762084339989087,
208
- "bytes": 10342096,
209
- "pages": 40,
210
- "shards": 3
211
  },
212
  {
213
- "seconds": 1.164775548999387,
214
- "bytes": 5765223,
215
- "pages": 29,
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 0.9962533819998498,
220
- "bytes": 8524201,
221
- "pages": 41,
222
  "shards": 3
223
  },
224
  {
225
- "seconds": 1.0446712169996317,
226
- "bytes": 5880079,
227
  "pages": 28,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.5804328610010998,
232
- "bytes": 12088466,
233
- "pages": 42,
234
- "shards": 4
235
  },
236
  {
237
- "seconds": 0.7921914839989768,
238
- "bytes": 902002,
239
- "pages": 27,
240
- "shards": 1
241
  },
242
  {
243
- "seconds": 1.0630905629986955,
244
- "bytes": 6092917,
245
  "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 0.9817484880004486,
250
- "bytes": 2589304,
251
- "pages": 36,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.2340881659984007,
256
- "bytes": 16290946,
257
- "pages": 38,
258
- "shards": 4
259
- },
260
- {
261
- "seconds": 1.10198693500206,
262
- "bytes": 7334244,
263
- "pages": 30,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 0.7830869729987171,
268
- "bytes": 1161196,
269
- "pages": 35,
270
  "shards": 1
271
  },
272
  {
273
- "seconds": 1.0245182449980348,
274
- "bytes": 6075842,
275
- "pages": 36,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 1.3879416819981998,
280
- "bytes": 11260844,
281
  "pages": 36,
282
- "shards": 4
283
  },
284
  {
285
- "seconds": 1.0456256510005915,
286
- "bytes": 862691,
287
- "pages": 26,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 2.013755919000687,
292
- "bytes": 11740969,
293
- "pages": 33,
294
- "shards": 3
295
  },
296
  {
297
- "seconds": 0.9664811299990106,
298
- "bytes": 2719675,
299
  "pages": 37,
300
  "shards": 2
301
  },
302
  {
303
- "seconds": 1.2011403209980926,
304
- "bytes": 11095830,
305
- "pages": 34,
 
 
 
 
 
 
306
  "shards": 3
307
  },
308
  {
309
- "seconds": 1.0495139129998279,
310
- "bytes": 7453354,
311
  "pages": 39,
312
  "shards": 3
313
  },
314
  {
315
- "seconds": 1.6562018110016652,
316
- "bytes": 1121250,
317
- "pages": 33,
318
- "shards": 1
319
  },
320
  {
321
- "seconds": 0.9384285160012951,
322
- "bytes": 5824612,
323
- "pages": 28,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 0.8717172680007934,
328
- "bytes": 283568,
329
- "pages": 8,
 
 
 
 
 
 
330
  "shards": 1
331
  },
332
  {
333
- "seconds": 14.44776459899731,
334
- "bytes": 11619395,
335
- "pages": 28,
336
- "shards": 4
337
  },
338
  {
339
- "seconds": 1.2608952039990982,
340
- "bytes": 6128515,
341
- "pages": 36,
 
 
 
 
 
 
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 0.9641569820014411,
346
- "bytes": 6035455,
347
  "pages": 34,
 
 
 
 
 
 
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.0038025159992685,
352
- "bytes": 2414763,
353
  "pages": 27,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.1136625049985014,
358
- "bytes": 10469225,
359
  "pages": 39,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 0.9355267249993631,
364
- "bytes": 7522148,
365
- "pages": 39,
 
 
 
 
 
 
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 1.2877170510000724,
370
- "bytes": 6767236,
371
- "pages": 27,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.0314974919965607,
376
- "bytes": 5945434,
377
- "pages": 28,
378
- "shards": 2
379
  },
380
  {
381
- "seconds": 1.1649709810008062,
382
- "bytes": 7159686,
383
- "pages": 37,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.206743112001277,
388
- "bytes": 6257240,
389
- "pages": 36,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.0216239910005243,
394
- "bytes": 7839111,
395
- "pages": 37,
396
- "shards": 4
397
- },
398
- {
399
- "seconds": 1.0000162349970196,
400
- "bytes": 885768,
401
- "pages": 25,
402
- "shards": 1
403
  },
404
  {
405
- "seconds": 1.2385705720007536,
406
- "bytes": 15798086,
407
  "pages": 39,
408
  "shards": 4
409
  },
410
  {
411
- "seconds": 0.9273680320002313,
412
- "bytes": 2684614,
413
- "pages": 37,
414
- "shards": 2
415
- },
416
- {
417
- "seconds": 1.003430278000451,
418
- "bytes": 5945767,
419
- "pages": 28,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 0.8994475729996338,
424
- "bytes": 6104947,
425
  "pages": 37,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 1.0381844509975053,
430
- "bytes": 7741032,
431
  "pages": 38,
432
- "shards": 3
433
  },
434
  {
435
- "seconds": 1.5200233909999952,
436
- "bytes": 6160062,
437
  "pages": 28,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.1740434719977202,
442
- "bytes": 8468925,
443
- "pages": 41,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.0248841799984802,
448
- "bytes": 5963815,
449
- "pages": 27,
450
- "shards": 2
451
  },
452
  {
453
- "seconds": 1.231629100999271,
454
- "bytes": 4969209,
455
  "pages": 37,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.249853739001992,
460
- "bytes": 7694709,
461
- "pages": 39,
462
- "shards": 3
463
- },
464
- {
465
- "seconds": 1.2034739760019875,
466
- "bytes": 5697150,
467
  "pages": 28,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 0.9529857010020351,
472
- "bytes": 6624046,
473
- "pages": 40,
474
- "shards": 3
475
- },
476
- {
477
- "seconds": 1.1070565920017543,
478
- "bytes": 11831537,
479
- "pages": 32,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.0998464429976593,
484
- "bytes": 6314908,
485
- "pages": 36,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.0093370079994202,
490
- "bytes": 2319025,
491
- "pages": 29,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.116415386000881,
496
- "bytes": 10937725,
497
- "pages": 41,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 0.8731399669995881,
502
- "bytes": 778500,
503
- "pages": 25,
504
- "shards": 1
505
  },
506
  {
507
- "seconds": 1.033372367000993,
508
- "bytes": 7683491,
509
- "pages": 38,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 1.4393485709988454,
514
- "bytes": 11721944,
515
- "pages": 36,
516
- "shards": 3
517
  },
518
  {
519
- "seconds": 1.162564084999758,
520
- "bytes": 5892604,
521
- "pages": 23,
522
- "shards": 3
523
  },
524
  {
525
- "seconds": 1.082973058000789,
526
- "bytes": 6647343,
527
- "pages": 45,
528
- "shards": 2
529
  },
530
  {
531
- "seconds": 1.1072141639997426,
532
- "bytes": 6205762,
533
- "pages": 37,
534
- "shards": 2
535
  },
536
  {
537
- "seconds": 0.9200315849993785,
538
- "bytes": 2298271,
539
- "pages": 29,
540
- "shards": 2
541
  },
542
  {
543
- "seconds": 1.584766658001172,
544
- "bytes": 10894182,
545
- "pages": 37,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.1065202139980101,
550
- "bytes": 8432247,
551
- "pages": 39,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.2434194350025791,
556
- "bytes": 6179718,
557
- "pages": 28,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 1.03145402500013,
562
- "bytes": 6170934,
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": 76980.58360649772,
26
+ "required_pages_per_second_upper_estimate": 0.005325695393460453
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": 9622.572950812215,
40
+ "required_pages_per_second_upper_estimate": 0.004028770352331965
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": 19245.14590162443,
54
+ "required_pages_per_second_upper_estimate": 0.004008256362681121
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": 28867.718852436643,
68
+ "required_pages_per_second_upper_estimate": 0.0017305756667491143
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": 28867.718852436643,
82
+ "required_pages_per_second_upper_estimate": 0.00224433816510092
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": 28867.718852436643,
96
+ "required_pages_per_second_upper_estimate": 0.3702841016974724
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": 0.4470424027734907,
162
+ "producer_p95_seconds": 0.30381212399879587,
163
+ "gpu_update_seconds": 0.9577480001564553,
164
+ "producer_batches_per_second": 2.23692426891926,
165
+ "gpu_batches_per_second": 1.0441159885863949,
166
+ "producer_headroom": 2.142409745058862,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.338124009856685,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 77,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 399755.21465372964,
174
+ "upload_bytes_per_second": 5963845.868077777,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 85.77816323956411,
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": 1.0381844509975053,
184
+ "bytes": 7741032,
185
+ "pages": 38,
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
+ "seconds": 1.5200233909999952,
190
+ "bytes": 6160062,
191
+ "pages": 28,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.1740434719977202,
196
+ "bytes": 8468925,
197
+ "pages": 41,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.0248841799984802,
202
+ "bytes": 5963815,
203
+ "pages": 27,
204
+ "shards": 2
205
  },
206
  {
207
+ "seconds": 1.231629100999271,
208
+ "bytes": 4969209,
209
+ "pages": 37,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.249853739001992,
214
+ "bytes": 7694709,
215
+ "pages": 39,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.2034739760019875,
220
+ "bytes": 5697150,
221
  "pages": 28,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 0.9529857010020351,
226
+ "bytes": 6624046,
227
+ "pages": 40,
228
+ "shards": 3
229
  },
230
  {
231
+ "seconds": 1.1070565920017543,
232
+ "bytes": 11831537,
233
+ "pages": 32,
234
+ "shards": 3
235
  },
236
  {
237
+ "seconds": 1.0998464429976593,
238
+ "bytes": 6314908,
239
  "pages": 36,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.0093370079994202,
244
+ "bytes": 2319025,
245
+ "pages": 29,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.116415386000881,
250
+ "bytes": 10937725,
251
+ "pages": 41,
 
 
 
 
 
 
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 0.8731399669995881,
256
+ "bytes": 778500,
257
+ "pages": 25,
258
  "shards": 1
259
  },
260
  {
261
+ "seconds": 1.033372367000993,
262
+ "bytes": 7683491,
263
+ "pages": 38,
264
+ "shards": 3
265
  },
266
  {
267
+ "seconds": 1.4393485709988454,
268
+ "bytes": 11721944,
269
  "pages": 36,
270
+ "shards": 3
271
  },
272
  {
273
+ "seconds": 1.162564084999758,
274
+ "bytes": 5892604,
275
+ "pages": 23,
276
+ "shards": 3
277
  },
278
  {
279
+ "seconds": 1.082973058000789,
280
+ "bytes": 6647343,
281
+ "pages": 45,
282
+ "shards": 2
283
  },
284
  {
285
+ "seconds": 1.1072141639997426,
286
+ "bytes": 6205762,
287
  "pages": 37,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 0.9200315849993785,
292
+ "bytes": 2298271,
293
+ "pages": 29,
294
+ "shards": 2
295
+ },
296
+ {
297
+ "seconds": 1.584766658001172,
298
+ "bytes": 10894182,
299
+ "pages": 37,
300
  "shards": 3
301
  },
302
  {
303
+ "seconds": 1.1065202139980101,
304
+ "bytes": 8432247,
305
  "pages": 39,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.2434194350025791,
310
+ "bytes": 6179718,
311
+ "pages": 28,
312
+ "shards": 2
313
  },
314
  {
315
+ "seconds": 1.03145402500013,
316
+ "bytes": 6170934,
317
+ "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 0.9970344739995198,
322
+ "bytes": 6228444,
323
+ "pages": 33,
324
+ "shards": 3
325
+ },
326
+ {
327
+ "seconds": 0.8807951420021709,
328
+ "bytes": 343455,
329
+ "pages": 9,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 1.1079528180016496,
334
+ "bytes": 5894448,
335
+ "pages": 27,
336
+ "shards": 3
337
  },
338
  {
339
+ "seconds": 1.0535939100009273,
340
+ "bytes": 7640568,
341
+ "pages": 40,
342
+ "shards": 3
343
+ },
344
+ {
345
+ "seconds": 0.9312373259999731,
346
+ "bytes": 6794347,
347
+ "pages": 32,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.2384252730007574,
352
+ "bytes": 10759915,
353
  "pages": 34,
354
+ "shards": 3
355
+ },
356
+ {
357
+ "seconds": 0.8831225600006292,
358
+ "bytes": 2598975,
359
+ "pages": 38,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 1.0244675720023224,
364
+ "bytes": 5791149,
365
  "pages": 27,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 1.2756652919997578,
370
+ "bytes": 7797831,
371
  "pages": 39,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 0.9431873399989854,
376
+ "bytes": 5431625,
377
+ "pages": 32,
378
+ "shards": 2
379
+ },
380
+ {
381
+ "seconds": 1.3104591960000107,
382
+ "bytes": 11823249,
383
+ "pages": 35,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.1533994220008026,
388
+ "bytes": 7669461,
389
+ "pages": 39,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 0.8288239209978201,
394
+ "bytes": 961455,
395
+ "pages": 27,
396
+ "shards": 1
397
  },
398
  {
399
+ "seconds": 1.3137797929994122,
400
+ "bytes": 10288628,
401
+ "pages": 39,
402
  "shards": 3
403
  },
404
  {
405
+ "seconds": 1.1022804700005508,
406
+ "bytes": 2387118,
407
+ "pages": 29,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.1524969450001663,
412
+ "bytes": 6155701,
413
+ "pages": 35,
414
+ "shards": 2
 
 
 
 
 
 
415
  },
416
  {
417
+ "seconds": 1.4659404199992423,
418
+ "bytes": 13390262,
419
  "pages": 39,
420
  "shards": 4
421
  },
422
  {
423
+ "seconds": 0.8803229269979056,
424
+ "bytes": 4831371,
425
+ "pages": 29,
 
 
 
 
 
 
426
  "shards": 2
427
  },
428
  {
429
+ "seconds": 1.1803598539991071,
430
+ "bytes": 6473769,
431
  "pages": 37,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.0211126419999346,
436
+ "bytes": 2557759,
437
  "pages": 38,
438
+ "shards": 2
439
  },
440
  {
441
+ "seconds": 1.475009352001507,
442
+ "bytes": 6011684,
443
  "pages": 28,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.40803722399869,
448
+ "bytes": 10317954,
449
+ "pages": 35,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.058006204002595,
454
+ "bytes": 8327609,
455
+ "pages": 35,
456
+ "shards": 3
457
  },
458
  {
459
+ "seconds": 1.1640828560011869,
460
+ "bytes": 6135465,
461
  "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 0.8637312579994614,
466
+ "bytes": 2324435,
 
 
 
 
 
 
467
  "pages": 28,
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 1.1669898459986143,
472
+ "bytes": 10131671,
473
+ "pages": 41,
 
 
 
 
 
 
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 1.0318543029970897,
478
+ "bytes": 5757385,
479
+ "pages": 28,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 0.9449253849998058,
484
+ "bytes": 2649946,
485
+ "pages": 37,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.2970311710014357,
490
+ "bytes": 11562430,
491
+ "pages": 33,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 1.4141549599989958,
496
+ "bytes": 12336426,
497
+ "pages": 42,
498
+ "shards": 4
499
  },
500
  {
501
+ "seconds": 0.8797184569993988,
502
+ "bytes": 870056,
503
+ "pages": 27,
504
+ "shards": 1
505
  },
506
  {
507
+ "seconds": 1.3180106779982452,
508
+ "bytes": 6077154,
509
+ "pages": 37,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 0.8500619599981292,
514
+ "bytes": 2427201,
515
+ "pages": 31,
516
+ "shards": 2
517
  },
518
  {
519
+ "seconds": 1.7313189029991918,
520
+ "bytes": 10259217,
521
+ "pages": 40,
522
+ "shards": 4
523
  },
524
  {
525
+ "seconds": 1.5030511410004692,
526
+ "bytes": 13228107,
527
+ "pages": 34,
528
+ "shards": 4
529
  },
530
  {
531
+ "seconds": 0.8756032579985913,
532
+ "bytes": 1232264,
533
+ "pages": 34,
534
+ "shards": 1
535
  },
536
  {
537
+ "seconds": 1.2522269889996096,
538
+ "bytes": 10696457,
539
+ "pages": 36,
540
  "shards": 3
541
  },
542
  {
543
+ "seconds": 1.1242637430004834,
544
+ "bytes": 7345030,
545
+ "pages": 30,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 0.8663610330004303,
550
+ "bytes": 1077995,
551
+ "pages": 35,
552
+ "shards": 1
553
  },
554
  {
555
+ "seconds": 0.9089211409991549,
556
+ "bytes": 5857565,
557
+ "pages": 34,
558
  "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.1926017650002905,
562
+ "bytes": 13608076,
563
+ "pages": 35,
564
+ "shards": 4
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9af8f113fcd330718ea1f4faa81d21b284817f06e6050316baaf832fa2400fa0
3
- size 721762431
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b280b0d64c4eb0d14a1021d0c9cdbedc4c0bf8eb245a63c69257180b96af6158
3
+ size 721780415