appvoid commited on
Commit
861d799
·
verified ·
1 Parent(s): 552a78c

Qwen90M step 55000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +51 -51
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +248 -224
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 50000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "681b4da7833f165cc654739e94f0e7a807bd7a54f8adefbfe0a179fa5fd48b20",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "ad5414aa9cc5e09a287ca937f00d1a790494a4a5870ec4a0ca71e8a2bf9bfdc0",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f",
14
- "shard_scheduler.json": "7b240c248709a7609872c5ec8f72f1fa0583a80ceadb696ffbc883b68d067333",
15
- "coverage.json": "5e6ec826d3400c038f929348e9886127bd7b7fc2f4b301d6ad5272d1b42e325c",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 55000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f",
14
+ "shard_scheduler.json": "0b7760fce951fcc476bd9ce864c73c4a8111a03bf92e84c8373e26e818a1f854",
15
+ "coverage.json": "a90e5fa0fd4f8822ac17717d8ddbef2645fe20188ab34ad850e1754843896bc9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 50000,
3
  "stats": {
4
- "climbmix/blocks": 1800000,
5
- "climbmix/targets": 3007990030,
6
- "climbmix/input_tokens": 3008180540,
7
- "ultra_qa/blocks": 675000,
8
- "ultra_qa/targets": 1101475594,
9
- "ultra_qa/input_tokens": 1101477776,
10
- "cortex/blocks": 675000,
11
- "cortex/targets": 25668963,
12
- "cortex/input_tokens": 1073030847,
13
- "ultra_style/blocks": 675000,
14
- "ultra_style/targets": 1067349516,
15
- "ultra_style/input_tokens": 1067405409,
16
- "openmath/blocks": 450000,
17
- "openmath/targets": 695410628,
18
- "openmath/input_tokens": 695642540,
19
- "rewrite6/blocks": 225000,
20
- "rewrite6/targets": 368835030,
21
- "rewrite6/input_tokens": 369239542
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 246,
26
- "row": 1907,
27
  "start": {
28
- "file": 11,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 73,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 108544,
37
- "documents": 1009523
38
  },
39
  "rewrite6": {
40
- "page": 40,
41
- "row": 1649,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 163840
47
  }
48
  },
49
- "epoch": 2,
50
- "offset": 0,
51
- "documents": 576297
52
  },
53
  "openmath": {
54
- "page": 154,
55
- "row": 365,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 194,
61
- "row": 134
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 631149
67
  },
68
  "ultra_qa": {
69
- "page": 66,
70
- "row": 2711,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 12002
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 273047
82
  },
83
  "ultra_style": {
84
- "page": 101,
85
- "row": 3047,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 4,
91
- "row": 73675
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 4096,
96
- "documents": 416743
97
  },
98
  "cortex": {
99
- "page": 13088,
100
- "row": 13,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 1675277
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 55000,
3
  "stats": {
4
+ "climbmix/blocks": 1980000,
5
+ "climbmix/targets": 3309029909,
6
+ "climbmix/input_tokens": 3309239389,
7
+ "ultra_qa/blocks": 742500,
8
+ "ultra_qa/targets": 1211473004,
9
+ "ultra_qa/input_tokens": 1211475401,
10
+ "cortex/blocks": 742500,
11
+ "cortex/targets": 28237480,
12
+ "cortex/input_tokens": 1180342911,
13
+ "ultra_style/blocks": 742500,
14
+ "ultra_style/targets": 1174419713,
15
+ "ultra_style/input_tokens": 1174481204,
16
+ "openmath/blocks": 495000,
17
+ "openmath/targets": 764845062,
18
+ "openmath/input_tokens": 765100091,
19
+ "rewrite6/blocks": 247500,
20
+ "rewrite6/targets": 405729377,
21
+ "rewrite6/input_tokens": 406174162
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 271,
26
+ "row": 208,
27
  "start": {
28
+ "file": 13,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 9,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 53248,
37
+ "documents": 1110224
38
  },
39
  "rewrite6": {
40
+ "page": 4,
41
+ "row": 1053,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 16384
47
  }
48
  },
49
+ "epoch": 3,
50
+ "offset": 2048,
51
+ "documents": 633649
52
  },
53
  "openmath": {
54
+ "page": 169,
55
+ "row": 1982,
56
  "start": {
57
  "file": 1,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 255,
61
+ "row": 574
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 694206
67
  },
68
  "ultra_qa": {
69
+ "page": 73,
70
+ "row": 1336,
71
  "start": {
72
  "file": 0,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 40674
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 300344
82
  },
83
  "ultra_style": {
84
+ "page": 111,
85
+ "row": 3394,
86
  "start": {
87
  "file": 0,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 5,
91
+ "row": 28852
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 458050
97
  },
98
  "cortex": {
99
+ "page": 14397,
100
+ "row": 119,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 1842935
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 50000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "681b4da7833f165cc654739e94f0e7a807bd7a54f8adefbfe0a179fa5fd48b20", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ad5414aa9cc5e09a287ca937f00d1a790494a4a5870ec4a0ca71e8a2bf9bfdc0", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "7b240c248709a7609872c5ec8f72f1fa0583a80ceadb696ffbc883b68d067333", "coverage.json": "5e6ec826d3400c038f929348e9886127bd7b7fc2f4b301d6ad5272d1b42e325c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000050000"}
 
1
+ {"step": 55000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "d14cb208816158eb276e72d732c3b9e6358b2a75509678412180fc272350709f", "shard_scheduler.json": "0b7760fce951fcc476bd9ce864c73c4a8111a03bf92e84c8373e26e818a1f854", "coverage.json": "a90e5fa0fd4f8822ac17717d8ddbef2645fe20188ab34ad850e1754843896bc9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000055000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:681b4da7833f165cc654739e94f0e7a807bd7a54f8adefbfe0a179fa5fd48b20
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f37360b73b215cfbd1cab3608606dbff58530c087eebca9f4fe5d5a1c20784da
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": 76046.63191128224,
26
- "required_pages_per_second_upper_estimate": 0.006591529315643466
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": 9505.82898891028,
40
- "required_pages_per_second_upper_estimate": 0.0037920173084850326
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": 19011.65797782056,
54
- "required_pages_per_second_upper_estimate": 0.004398012479416172
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": 28517.48696673084,
68
- "required_pages_per_second_upper_estimate": 0.0017365653365830128
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": 28517.48696673084,
82
- "required_pages_per_second_upper_estimate": 0.002195355088155342
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": 28517.48696673084,
96
- "required_pages_per_second_upper_estimate": 0.3540917462375162
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,386 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.07752646051569911,
162
- "producer_p95_seconds": 0.21125603599830356,
163
- "gpu_update_seconds": 0.9695103931231667,
164
- "producer_batches_per_second": 12.898821813198861,
165
- "gpu_batches_per_second": 1.0314484579980774,
166
- "producer_headroom": 12.505541806940107,
167
  "sustainable": true,
168
- "buffer_seconds": 60.109644373636336,
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": 376788.25974089396,
174
- "upload_bytes_per_second": 6017077.810174047,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 80.84998935861697,
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": 1.0632401930001834,
184
- "bytes": 22792,
185
- "pages": 1,
186
- "shards": 1
187
  },
188
  {
189
- "seconds": 0.9811290180000469,
190
- "bytes": 1073130,
191
- "pages": 29,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
192
  "shards": 1
193
  },
194
  {
195
- "seconds": 1.3945427510002446,
196
- "bytes": 16977562,
197
- "pages": 43,
198
- "shards": 5
199
  },
200
  {
201
- "seconds": 0.7539118750000853,
202
- "bytes": 789098,
203
- "pages": 25,
204
  "shards": 1
205
  },
206
  {
207
- "seconds": 0.9344837560001906,
208
- "bytes": 6249982,
209
  "pages": 35,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 1.0289287979999244,
214
- "bytes": 7839831,
215
- "pages": 35,
216
- "shards": 3
217
  },
218
  {
219
- "seconds": 1.02437976300007,
220
- "bytes": 6057442,
221
- "pages": 31,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.1024301279999236,
226
- "bytes": 7712415,
227
  "pages": 37,
228
  "shards": 3
229
  },
230
  {
231
- "seconds": 0.9960975080002754,
232
- "bytes": 6877408,
233
- "pages": 35,
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.0195401410001068,
238
- "bytes": 6014632,
239
- "pages": 34,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 1.221793560000151,
244
- "bytes": 10970126,
245
- "pages": 34,
246
  "shards": 4
247
  },
248
  {
249
- "seconds": 0.999679897000533,
250
- "bytes": 1109612,
251
- "pages": 33,
252
- "shards": 1
253
  },
254
  {
255
- "seconds": 0.9920720150003035,
256
- "bytes": 6122911,
257
- "pages": 35,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 1.0328473609997673,
262
- "bytes": 7480746,
263
- "pages": 29,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.141071416000159,
268
- "bytes": 10868715,
269
- "pages": 32,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 1.1746170310007074,
274
- "bytes": 7476573,
275
  "pages": 37,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.0039610859994355,
280
- "bytes": 1092998,
281
- "pages": 34,
282
  "shards": 1
283
  },
284
  {
285
- "seconds": 1.1609742079999705,
286
- "bytes": 10556650,
287
- "pages": 33,
288
  "shards": 3
289
  },
290
  {
291
- "seconds": 1.1113881370001764,
292
- "bytes": 7415466,
293
- "pages": 37,
294
- "shards": 3
295
  },
296
  {
297
- "seconds": 1.0273279060002096,
298
- "bytes": 1056303,
299
- "pages": 33,
300
  "shards": 1
301
  },
302
  {
303
- "seconds": 1.1538401379993957,
304
- "bytes": 11570054,
305
- "pages": 22,
 
 
 
 
 
 
 
 
 
 
 
 
306
  "shards": 3
307
  },
308
  {
309
- "seconds": 0.7532719520004321,
310
- "bytes": 2586371,
311
  "pages": 35,
312
  "shards": 2
313
  },
314
  {
315
- "seconds": 0.9901489570002013,
316
- "bytes": 9709695,
317
- "pages": 35,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 0.8841591720001816,
322
- "bytes": 855761,
323
- "pages": 25,
324
- "shards": 1
325
  },
326
  {
327
- "seconds": 1.0832384449995516,
328
- "bytes": 7778060,
329
- "pages": 38,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 0.9103183699999136,
334
- "bytes": 5940010,
335
  "pages": 35,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 0.9267724249993989,
340
- "bytes": 5041964,
341
- "pages": 22,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.078922188999968,
346
- "bytes": 13377142,
347
- "pages": 44,
348
- "shards": 5
349
  },
350
  {
351
- "seconds": 0.9435872209996887,
352
- "bytes": 6172084,
353
- "pages": 36,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 0.8994586850003543,
358
- "bytes": 2397832,
359
  "pages": 28,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.344766387999698,
364
- "bytes": 10673063,
365
  "pages": 35,
366
  "shards": 3
367
  },
368
  {
369
- "seconds": 2.8867894830000296,
370
- "bytes": 1267136,
371
  "pages": 34,
372
  "shards": 1
373
  },
374
  {
375
- "seconds": 1.532430633000331,
376
- "bytes": 13601194,
377
- "pages": 37,
378
- "shards": 4
379
- },
380
- {
381
- "seconds": 1.1585651320001489,
382
- "bytes": 6030177,
383
- "pages": 31,
384
- "shards": 2
385
- },
386
- {
387
- "seconds": 0.9087623919995167,
388
- "bytes": 6312525,
389
- "pages": 37,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.0277426739994553,
394
- "bytes": 6105617,
395
- "pages": 35,
396
- "shards": 2
397
  },
398
  {
399
- "seconds": 0.9305132150002464,
400
- "bytes": 1095891,
401
- "pages": 34,
402
  "shards": 1
403
  },
404
  {
405
- "seconds": 1.8200730230000772,
406
- "bytes": 7217811,
407
- "pages": 25,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 0.9584536009997464,
412
- "bytes": 5627627,
413
  "pages": 33,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.156600534000063,
418
- "bytes": 13506264,
419
- "pages": 39,
420
- "shards": 4
421
  },
422
  {
423
- "seconds": 0.817877588000556,
424
- "bytes": 6663126,
425
- "pages": 35,
426
  "shards": 2
427
  },
428
  {
429
- "seconds": 0.9133121940003548,
430
- "bytes": 1030486,
431
- "pages": 31,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.181683423999857,
436
- "bytes": 12141179,
437
- "pages": 34,
438
- "shards": 4
439
  },
440
  {
441
- "seconds": 1.1883426349995716,
442
- "bytes": 1212298,
443
- "pages": 33,
444
- "shards": 1
445
  },
446
  {
447
- "seconds": 1.1447565099988424,
448
- "bytes": 5933026,
449
- "pages": 35,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.2180850740005553,
454
- "bytes": 13486469,
455
- "pages": 33,
456
- "shards": 4
457
  },
458
  {
459
- "seconds": 1.0477722719988378,
460
- "bytes": 4809637,
461
- "pages": 36,
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.1933620469990274,
466
- "bytes": 7601419,
467
- "pages": 37,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.0362921949999873,
472
- "bytes": 5998979,
473
- "pages": 33,
474
- "shards": 2
475
  },
476
  {
477
- "seconds": 0.7888650649983902,
478
- "bytes": 730238,
479
- "pages": 23,
480
- "shards": 1
481
  },
482
  {
483
- "seconds": 1.1390384229998745,
484
- "bytes": 12216394,
485
- "pages": 44,
486
  "shards": 4
487
  },
488
  {
489
- "seconds": 0.9729804899998271,
490
- "bytes": 6782943,
491
- "pages": 30,
492
- "shards": 2
493
- },
494
- {
495
- "seconds": 1.0282262380005704,
496
- "bytes": 6456198,
497
- "pages": 36,
498
- "shards": 2
499
  },
500
  {
501
- "seconds": 1.007465778999176,
502
- "bytes": 7569670,
503
- "pages": 36,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.248841076001554,
508
- "bytes": 5066284,
509
  "pages": 28,
510
  "shards": 2
511
  },
512
  {
513
- "seconds": 1.0381730779990903,
514
- "bytes": 7573812,
515
- "pages": 37,
516
  "shards": 3
517
  },
518
  {
519
- "seconds": 0.756838974000857,
520
- "bytes": 1060731,
521
- "pages": 33,
 
 
 
 
 
 
522
  "shards": 1
523
  },
524
  {
525
- "seconds": 1.1045795680001902,
526
- "bytes": 11876139,
527
- "pages": 32,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 1.0963092559995857,
532
- "bytes": 12324311,
533
- "pages": 39,
534
- "shards": 5
535
  },
536
  {
537
- "seconds": 0.7185119169989775,
538
- "bytes": 873551,
539
- "pages": 25,
540
  "shards": 1
 
 
 
 
 
 
541
  }
542
  ],
543
  "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": 76127.77658254562,
26
+ "required_pages_per_second_upper_estimate": 0.006598562730089286
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": 9515.972072818202,
40
+ "required_pages_per_second_upper_estimate": 0.0037960635363085216
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": 19031.944145636404,
54
+ "required_pages_per_second_upper_estimate": 0.004402705327316008
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": 28547.916218454604,
68
+ "required_pages_per_second_upper_estimate": 0.0017384183183630445
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": 28547.916218454604,
82
+ "required_pages_per_second_upper_estimate": 0.002197697616186597
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": 28547.916218454604,
96
+ "required_pages_per_second_upper_estimate": 0.35446957570377097
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": 0.09582121260153542,
162
+ "producer_p95_seconds": 0.27117204399837647,
163
+ "gpu_update_seconds": 0.9684769910501257,
164
+ "producer_batches_per_second": 10.436102537737831,
165
+ "gpu_batches_per_second": 1.0325490530401695,
166
+ "producer_headroom": 10.107125184038916,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.0455734451078,
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": 416734.52252620424,
174
+ "upload_bytes_per_second": 6385147.086400999,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 89.42152745093914,
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": 1.8200730230000772,
184
+ "bytes": 7217811,
185
+ "pages": 25,
186
+ "shards": 3
187
  },
188
  {
189
+ "seconds": 0.9584536009997464,
190
+ "bytes": 5627627,
191
+ "pages": 33,
192
+ "shards": 2
193
+ },
194
+ {
195
+ "seconds": 1.156600534000063,
196
+ "bytes": 13506264,
197
+ "pages": 39,
198
+ "shards": 4
199
+ },
200
+ {
201
+ "seconds": 0.817877588000556,
202
+ "bytes": 6663126,
203
+ "pages": 35,
204
+ "shards": 2
205
+ },
206
+ {
207
+ "seconds": 0.9133121940003548,
208
+ "bytes": 1030486,
209
+ "pages": 31,
210
  "shards": 1
211
  },
212
  {
213
+ "seconds": 1.181683423999857,
214
+ "bytes": 12141179,
215
+ "pages": 34,
216
+ "shards": 4
217
  },
218
  {
219
+ "seconds": 1.1883426349995716,
220
+ "bytes": 1212298,
221
+ "pages": 33,
222
  "shards": 1
223
  },
224
  {
225
+ "seconds": 1.1447565099988424,
226
+ "bytes": 5933026,
227
  "pages": 35,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.2180850740005553,
232
+ "bytes": 13486469,
233
+ "pages": 33,
234
+ "shards": 4
235
  },
236
  {
237
+ "seconds": 1.0477722719988378,
238
+ "bytes": 4809637,
239
+ "pages": 36,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.1933620469990274,
244
+ "bytes": 7601419,
245
  "pages": 37,
246
  "shards": 3
247
  },
248
  {
249
+ "seconds": 1.0362921949999873,
250
+ "bytes": 5998979,
251
+ "pages": 33,
252
  "shards": 2
253
  },
254
  {
255
+ "seconds": 0.7888650649983902,
256
+ "bytes": 730238,
257
+ "pages": 23,
258
+ "shards": 1
259
  },
260
  {
261
+ "seconds": 1.1390384229998745,
262
+ "bytes": 12216394,
263
+ "pages": 44,
264
  "shards": 4
265
  },
266
  {
267
+ "seconds": 0.9729804899998271,
268
+ "bytes": 6782943,
269
+ "pages": 30,
270
+ "shards": 2
271
  },
272
  {
273
+ "seconds": 1.0282262380005704,
274
+ "bytes": 6456198,
275
+ "pages": 36,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.007465778999176,
280
+ "bytes": 7569670,
281
+ "pages": 36,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.248841076001554,
286
+ "bytes": 5066284,
287
+ "pages": 28,
288
+ "shards": 2
289
  },
290
  {
291
+ "seconds": 1.0381730779990903,
292
+ "bytes": 7573812,
293
  "pages": 37,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 0.756838974000857,
298
+ "bytes": 1060731,
299
+ "pages": 33,
300
  "shards": 1
301
  },
302
  {
303
+ "seconds": 1.1045795680001902,
304
+ "bytes": 11876139,
305
+ "pages": 32,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.0963092559995857,
310
+ "bytes": 12324311,
311
+ "pages": 39,
312
+ "shards": 5
313
  },
314
  {
315
+ "seconds": 0.7185119169989775,
316
+ "bytes": 873551,
317
+ "pages": 25,
318
  "shards": 1
319
  },
320
  {
321
+ "seconds": 0.8877478260001226,
322
+ "bytes": 6165145,
323
+ "pages": 35,
324
+ "shards": 2
325
+ },
326
+ {
327
+ "seconds": 0.9784168599999248,
328
+ "bytes": 2018917,
329
+ "pages": 19,
330
+ "shards": 2
331
+ },
332
+ {
333
+ "seconds": 1.1833946610004205,
334
+ "bytes": 10020680,
335
+ "pages": 36,
336
  "shards": 3
337
  },
338
  {
339
+ "seconds": 0.9556004539990681,
340
+ "bytes": 6956463,
341
  "pages": 35,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.286864922998575,
346
+ "bytes": 7382695,
347
+ "pages": 34,
348
  "shards": 3
349
  },
350
  {
351
+ "seconds": 0.9901690320002672,
352
+ "bytes": 6106536,
353
+ "pages": 30,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.1473440179997851,
358
+ "bytes": 6606039,
359
+ "pages": 35,
360
  "shards": 3
361
  },
362
  {
363
+ "seconds": 1.092060394001237,
364
+ "bytes": 6397904,
365
  "pages": 35,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 0.9357377939995786,
370
+ "bytes": 1197833,
371
+ "pages": 33,
372
+ "shards": 1
373
  },
374
  {
375
+ "seconds": 1.0255696429994714,
376
+ "bytes": 13323829,
377
+ "pages": 36,
378
+ "shards": 4
379
  },
380
  {
381
+ "seconds": 1.3980831969984138,
382
+ "bytes": 10844649,
383
+ "pages": 37,
384
+ "shards": 3
385
  },
386
  {
387
+ "seconds": 0.9555492769995908,
388
+ "bytes": 915953,
389
  "pages": 28,
390
+ "shards": 1
391
  },
392
  {
393
+ "seconds": 1.0586995369994838,
394
+ "bytes": 7742624,
395
  "pages": 35,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 0.8267605399996683,
400
+ "bytes": 1098322,
401
  "pages": 34,
402
  "shards": 1
403
  },
404
  {
405
+ "seconds": 1.0487060740015295,
406
+ "bytes": 7770470,
407
+ "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.5646253450013319,
412
+ "bytes": 15407220,
413
+ "pages": 37,
414
+ "shards": 4
415
  },
416
  {
417
+ "seconds": 0.8243736469994474,
418
+ "bytes": 1000268,
419
+ "pages": 30,
420
  "shards": 1
421
  },
422
  {
423
+ "seconds": 1.0672549959999742,
424
+ "bytes": 7538580,
425
+ "pages": 38,
426
  "shards": 3
427
  },
428
  {
429
+ "seconds": 1.1064888179989794,
430
+ "bytes": 5541578,
431
  "pages": 33,
432
  "shards": 2
433
  },
434
  {
435
+ "seconds": 0.9643621880004503,
436
+ "bytes": 7410728,
437
+ "pages": 34,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 0.8883736560001125,
442
+ "bytes": 6293157,
443
+ "pages": 36,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 0.9868076469992957,
448
+ "bytes": 927962,
449
+ "pages": 26,
450
  "shards": 1
451
  },
452
  {
453
+ "seconds": 1.2888746650005487,
454
+ "bytes": 18374728,
455
+ "pages": 40,
456
+ "shards": 5
457
  },
458
  {
459
+ "seconds": 1.1437757880012214,
460
+ "bytes": 6056576,
461
+ "pages": 36,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.102731920998849,
466
+ "bytes": 2598092,
467
+ "pages": 36,
468
  "shards": 2
469
  },
470
  {
471
+ "seconds": 1.1366644360005012,
472
+ "bytes": 5837515,
473
+ "pages": 27,
474
+ "shards": 2
475
  },
476
  {
477
+ "seconds": 0.8290303609992407,
478
+ "bytes": 5336251,
479
+ "pages": 39,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.0252480340004695,
484
+ "bytes": 7371694,
485
+ "pages": 29,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.1864866850010003,
490
+ "bytes": 11535625,
491
+ "pages": 35,
492
+ "shards": 3
493
  },
494
  {
495
+ "seconds": 0.9566465869993408,
496
+ "bytes": 1137762,
497
+ "pages": 35,
498
+ "shards": 2
499
  },
500
  {
501
+ "seconds": 1.0941519280004286,
502
+ "bytes": 11375746,
503
+ "pages": 34,
504
  "shards": 4
505
  },
506
  {
507
+ "seconds": 0.7203741030007222,
508
+ "bytes": 1052913,
509
+ "pages": 34,
510
+ "shards": 1
 
 
 
 
 
 
511
  },
512
  {
513
+ "seconds": 1.0976181589994667,
514
+ "bytes": 7695785,
515
+ "pages": 37,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 1.024126566000632,
520
+ "bytes": 5953181,
521
  "pages": 28,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.452807390000089,
526
+ "bytes": 11684781,
527
+ "pages": 40,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 0.8971567329990648,
532
+ "bytes": 7415356,
533
+ "pages": 29,
534
+ "shards": 3
535
+ },
536
+ {
537
+ "seconds": 0.9046226490008848,
538
+ "bytes": 1040439,
539
+ "pages": 34,
540
  "shards": 1
541
  },
542
  {
543
+ "seconds": 1.043625083999359,
544
+ "bytes": 6212867,
545
+ "pages": 35,
546
+ "shards": 2
547
  },
548
  {
549
+ "seconds": 1.3019511900001817,
550
+ "bytes": 12069002,
551
+ "pages": 36,
552
+ "shards": 4
553
  },
554
  {
555
+ "seconds": 0.8234439619991463,
556
+ "bytes": 1074472,
557
+ "pages": 32,
558
  "shards": 1
559
+ },
560
+ {
561
+ "seconds": 1.0881272910010011,
562
+ "bytes": 13091178,
563
+ "pages": 33,
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:ad5414aa9cc5e09a287ca937f00d1a790494a4a5870ec4a0ca71e8a2bf9bfdc0
3
- size 721772735
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54b03f408ed31fec8a13f34a9c659787ebd6b16e7fa218e2cea4dd5412383f2f
3
+ size 721776959