appvoid commited on
Commit
ca0d798
·
verified ·
1 Parent(s): 3095066

Qwen90M step 150000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +48 -48
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +244 -244
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 145000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "f82ef0b41a9f5a444a6d4c213cf42228dda7b7856086639bd309c8f1264486c9",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "81681141f6dfe127e198e16da88d8c38af13ada9c9f0a4963e0cd992eb0b365c",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4",
14
- "shard_scheduler.json": "33c9db18de388458070d6df886e5bda2cedf9dabd3e716f543d2e756f18de9cc",
15
- "coverage.json": "088c76466a16e8303fb88ede838eb2896c816dbf26f02d114c54169dacd144e5",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 150000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4",
14
+ "shard_scheduler.json": "34c9075a0f66a3d08c183399d506bb22fcf18c184ad147564967e4708a38ba3f",
15
+ "coverage.json": "a2f785366449384f55526192ff2e3551a8f132be1f93c9467f8d4fa84286825c",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 145000,
3
  "stats": {
4
- "climbmix/blocks": 5220000,
5
- "climbmix/targets": 8723475320,
6
- "climbmix/input_tokens": 8724028649,
7
- "ultra_qa/blocks": 1957500,
8
- "ultra_qa/targets": 3193055973,
9
- "ultra_qa/input_tokens": 3193062132,
10
- "cortex/blocks": 1957500,
11
- "cortex/targets": 74442843,
12
- "cortex/input_tokens": 3111300467,
13
- "ultra_style/blocks": 1957500,
14
- "ultra_style/targets": 3096311658,
15
- "ultra_style/input_tokens": 3096473872,
16
- "openmath/blocks": 1305000,
17
- "openmath/targets": 2016316779,
18
- "openmath/input_tokens": 2016988997,
19
- "rewrite6/blocks": 652500,
20
- "rewrite6/targets": 1069684284,
21
- "rewrite6/input_tokens": 1070857144
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 715,
26
- "row": 2681,
27
  "start": {
28
- "file": 34,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 44,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 0,
37
- "documents": 2931321
38
  },
39
  "rewrite6": {
40
- "page": 6,
41
- "row": 3005,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 24576
47
  }
48
  },
49
  "epoch": 8,
50
- "offset": 2048,
51
- "documents": 1670813
52
  },
53
  "openmath": {
54
- "page": 446,
55
- "row": 3868,
56
  "start": {
57
  "file": 4,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 80,
61
- "row": 216
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 1830684
67
  },
68
  "ultra_qa": {
69
- "page": 193,
70
- "row": 848,
71
  "start": {
72
  "file": 1,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 12512
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 0,
81
- "documents": 791376
82
  },
83
  "ultra_style": {
84
- "page": 294,
85
- "row": 3383,
86
  "start": {
87
  "file": 1,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
- "row": 9157
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 1207607
97
  },
98
  "cortex": {
99
- "page": 37962,
100
- "row": 8,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 4859144
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 150000,
3
  "stats": {
4
+ "climbmix/blocks": 5400000,
5
+ "climbmix/targets": 9024151403,
6
+ "climbmix/input_tokens": 9024723849,
7
+ "ultra_qa/blocks": 2025000,
8
+ "ultra_qa/targets": 3303216837,
9
+ "ultra_qa/input_tokens": 3303223204,
10
+ "cortex/blocks": 2025000,
11
+ "cortex/targets": 77013811,
12
+ "cortex/input_tokens": 3218659538,
13
+ "ultra_style/blocks": 2025000,
14
+ "ultra_style/targets": 3203109017,
15
+ "ultra_style/input_tokens": 3203277056,
16
+ "openmath/blocks": 1350000,
17
+ "openmath/targets": 2085884301,
18
+ "openmath/input_tokens": 2086579643,
19
+ "rewrite6/blocks": 675000,
20
+ "rewrite6/targets": 1106511308,
21
+ "rewrite6/input_tokens": 1107724623
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 740,
26
+ "row": 1687,
27
  "start": {
28
+ "file": 35,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 60,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 3032727
38
  },
39
  "rewrite6": {
40
+ "page": 20,
41
+ "row": 3357,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 81920
47
  }
48
  },
49
  "epoch": 8,
50
+ "offset": 0,
51
+ "documents": 1728509
52
  },
53
  "openmath": {
54
+ "page": 462,
55
+ "row": 1399,
56
  "start": {
57
  "file": 4,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 145,
61
+ "row": 752
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 1893751
67
  },
68
  "ultra_qa": {
69
+ "page": 199,
70
+ "row": 3551,
71
  "start": {
72
  "file": 1,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 37088
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 818655
82
  },
83
  "ultra_style": {
84
+ "page": 305,
85
+ "row": 256,
86
  "start": {
87
  "file": 1,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
+ "row": 54213
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 1249536
97
  },
98
  "cortex": {
99
+ "page": 39271,
100
+ "row": 122,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 5026810
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 145000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "f82ef0b41a9f5a444a6d4c213cf42228dda7b7856086639bd309c8f1264486c9", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "81681141f6dfe127e198e16da88d8c38af13ada9c9f0a4963e0cd992eb0b365c", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4", "shard_scheduler.json": "33c9db18de388458070d6df886e5bda2cedf9dabd3e716f543d2e756f18de9cc", "coverage.json": "088c76466a16e8303fb88ede838eb2896c816dbf26f02d114c54169dacd144e5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000145000"}
 
1
+ {"step": 150000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "10fa55d76d1a4ec4f9132ef494d950eba8aa9e6cecd5fa1ceb70ba6de35537e4", "shard_scheduler.json": "34c9075a0f66a3d08c183399d506bb22fcf18c184ad147564967e4708a38ba3f", "coverage.json": "a2f785366449384f55526192ff2e3551a8f132be1f93c9467f8d4fa84286825c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000150000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f82ef0b41a9f5a444a6d4c213cf42228dda7b7856086639bd309c8f1264486c9
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf66ab2f55cbc1fb6e95bfaef3a16d5c7c8084d9d77f534a346ea64534d3636b
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": 75819.42597709605,
26
- "required_pages_per_second_upper_estimate": 0.006571835681116382
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": 9477.428247137006,
40
- "required_pages_per_second_upper_estimate": 0.003780687827962744
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": 18954.856494274012,
54
- "required_pages_per_second_upper_estimate": 0.0043848724558696464
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": 28432.28474141102,
68
- "required_pages_per_second_upper_estimate": 0.0017313769680825642
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": 28432.28474141102,
82
- "required_pages_per_second_upper_estimate": 0.002188795984995367
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": 28432.28474141102,
96
- "required_pages_per_second_upper_estimate": 0.35303381975254877
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,409 +158,409 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.22319741781251423,
162
- "producer_p95_seconds": 0.37801911600035965,
163
- "gpu_update_seconds": 0.9724156975584616,
164
- "producer_batches_per_second": 4.4803385711209245,
165
- "gpu_batches_per_second": 1.0283667802883036,
166
- "producer_headroom": 4.3567515569346345,
167
  "sustainable": true,
168
- "buffer_seconds": 60.28977324862462,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 80,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 549008.0145004862,
174
- "upload_bytes_per_second": 5070479.625359649,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 117.80433965931834,
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": 0.8942368999996688,
184
- "bytes": 850966,
185
- "pages": 25,
186
- "shards": 1
187
  },
188
  {
189
- "seconds": 1.4387539480012492,
190
- "bytes": 11588272,
191
- "pages": 38,
192
- "shards": 4
193
  },
194
  {
195
- "seconds": 1.6636693889995513,
196
- "bytes": 11864781,
197
- "pages": 37,
 
 
 
 
 
 
198
  "shards": 3
199
  },
200
  {
201
- "seconds": 1.1722596879990306,
202
- "bytes": 1208158,
203
- "pages": 35,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 1.2069239850025042,
208
- "bytes": 7251816,
209
- "pages": 27,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.169119137000962,
214
- "bytes": 6109087,
215
- "pages": 37,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.4141185940025025,
220
- "bytes": 7704490,
221
- "pages": 38,
 
 
 
 
 
 
222
  "shards": 3
223
  },
224
  {
225
- "seconds": 1.2197817560008843,
226
- "bytes": 5784149,
227
- "pages": 28,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.3641967180010397,
232
- "bytes": 7009240,
233
  "pages": 38,
234
- "shards": 2
235
  },
236
  {
237
- "seconds": 1.4704428500008362,
238
- "bytes": 11028707,
239
  "pages": 30,
240
- "shards": 4
241
  },
242
  {
243
- "seconds": 0.9246157949964982,
244
- "bytes": 1204930,
245
- "pages": 34,
246
  "shards": 1
247
  },
248
  {
249
- "seconds": 1.0427682160006952,
250
- "bytes": 5985007,
 
 
 
 
 
 
 
 
 
 
 
 
251
  "pages": 36,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.1093607039983908,
256
- "bytes": 7338185,
257
  "pages": 30,
258
  "shards": 3
259
  },
260
  {
261
- "seconds": 0.9098767819996283,
262
- "bytes": 5901703,
263
- "pages": 35,
264
- "shards": 2
265
  },
266
  {
267
- "seconds": 1.1044195929971465,
268
- "bytes": 13283238,
269
- "pages": 34,
270
- "shards": 4
271
  },
272
  {
273
- "seconds": 0.8937192140001571,
274
- "bytes": 6207827,
275
- "pages": 36,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 0.9678468759993848,
280
- "bytes": 1151038,
281
  "pages": 35,
282
  "shards": 1
283
  },
284
  {
285
- "seconds": 1.3402203920013562,
286
- "bytes": 12178587,
287
- "pages": 33,
288
  "shards": 4
289
  },
290
  {
291
- "seconds": 0.8894950190006057,
292
- "bytes": 1180133,
293
- "pages": 34,
294
  "shards": 1
295
  },
296
  {
297
- "seconds": 1.1408973270008573,
298
- "bytes": 5952533,
299
- "pages": 28,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 1.4523514239990618,
304
- "bytes": 13500548,
305
- "pages": 42,
306
- "shards": 4
307
  },
308
  {
309
- "seconds": 0.8281688100032625,
310
- "bytes": 661491,
311
  "pages": 19,
312
- "shards": 1
313
  },
314
  {
315
- "seconds": 1.42470605500057,
316
- "bytes": 11031230,
317
- "pages": 41,
318
  "shards": 4
319
  },
320
  {
321
- "seconds": 0.8167821940005524,
322
- "bytes": 1161839,
323
- "pages": 34,
324
- "shards": 1
325
  },
326
  {
327
- "seconds": 0.9240367780002998,
328
- "bytes": 5199805,
329
- "pages": 10,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.4070672229972843,
334
- "bytes": 1010316,
335
- "pages": 29,
336
- "shards": 1
337
  },
338
  {
339
- "seconds": 1.181387053002254,
340
- "bytes": 11521167,
341
- "pages": 38,
342
  "shards": 4
343
  },
344
  {
345
- "seconds": 1.134901017001539,
346
- "bytes": 6770130,
347
- "pages": 32,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.1226096480022534,
352
- "bytes": 7935280,
353
- "pages": 38,
354
- "shards": 3
355
- },
356
- {
357
- "seconds": 0.982757900001161,
358
- "bytes": 5911177,
359
- "pages": 25,
360
  "shards": 2
361
  },
362
  {
363
- "seconds": 1.134857662000286,
364
- "bytes": 5917855,
365
- "pages": 41,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.5188524260011036,
370
- "bytes": 7655478,
371
- "pages": 39,
372
  "shards": 3
373
  },
374
  {
375
- "seconds": 0.7887198950011225,
376
- "bytes": 905384,
377
- "pages": 25,
378
- "shards": 1
379
- },
380
- {
381
- "seconds": 4.007056702997943,
382
- "bytes": 13411793,
383
- "pages": 36,
384
- "shards": 4
385
  },
386
  {
387
- "seconds": 1.3467935920016316,
388
- "bytes": 10736953,
389
- "pages": 38,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 0.8973097269990831,
394
- "bytes": 792959,
395
- "pages": 26,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.2533608519988775,
400
- "bytes": 7837535,
401
- "pages": 39,
402
- "shards": 3
403
  },
404
  {
405
- "seconds": 1.1466966719999618,
406
- "bytes": 6019789,
407
- "pages": 36,
408
- "shards": 2
409
  },
410
  {
411
- "seconds": 1.0688578599983884,
412
- "bytes": 4660632,
413
- "pages": 31,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 1.4090757320009288,
418
- "bytes": 13354980,
419
  "pages": 37,
 
 
 
 
 
 
420
  "shards": 4
421
  },
422
  {
423
- "seconds": 1.0085502450019703,
424
- "bytes": 1017261,
425
- "pages": 26,
426
  "shards": 1
427
  },
428
  {
429
- "seconds": 1.7747774790004769,
430
- "bytes": 7374031,
431
- "pages": 33,
432
  "shards": 3
433
  },
434
  {
435
- "seconds": 0.9477112800013856,
436
- "bytes": 5792070,
437
- "pages": 34,
438
- "shards": 2
439
- },
440
- {
441
- "seconds": 1.0662493220006581,
442
- "bytes": 5942385,
443
  "pages": 30,
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 1.245676413000183,
448
- "bytes": 7693974,
449
- "pages": 38,
450
  "shards": 3
451
  },
452
  {
453
- "seconds": 1.0202759780004271,
454
- "bytes": 6615812,
455
- "pages": 28,
456
  "shards": 2
457
  },
458
  {
459
- "seconds": 1.4830753700007335,
460
- "bytes": 9974834,
461
- "pages": 35,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 1.272736505001376,
466
- "bytes": 7611348,
467
- "pages": 38,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 0.8344998959983059,
472
- "bytes": 1093697,
473
- "pages": 33,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.3662152600008994,
478
- "bytes": 7153184,
479
- "pages": 30,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 1.6717624239972793,
484
- "bytes": 5422811,
485
- "pages": 35,
486
  "shards": 2
487
  },
488
  {
489
- "seconds": 1.1521110989997396,
490
- "bytes": 12146138,
491
- "pages": 38,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.1019259579989011,
496
- "bytes": 7338055,
497
- "pages": 30,
 
 
 
 
 
 
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 0.7986074580003333,
502
- "bytes": 1164850,
503
- "pages": 35,
504
- "shards": 1
505
  },
506
  {
507
- "seconds": 1.6704007100015588,
508
- "bytes": 10030549,
509
  "pages": 31,
510
- "shards": 3
511
  },
512
  {
513
- "seconds": 2.951283398000669,
514
- "bytes": 2617533,
515
- "pages": 35,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 1.1606040469996515,
520
- "bytes": 6236824,
521
- "pages": 36,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 4.655135680000967,
526
- "bytes": 7374926,
527
- "pages": 30,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 1.311550709000585,
532
- "bytes": 11722471,
533
  "pages": 42,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 0.7739134580006066,
538
- "bytes": 5816616,
539
- "pages": 28,
540
- "shards": 2
541
- },
542
- {
543
- "seconds": 1.117633730002126,
544
- "bytes": 7565346,
545
- "pages": 38,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 0.8082868439996673,
550
- "bytes": 1195987,
551
- "pages": 35,
552
- "shards": 1
553
- },
554
- {
555
- "seconds": 1.157742789000622,
556
- "bytes": 11040923,
557
- "pages": 34,
558
- "shards": 4
559
  },
560
  {
561
- "seconds": 0.8454459890017461,
562
- "bytes": 981466,
563
- "pages": 29,
564
  "shards": 1
565
  }
566
  ],
 
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": 75368.18777639217,
26
+ "required_pages_per_second_upper_estimate": 0.006532723497532133
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": 9421.02347204902,
40
+ "required_pages_per_second_upper_estimate": 0.0037581871198536063
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": 18842.04694409804,
54
+ "required_pages_per_second_upper_estimate": 0.004358775951816709
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": 28263.070416147064,
68
+ "required_pages_per_second_upper_estimate": 0.0017210727034729413
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": 28263.070416147064,
82
+ "required_pages_per_second_upper_estimate": 0.0021757693978213155
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": 28263.070416147064,
96
+ "required_pages_per_second_upper_estimate": 0.3509327441566865
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": 0.1606457544066302,
162
+ "producer_p95_seconds": 0.3911977770039812,
163
+ "gpu_update_seconds": 0.9782376646595458,
164
+ "producer_batches_per_second": 6.2248766155922,
165
+ "gpu_batches_per_second": 1.0222464704914302,
166
+ "producer_headroom": 6.089408763230731,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.65073520889184,
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": 362750.0021895322,
174
+ "upload_bytes_per_second": 5470118.08714139,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 77.83770608200525,
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": 1.7747774790004769,
184
+ "bytes": 7374031,
185
+ "pages": 33,
186
+ "shards": 3
187
  },
188
  {
189
+ "seconds": 0.9477112800013856,
190
+ "bytes": 5792070,
191
+ "pages": 34,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.0662493220006581,
196
+ "bytes": 5942385,
197
+ "pages": 30,
198
+ "shards": 2
199
+ },
200
+ {
201
+ "seconds": 1.245676413000183,
202
+ "bytes": 7693974,
203
+ "pages": 38,
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 1.0202759780004271,
208
+ "bytes": 6615812,
209
+ "pages": 28,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.4830753700007335,
214
+ "bytes": 9974834,
215
+ "pages": 35,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.272736505001376,
220
+ "bytes": 7611348,
221
+ "pages": 38,
222
+ "shards": 3
223
  },
224
  {
225
+ "seconds": 0.8344998959983059,
226
+ "bytes": 1093697,
227
+ "pages": 33,
228
+ "shards": 1
229
+ },
230
+ {
231
+ "seconds": 1.3662152600008994,
232
+ "bytes": 7153184,
233
+ "pages": 30,
234
  "shards": 3
235
  },
236
  {
237
+ "seconds": 1.6717624239972793,
238
+ "bytes": 5422811,
239
+ "pages": 35,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.1521110989997396,
244
+ "bytes": 12146138,
245
  "pages": 38,
246
+ "shards": 3
247
  },
248
  {
249
+ "seconds": 1.1019259579989011,
250
+ "bytes": 7338055,
251
  "pages": 30,
252
+ "shards": 3
253
  },
254
  {
255
+ "seconds": 0.7986074580003333,
256
+ "bytes": 1164850,
257
+ "pages": 35,
258
  "shards": 1
259
  },
260
  {
261
+ "seconds": 1.6704007100015588,
262
+ "bytes": 10030549,
263
+ "pages": 31,
264
+ "shards": 3
265
+ },
266
+ {
267
+ "seconds": 2.951283398000669,
268
+ "bytes": 2617533,
269
+ "pages": 35,
270
+ "shards": 2
271
+ },
272
+ {
273
+ "seconds": 1.1606040469996515,
274
+ "bytes": 6236824,
275
  "pages": 36,
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 4.655135680000967,
280
+ "bytes": 7374926,
281
  "pages": 30,
282
  "shards": 3
283
  },
284
  {
285
+ "seconds": 1.311550709000585,
286
+ "bytes": 11722471,
287
+ "pages": 42,
288
+ "shards": 3
289
  },
290
  {
291
+ "seconds": 0.7739134580006066,
292
+ "bytes": 5816616,
293
+ "pages": 28,
294
+ "shards": 2
295
  },
296
  {
297
+ "seconds": 1.117633730002126,
298
+ "bytes": 7565346,
299
+ "pages": 38,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 0.8082868439996673,
304
+ "bytes": 1195987,
305
  "pages": 35,
306
  "shards": 1
307
  },
308
  {
309
+ "seconds": 1.157742789000622,
310
+ "bytes": 11040923,
311
+ "pages": 34,
312
  "shards": 4
313
  },
314
  {
315
+ "seconds": 0.8454459890017461,
316
+ "bytes": 981466,
317
+ "pages": 29,
318
  "shards": 1
319
  },
320
  {
321
+ "seconds": 1.0536112129993853,
322
+ "bytes": 11943546,
323
+ "pages": 39,
324
+ "shards": 3
 
 
 
 
 
 
325
  },
326
  {
327
+ "seconds": 1.0575292020002962,
328
+ "bytes": 2082389,
329
  "pages": 19,
330
+ "shards": 2
331
  },
332
  {
333
+ "seconds": 1.061086557998351,
334
+ "bytes": 10157591,
335
+ "pages": 32,
336
  "shards": 4
337
  },
338
  {
339
+ "seconds": 1.351508649000607,
340
+ "bytes": 6405900,
341
+ "pages": 37,
342
+ "shards": 2
343
  },
344
  {
345
+ "seconds": 0.9840362190006999,
346
+ "bytes": 2390157,
347
+ "pages": 28,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.0104034229989338,
352
+ "bytes": 6117004,
353
+ "pages": 36,
354
+ "shards": 2
355
  },
356
  {
357
+ "seconds": 1.275026989998878,
358
+ "bytes": 16016482,
359
+ "pages": 37,
360
  "shards": 4
361
  },
362
  {
363
+ "seconds": 0.8882673849984712,
364
+ "bytes": 2492218,
365
+ "pages": 31,
366
  "shards": 2
367
  },
368
  {
369
+ "seconds": 1.2180680759993265,
370
+ "bytes": 6022567,
371
+ "pages": 36,
 
 
 
 
 
 
372
  "shards": 2
373
  },
374
  {
375
+ "seconds": 1.038481719999254,
376
+ "bytes": 2380066,
377
+ "pages": 28,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.0937937710004917,
382
+ "bytes": 10154198,
383
+ "pages": 42,
384
  "shards": 3
385
  },
386
  {
387
+ "seconds": 1.047995336000895,
388
+ "bytes": 5924124,
389
+ "pages": 27,
390
+ "shards": 2
 
 
 
 
 
 
391
  },
392
  {
393
+ "seconds": 1.1032993179978803,
394
+ "bytes": 8345826,
395
+ "pages": 35,
396
  "shards": 3
397
  },
398
  {
399
+ "seconds": 0.9737631519965362,
400
+ "bytes": 6188431,
401
+ "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
+ "seconds": 1.3008259939997515,
406
+ "bytes": 12099101,
407
+ "pages": 36,
408
+ "shards": 4
409
  },
410
  {
411
+ "seconds": 0.8470429570006672,
412
+ "bytes": 932725,
413
+ "pages": 27,
414
+ "shards": 1
415
  },
416
  {
417
+ "seconds": 1.1173031580001407,
418
+ "bytes": 6247577,
419
+ "pages": 37,
420
  "shards": 2
421
  },
422
  {
423
+ "seconds": 0.892657821001194,
424
+ "bytes": 2615407,
425
  "pages": 37,
426
+ "shards": 2
427
+ },
428
+ {
429
+ "seconds": 1.2370568289989023,
430
+ "bytes": 16078869,
431
+ "pages": 34,
432
  "shards": 4
433
  },
434
  {
435
+ "seconds": 0.8470632479984488,
436
+ "bytes": 947755,
437
+ "pages": 29,
438
  "shards": 1
439
  },
440
  {
441
+ "seconds": 1.2166951500003051,
442
+ "bytes": 7600114,
443
+ "pages": 37,
444
  "shards": 3
445
  },
446
  {
447
+ "seconds": 1.056514903997595,
448
+ "bytes": 5902748,
 
 
 
 
 
 
449
  "pages": 30,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 1.0406464490006329,
454
+ "bytes": 6219293,
455
+ "pages": 34,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 1.1162217770033749,
460
+ "bytes": 6495200,
461
+ "pages": 37,
462
  "shards": 2
463
  },
464
  {
465
+ "seconds": 1.1027901979978196,
466
+ "bytes": 11991791,
467
+ "pages": 37,
468
  "shards": 3
469
  },
470
  {
471
+ "seconds": 1.5872819810028886,
472
+ "bytes": 6214149,
473
+ "pages": 24,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 1.0351565999953891,
478
+ "bytes": 6216009,
479
+ "pages": 41,
480
+ "shards": 2
481
  },
482
  {
483
+ "seconds": 1.4164216759963892,
484
+ "bytes": 2450413,
485
+ "pages": 33,
486
+ "shards": 2
487
  },
488
  {
489
+ "seconds": 1.0167613629964762,
490
+ "bytes": 6214181,
491
+ "pages": 34,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.3737143940015812,
496
+ "bytes": 10463664,
497
+ "pages": 36,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 0.8019685460021719,
502
+ "bytes": 2186825,
503
+ "pages": 23,
504
+ "shards": 2
505
+ },
506
+ {
507
+ "seconds": 1.2952469239971833,
508
+ "bytes": 12034549,
509
+ "pages": 44,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 0.8924442770003225,
514
+ "bytes": 6155396,
515
+ "pages": 37,
516
+ "shards": 2
517
  },
518
  {
519
+ "seconds": 1.1965257189949625,
520
+ "bytes": 6621394,
521
  "pages": 31,
522
+ "shards": 4
523
  },
524
  {
525
+ "seconds": 1.1455287000062526,
526
+ "bytes": 6484150,
527
+ "pages": 37,
528
  "shards": 2
529
  },
530
  {
531
+ "seconds": 0.8978742559993407,
532
+ "bytes": 2461928,
533
+ "pages": 30,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 0.9092525360028958,
538
+ "bytes": 6015320,
539
+ "pages": 26,
540
+ "shards": 2
541
  },
542
  {
543
+ "seconds": 1.3533267960010562,
544
+ "bytes": 11703451,
545
  "pages": 42,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 1.2227107450016774,
550
+ "bytes": 7523214,
551
+ "pages": 33,
 
 
 
 
 
 
552
  "shards": 3
553
  },
554
  {
555
+ "seconds": 1.0604035779979313,
556
+ "bytes": 6119324,
557
+ "pages": 36,
558
+ "shards": 2
 
 
 
 
 
 
559
  },
560
  {
561
+ "seconds": 1.0826924209977733,
562
+ "bytes": 902486,
563
+ "pages": 27,
564
  "shards": 1
565
  }
566
  ],
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:81681141f6dfe127e198e16da88d8c38af13ada9c9f0a4963e0cd992eb0b365c
3
- size 721762943
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f838656d4de1ef990d79c3418bb1cd36ab78cb5ecc62a6d1f5a9fbe1d6a5d857
3
+ size 721765055