appvoid commited on
Commit
b59a6a0
·
verified ·
1 Parent(s): 768f18a

Qwen90M step 20000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +47 -47
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +251 -251
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 15000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "cd4b4a336ca854cc63b56daafd448bbdabc975be153afb5b40dbcfeed84c0a5c",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "9ee3cdbc8544ee947af4a7d860b9cd7d6a5386d7eb87fa5e677065b54efe108a",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
- "shard_scheduler.json": "4d92f7ac5793b7b10fc6dcd78ea5d5fc6a289ca95e0d0b954b4effa3df49ceff",
15
- "coverage.json": "ae6b95743a2e517acaa8daca9ca6ec0408a66feeeb897332ac3aa2b7585e4311",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 20000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "8ac28a22873400a2ad81f120bca87505f54dd1c44997b9d1fb50fb7b5e06c38a",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "f18e92b7be914c509e752f35a75c1d1da97d9597baf2c06a4ed1fd4a3d17d365",
12
  "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2",
13
  "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e",
14
+ "shard_scheduler.json": "ba70b7966dd2918d6c10d0e338ae611c7dc6c0c15e14439a5011a3b293e76f49",
15
+ "coverage.json": "4e370713d47399509a8d42fcc8692de21ea4cb81a32e0196eac9278823ef761f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 15000,
3
  "stats": {
4
- "climbmix/blocks": 540000,
5
- "climbmix/targets": 901514897,
6
- "climbmix/input_tokens": 901572469,
7
- "ultra_qa/blocks": 202500,
8
- "ultra_qa/targets": 330561975,
9
- "ultra_qa/input_tokens": 330562626,
10
- "cortex/blocks": 202500,
11
- "cortex/targets": 7710129,
12
- "cortex/input_tokens": 321940622,
13
- "ultra_style/blocks": 202500,
14
- "ultra_style/targets": 320384397,
15
- "ultra_style/input_tokens": 320401468,
16
- "openmath/blocks": 135000,
17
- "openmath/targets": 208599081,
18
- "openmath/input_tokens": 208668782,
19
- "rewrite6/blocks": 67500,
20
- "rewrite6/targets": 110633666,
21
- "rewrite6/input_tokens": 110755126
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 74,
26
- "row": 1396,
27
  "start": {
28
- "file": 3,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 44,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 304500
38
  },
39
  "rewrite6": {
40
- "page": 42,
41
- "row": 1009,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 172032
47
  }
48
  },
49
- "epoch": 0,
50
- "offset": 0,
51
- "documents": 173041
52
  },
53
  "openmath": {
54
- "page": 46,
55
- "row": 1133,
56
  "start": {
57
  "file": 0,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 188,
61
- "row": 416
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 189549
67
  },
68
  "ultra_qa": {
69
- "page": 20,
70
- "row": 1,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 1,
76
- "row": 17350
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 81921
82
  },
83
  "ultra_style": {
84
- "page": 30,
85
- "row": 2561,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 1,
91
- "row": 37513
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 125441
97
  },
98
  "cortex": {
99
- "page": 3926,
100
- "row": 10,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 502538
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 20000,
3
  "stats": {
4
+ "climbmix/blocks": 720000,
5
+ "climbmix/targets": 1202690209,
6
+ "climbmix/input_tokens": 1202766834,
7
+ "ultra_qa/blocks": 270000,
8
+ "ultra_qa/targets": 440598577,
9
+ "ultra_qa/input_tokens": 440599431,
10
+ "cortex/blocks": 270000,
11
+ "cortex/targets": 10270926,
12
+ "cortex/input_tokens": 429238220,
13
+ "ultra_style/blocks": 270000,
14
+ "ultra_style/targets": 426974586,
15
+ "ultra_style/input_tokens": 426997226,
16
+ "openmath/blocks": 180000,
17
+ "openmath/targets": 278145153,
18
+ "openmath/input_tokens": 278238017,
19
+ "rewrite6/blocks": 90000,
20
+ "rewrite6/targets": 147537674,
21
+ "rewrite6/input_tokens": 147699508
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 98,
26
+ "row": 3561,
27
  "start": {
28
+ "file": 4,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 58,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 404969
38
  },
39
  "rewrite6": {
40
+ "page": 6,
41
+ "row": 493,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 24576
47
  }
48
  },
49
+ "epoch": 1,
50
+ "offset": 2048,
51
+ "documents": 230473
52
  },
53
  "openmath": {
54
+ "page": 61,
55
+ "row": 2771,
56
  "start": {
57
  "file": 0,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 249,
61
+ "row": 856
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 252627
67
  },
68
  "ultra_qa": {
69
+ "page": 26,
70
+ "row": 2708,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 1,
76
+ "row": 41926
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 109204
82
  },
83
  "ultra_style": {
84
+ "page": 40,
85
+ "row": 3229,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 1,
91
+ "row": 78473
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 167069
97
  },
98
  "cortex": {
99
+ "page": 5235,
100
+ "row": 127,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 670207
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 15000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "cd4b4a336ca854cc63b56daafd448bbdabc975be153afb5b40dbcfeed84c0a5c", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9ee3cdbc8544ee947af4a7d860b9cd7d6a5386d7eb87fa5e677065b54efe108a", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "4d92f7ac5793b7b10fc6dcd78ea5d5fc6a289ca95e0d0b954b4effa3df49ceff", "coverage.json": "ae6b95743a2e517acaa8daca9ca6ec0408a66feeeb897332ac3aa2b7585e4311", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000015000"}
 
1
+ {"step": 20000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "8ac28a22873400a2ad81f120bca87505f54dd1c44997b9d1fb50fb7b5e06c38a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f18e92b7be914c509e752f35a75c1d1da97d9597baf2c06a4ed1fd4a3d17d365", "training_config.json": "07f331dbc4eba0ea92b45763c59c20af4874302e74a327f3562642d69645fed2", "performance.json": "b95a1befd575f8d9a02b43fd0cc41c54243f5297127b93d2656e0292ba15e77e", "shard_scheduler.json": "ba70b7966dd2918d6c10d0e338ae611c7dc6c0c15e14439a5011a3b293e76f49", "coverage.json": "4e370713d47399509a8d42fcc8692de21ea4cb81a32e0196eac9278823ef761f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000020000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cd4b4a336ca854cc63b56daafd448bbdabc975be153afb5b40dbcfeed84c0a5c
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ac28a22873400a2ad81f120bca87505f54dd1c44997b9d1fb50fb7b5e06c38a
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75908.27937948794,
26
- "required_pages_per_second_upper_estimate": 0.006579537268838822
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9488.534922435992,
40
- "required_pages_per_second_upper_estimate": 0.0037851184467991034
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18977.069844871985,
54
- "required_pages_per_second_upper_estimate": 0.004390011123588869
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28465.604767307974,
68
- "required_pages_per_second_upper_estimate": 0.0017334059828430235
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28465.604767307974,
82
- "required_pages_per_second_upper_estimate": 0.0021913610528246498
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28465.604767307974,
96
- "required_pages_per_second_upper_estimate": 0.3534475429592358
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.6275874297735697,
162
- "producer_p95_seconds": 0.29346390000137035,
163
- "gpu_update_seconds": 0.9712774496101002,
164
- "producer_batches_per_second": 1.5934034885956763,
165
- "gpu_batches_per_second": 1.029571931687933,
166
- "producer_headroom": 1.5476368766030448,
167
  "sustainable": true,
168
- "buffer_seconds": 60.219201875826215,
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": 234605.4285099929,
174
- "upload_bytes_per_second": 6302700.702389904,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 50.340863623379136,
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": 1.0970937430001868,
184
- "bytes": 8464356,
185
- "pages": 36,
186
- "shards": 4
187
  },
188
  {
189
- "seconds": 1.18106462400101,
190
- "bytes": 6740941,
191
- "pages": 29,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.0259071060008864,
196
- "bytes": 6033717,
197
  "pages": 35,
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.3882136779993743,
202
- "bytes": 13040894,
203
- "pages": 40,
204
- "shards": 5
205
- },
206
- {
207
- "seconds": 0.7154928489999293,
208
- "bytes": 835967,
209
- "pages": 25,
210
- "shards": 1
211
  },
212
  {
213
- "seconds": 1.0692436869994708,
214
- "bytes": 7577178,
215
- "pages": 36,
216
  "shards": 3
217
  },
218
  {
219
- "seconds": 1.0859860309992655,
220
- "bytes": 2128705,
221
- "pages": 34,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.1122738659996685,
226
- "bytes": 11663489,
227
- "pages": 30,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 1.1172626269999455,
232
- "bytes": 11245903,
233
  "pages": 36,
234
  "shards": 4
235
  },
236
  {
237
- "seconds": 0.9253301299995655,
238
- "bytes": 2328336,
239
- "pages": 35,
240
- "shards": 2
241
- },
242
- {
243
- "seconds": 1.028129626000009,
244
- "bytes": 6087526,
245
- "pages": 35,
246
- "shards": 2
247
  },
248
  {
249
- "seconds": 0.7597279170004185,
250
- "bytes": 2156570,
251
  "pages": 21,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 0.9871070610006427,
256
- "bytes": 12008396,
257
- "pages": 45,
258
  "shards": 4
259
  },
260
  {
261
- "seconds": 1.228597503999481,
262
- "bytes": 13077636,
263
  "pages": 34,
264
- "shards": 4
265
- },
266
- {
267
- "seconds": 1.1013024290004978,
268
- "bytes": 1132593,
269
- "pages": 33,
270
- "shards": 1
271
- },
272
- {
273
- "seconds": 0.9999866519992793,
274
- "bytes": 6918002,
275
- "pages": 32,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.2065209279990086,
280
- "bytes": 11423370,
281
- "pages": 39,
282
- "shards": 4
283
  },
284
  {
285
- "seconds": 1.0073676210013218,
286
- "bytes": 932900,
287
- "pages": 27,
288
- "shards": 1
289
  },
290
  {
291
- "seconds": 1.0279631710000103,
292
- "bytes": 7068347,
293
  "pages": 35,
294
- "shards": 3
295
  },
296
  {
297
- "seconds": 0.8251235360003193,
298
- "bytes": 2046300,
299
- "pages": 19,
300
- "shards": 2
301
  },
302
  {
303
- "seconds": 1.6571312910000415,
304
- "bytes": 16761254,
305
  "pages": 38,
306
- "shards": 4
307
  },
308
  {
309
- "seconds": 0.7977973630004271,
310
- "bytes": 2044807,
311
- "pages": 28,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 1.0466349409998656,
316
- "bytes": 7179273,
317
- "pages": 20,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 0.7437304710001627,
322
- "bytes": 927772,
323
- "pages": 26,
324
- "shards": 1
325
  },
326
  {
327
- "seconds": 1.1199327470003482,
328
- "bytes": 6072182,
329
- "pages": 30,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 0.9855451060011546,
334
- "bytes": 8735177,
335
- "pages": 38,
336
  "shards": 4
337
  },
338
  {
339
- "seconds": 0.964613582000311,
340
- "bytes": 5800981,
341
- "pages": 27,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 0.8288159710009495,
346
- "bytes": 7059901,
347
- "pages": 37,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 0.9880495930010511,
352
- "bytes": 8851756,
353
- "pages": 43,
354
- "shards": 4
355
- },
356
- {
357
- "seconds": 1.0367009409983439,
358
- "bytes": 10231090,
359
- "pages": 31,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 0.8111973979994218,
364
- "bytes": 875148,
365
- "pages": 25,
366
- "shards": 1
367
- },
368
- {
369
- "seconds": 1.2982671780009696,
370
- "bytes": 8839899,
371
- "pages": 39,
372
  "shards": 4
373
  },
374
  {
375
- "seconds": 0.756701359001454,
376
- "bytes": 1071685,
377
- "pages": 31,
378
  "shards": 1
379
  },
380
  {
381
- "seconds": 1.2869488490014191,
382
- "bytes": 16651954,
383
- "pages": 29,
384
- "shards": 5
385
  },
386
  {
387
- "seconds": 0.9651243050011544,
388
- "bytes": 7745172,
389
- "pages": 44,
390
- "shards": 4
391
  },
392
  {
393
- "seconds": 0.7505128630000399,
394
- "bytes": 1172556,
395
- "pages": 33,
396
- "shards": 1
397
  },
398
  {
399
- "seconds": 1.1141684859994712,
400
- "bytes": 8597889,
401
- "pages": 33,
402
- "shards": 4
403
  },
404
  {
405
- "seconds": 1.1277135729997099,
406
- "bytes": 5536865,
407
- "pages": 31,
408
  "shards": 2
409
  },
410
  {
411
- "seconds": 0.9057988900003693,
412
- "bytes": 5888539,
413
- "pages": 27,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.030597670998759,
418
- "bytes": 7980071,
419
- "pages": 26,
420
- "shards": 3
421
  },
422
  {
423
- "seconds": 0.9780282619994978,
424
- "bytes": 7468851,
425
- "pages": 44,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.20604977199946,
430
- "bytes": 11385510,
431
- "pages": 35,
432
  "shards": 4
433
  },
434
  {
435
- "seconds": 0.7792264220006473,
436
- "bytes": 714539,
437
- "pages": 19,
438
- "shards": 1
439
  },
440
  {
441
- "seconds": 1.1150978799996665,
442
- "bytes": 7568611,
443
- "pages": 45,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.1983624430013151,
448
- "bytes": 7531256,
449
- "pages": 35,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 1.255419142999017,
454
- "bytes": 11119971,
455
- "pages": 21,
456
- "shards": 3
457
  },
458
  {
459
- "seconds": 1.229843424000137,
460
- "bytes": 7832261,
461
- "pages": 45,
462
- "shards": 3
463
  },
464
  {
465
- "seconds": 0.9293085339995741,
466
- "bytes": 2351431,
467
- "pages": 32,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.204088584998317,
472
- "bytes": 7306502,
473
- "pages": 36,
474
- "shards": 3
475
  },
476
  {
477
- "seconds": 1.0363285590010491,
478
- "bytes": 11683428,
479
- "pages": 36,
480
- "shards": 4
481
  },
482
  {
483
- "seconds": 1.0476808559997153,
484
- "bytes": 995690,
485
- "pages": 29,
486
- "shards": 1
487
  },
488
  {
489
- "seconds": 1.5069925719999446,
490
- "bytes": 1775692,
491
- "pages": 21,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.199250067000321,
496
- "bytes": 13436756,
497
  "pages": 47,
498
  "shards": 4
499
  },
500
  {
501
- "seconds": 1.2015854260007472,
502
- "bytes": 10309364,
503
- "pages": 34,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 0.9169029310014594,
508
- "bytes": 7362656,
509
- "pages": 34,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 0.9597948790014925,
514
- "bytes": 2508350,
515
- "pages": 35,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.1389318259989523,
520
- "bytes": 6045358,
521
  "pages": 35,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.1304748169986851,
526
- "bytes": 12071070,
527
- "pages": 27,
 
 
 
 
 
 
 
 
 
 
 
 
528
  "shards": 4
529
  },
530
  {
531
- "seconds": 3.228665867998643,
532
- "bytes": 7676896,
533
- "pages": 38,
534
- "shards": 3
535
  },
536
  {
537
- "seconds": 0.9868028299988509,
538
- "bytes": 6101148,
539
- "pages": 35,
540
  "shards": 2
541
  },
542
  {
543
- "seconds": 0.8392134009991423,
544
- "bytes": 3718175,
545
- "pages": 34,
546
- "shards": 3
547
  },
548
  {
549
- "seconds": 1.0338640699992538,
550
- "bytes": 9654277,
551
- "pages": 35,
552
- "shards": 3
553
  },
554
  {
555
- "seconds": 0.8135462400005053,
556
- "bytes": 1110572,
557
  "pages": 33,
558
  "shards": 1
559
  },
560
  {
561
- "seconds": 1.3182481060011924,
562
- "bytes": 8487979,
563
- "pages": 27,
 
 
 
 
 
 
564
  "shards": 4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75780.15754960763,
26
+ "required_pages_per_second_upper_estimate": 0.006568431993346605
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9472.519693700953,
40
+ "required_pages_per_second_upper_estimate": 0.0037787297326076883
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18945.039387401906,
54
+ "required_pages_per_second_upper_estimate": 0.004382601440969964
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28417.55908110286,
68
+ "required_pages_per_second_upper_estimate": 0.001730480252629366
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28417.55908110286,
82
+ "required_pages_per_second_upper_estimate": 0.0021876623629016084
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28417.55908110286,
96
+ "required_pages_per_second_upper_estimate": 0.3528509763351361
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.09428515504697543,
162
+ "producer_p95_seconds": 0.2934562519985775,
163
+ "gpu_update_seconds": 0.9729195924637102,
164
+ "producer_batches_per_second": 10.60612351437268,
165
+ "gpu_batches_per_second": 1.0278341681533152,
166
+ "producer_headroom": 10.318905367223241,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.32101473275003,
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": 321461.5944819022,
174
+ "upload_bytes_per_second": 5587316.407845459,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 68.97817493288802,
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": 0.7792264220006473,
184
+ "bytes": 714539,
185
+ "pages": 19,
186
+ "shards": 1
187
  },
188
  {
189
+ "seconds": 1.1150978799996665,
190
+ "bytes": 7568611,
191
+ "pages": 45,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.1983624430013151,
196
+ "bytes": 7531256,
197
  "pages": 35,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.255419142999017,
202
+ "bytes": 11119971,
203
+ "pages": 21,
204
+ "shards": 3
 
 
 
 
 
 
205
  },
206
  {
207
+ "seconds": 1.229843424000137,
208
+ "bytes": 7832261,
209
+ "pages": 45,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 0.9293085339995741,
214
+ "bytes": 2351431,
215
+ "pages": 32,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.204088584998317,
220
+ "bytes": 7306502,
221
+ "pages": 36,
222
  "shards": 3
223
  },
224
  {
225
+ "seconds": 1.0363285590010491,
226
+ "bytes": 11683428,
227
  "pages": 36,
228
  "shards": 4
229
  },
230
  {
231
+ "seconds": 1.0476808559997153,
232
+ "bytes": 995690,
233
+ "pages": 29,
234
+ "shards": 1
 
 
 
 
 
 
235
  },
236
  {
237
+ "seconds": 1.5069925719999446,
238
+ "bytes": 1775692,
239
  "pages": 21,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.199250067000321,
244
+ "bytes": 13436756,
245
+ "pages": 47,
246
  "shards": 4
247
  },
248
  {
249
+ "seconds": 1.2015854260007472,
250
+ "bytes": 10309364,
251
  "pages": 34,
 
 
 
 
 
 
 
 
 
 
 
 
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 0.9169029310014594,
256
+ "bytes": 7362656,
257
+ "pages": 34,
258
+ "shards": 3
259
  },
260
  {
261
+ "seconds": 0.9597948790014925,
262
+ "bytes": 2508350,
263
+ "pages": 35,
264
+ "shards": 2
265
  },
266
  {
267
+ "seconds": 1.1389318259989523,
268
+ "bytes": 6045358,
269
  "pages": 35,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 1.1304748169986851,
274
+ "bytes": 12071070,
275
+ "pages": 27,
276
+ "shards": 4
277
  },
278
  {
279
+ "seconds": 3.228665867998643,
280
+ "bytes": 7676896,
281
  "pages": 38,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 0.9868028299988509,
286
+ "bytes": 6101148,
287
+ "pages": 35,
288
  "shards": 2
289
  },
290
  {
291
+ "seconds": 0.8392134009991423,
292
+ "bytes": 3718175,
293
+ "pages": 34,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.0338640699992538,
298
+ "bytes": 9654277,
299
+ "pages": 35,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 0.8135462400005053,
304
+ "bytes": 1110572,
305
+ "pages": 33,
306
+ "shards": 1
307
  },
308
  {
309
+ "seconds": 1.3182481060011924,
310
+ "bytes": 8487979,
311
+ "pages": 27,
312
  "shards": 4
313
  },
314
  {
315
+ "seconds": 1.0571894279983098,
316
+ "bytes": 6049663,
317
+ "pages": 10,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.061286989999644,
322
+ "bytes": 6189988,
323
+ "pages": 35,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.1359487030003947,
328
+ "bytes": 6670820,
329
+ "pages": 29,
 
 
 
 
 
 
330
  "shards": 3
331
  },
332
  {
333
+ "seconds": 0.9071375989988155,
334
+ "bytes": 7422614,
335
+ "pages": 37,
 
 
 
 
 
 
336
  "shards": 4
337
  },
338
  {
339
+ "seconds": 0.8035606240009656,
340
+ "bytes": 1171760,
341
+ "pages": 34,
342
  "shards": 1
343
  },
344
  {
345
+ "seconds": 1.2951343000004272,
346
+ "bytes": 7367275,
347
+ "pages": 27,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.1229757969995262,
352
+ "bytes": 5828242,
353
+ "pages": 34,
354
+ "shards": 3
355
  },
356
  {
357
+ "seconds": 1.539654905001953,
358
+ "bytes": 12116214,
359
+ "pages": 37,
360
+ "shards": 3
361
  },
362
  {
363
+ "seconds": 1.2743941750013619,
364
+ "bytes": 7473487,
365
+ "pages": 29,
366
+ "shards": 3
367
  },
368
  {
369
+ "seconds": 0.8017673559988907,
370
+ "bytes": 2201439,
371
+ "pages": 35,
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.2722073730001284,
376
+ "bytes": 12361137,
377
+ "pages": 39,
378
+ "shards": 4
379
  },
380
  {
381
+ "seconds": 0.8463285300022108,
382
+ "bytes": 752773,
383
+ "pages": 23,
384
+ "shards": 1
385
  },
386
  {
387
+ "seconds": 1.0778829359987867,
388
+ "bytes": 7678507,
389
+ "pages": 43,
390
  "shards": 3
391
  },
392
  {
393
+ "seconds": 1.0977264680004737,
394
+ "bytes": 13270237,
395
+ "pages": 39,
396
  "shards": 4
397
  },
398
  {
399
+ "seconds": 0.8527823730000819,
400
+ "bytes": 4778684,
401
+ "pages": 20,
402
+ "shards": 2
403
  },
404
  {
405
+ "seconds": 0.9591597530015861,
406
+ "bytes": 7365649,
407
+ "pages": 37,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.119754101997387,
412
+ "bytes": 2574245,
413
+ "pages": 36,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.0843618420003622,
418
+ "bytes": 5928833,
419
+ "pages": 24,
420
+ "shards": 2
421
  },
422
  {
423
+ "seconds": 1.330314847000409,
424
+ "bytes": 13379246,
425
+ "pages": 41,
426
+ "shards": 5
427
  },
428
  {
429
+ "seconds": 1.0224713220013655,
430
+ "bytes": 6909652,
431
+ "pages": 35,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 1.018912650000857,
436
+ "bytes": 5819747,
437
+ "pages": 22,
438
+ "shards": 2
439
  },
440
  {
441
+ "seconds": 0.8826371689974621,
442
+ "bytes": 4172226,
443
+ "pages": 46,
444
+ "shards": 3
445
  },
446
  {
447
+ "seconds": 2.508591282999987,
448
+ "bytes": 5665007,
449
+ "pages": 21,
450
+ "shards": 2
451
  },
452
  {
453
+ "seconds": 1.1267430049992981,
454
+ "bytes": 4735418,
455
+ "pages": 32,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.0588101130015275,
460
+ "bytes": 8209294,
461
  "pages": 47,
462
  "shards": 4
463
  },
464
  {
465
+ "seconds": 1.0802435129990045,
466
+ "bytes": 11583568,
467
+ "pages": 29,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.2433872929977952,
472
+ "bytes": 7696542,
473
+ "pages": 38,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 0.9942282960000739,
478
+ "bytes": 5730133,
479
+ "pages": 28,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 0.9958830260002287,
484
+ "bytes": 6097937,
485
  "pages": 35,
486
  "shards": 2
487
  },
488
  {
489
+ "seconds": 1.0204584569983126,
490
+ "bytes": 2643615,
491
+ "pages": 36,
492
+ "shards": 2
493
+ },
494
+ {
495
+ "seconds": 1.1489640459985822,
496
+ "bytes": 6010015,
497
+ "pages": 32,
498
+ "shards": 2
499
+ },
500
+ {
501
+ "seconds": 1.3807983080005215,
502
+ "bytes": 15814135,
503
+ "pages": 39,
504
  "shards": 4
505
  },
506
  {
507
+ "seconds": 1.1238995200001227,
508
+ "bytes": 2339467,
509
+ "pages": 27,
510
+ "shards": 2
511
  },
512
  {
513
+ "seconds": 1.365642591001233,
514
+ "bytes": 6320146,
515
+ "pages": 36,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0016697349965398,
520
+ "bytes": 761758,
521
+ "pages": 22,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.1483951949994662,
526
+ "bytes": 11735955,
527
+ "pages": 47,
528
+ "shards": 4
529
  },
530
  {
531
+ "seconds": 1.0657177410030272,
532
+ "bytes": 1129396,
533
  "pages": 33,
534
  "shards": 1
535
  },
536
  {
537
+ "seconds": 1.0460155590008071,
538
+ "bytes": 6970566,
539
+ "pages": 24,
540
+ "shards": 3
541
+ },
542
+ {
543
+ "seconds": 2.029252282998641,
544
+ "bytes": 16530913,
545
+ "pages": 39,
546
  "shards": 4
547
+ },
548
+ {
549
+ "seconds": 0.8704905230006261,
550
+ "bytes": 1186552,
551
+ "pages": 33,
552
+ "shards": 1
553
+ },
554
+ {
555
+ "seconds": 3.2236978209984954,
556
+ "bytes": 7889017,
557
+ "pages": 38,
558
+ "shards": 3
559
+ },
560
+ {
561
+ "seconds": 0.9988030870008515,
562
+ "bytes": 1104863,
563
+ "pages": 33,
564
+ "shards": 1
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9ee3cdbc8544ee947af4a7d860b9cd7d6a5386d7eb87fa5e677065b54efe108a
3
- size 721787967
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f18e92b7be914c509e752f35a75c1d1da97d9597baf2c06a4ed1fd4a3d17d365
3
+ size 721771775