appvoid commited on
Commit
dab8d92
·
verified ·
1 Parent(s): c9bf674

Qwen90M step 325000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +48 -48
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +277 -133
  6. trainer.pt +2 -2
  7. training_config.json +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 320000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "7af0a447c456f6760d4f6e84edc350c5e69416f8c3a7a61824547f910318f805",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "38c2f886ccf49b55cd34e215a6a6c1aa8127d46138bf70445d4ca70032231d6c",
12
- "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a",
14
- "shard_scheduler.json": "d5343011e97e001a66563c4b9f78ab277480b08424f02858b0f8f0948b4c217f",
15
- "coverage.json": "1f8b82802b7d23b4ac70b19b7e71ccea4fd98cb1ced9c8e431325fe81b59768d",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 325000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64",
12
+ "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a",
14
+ "shard_scheduler.json": "fcc8218f92e7a2b7ab38bc12e80b76d8e5163a9d89f4285f1044c2f3f6d27145",
15
+ "coverage.json": "877bcc297c777298fe977b7e68fa8d3e9711497b231150a19f8f8e0f118f5cab",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 320000,
3
  "stats": {
4
- "climbmix/blocks": 11520000,
5
- "climbmix/targets": 19257096294,
6
- "climbmix/input_tokens": 19258316982,
7
- "ultra_qa/blocks": 4320000,
8
- "ultra_qa/targets": 7047544993,
9
- "ultra_qa/input_tokens": 7047558585,
10
- "cortex/blocks": 4320000,
11
- "cortex/targets": 164372097,
12
- "cortex/input_tokens": 6867485637,
13
- "ultra_style/blocks": 4320000,
14
- "ultra_style/targets": 6833023501,
15
- "ultra_style/input_tokens": 6833381920,
16
- "openmath/blocks": 2880000,
17
- "openmath/targets": 4450371997,
18
- "openmath/input_tokens": 4451855399,
19
- "rewrite6/blocks": 1440000,
20
- "rewrite6/targets": 2360725625,
21
- "rewrite6/input_tokens": 2363313839
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1578,
26
- "row": 267,
27
  "start": {
28
- "file": 76,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 31,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 6463755
38
  },
39
  "rewrite6": {
40
- "page": 47,
41
- "row": 2882,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 192512
47
  }
48
  },
49
- "epoch": 17,
50
  "offset": 2048,
51
- "documents": 3687262
52
  },
53
  "openmath": {
54
- "page": 986,
55
- "row": 621,
56
  "start": {
57
  "file": 9,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 108,
61
- "row": 806
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 4039277
67
  },
68
  "ultra_qa": {
69
- "page": 426,
70
- "row": 1436,
71
  "start": {
72
  "file": 3,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 2,
76
- "row": 56755
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 4096,
81
- "documents": 1746332
82
  },
83
  "ultra_style": {
84
- "page": 650,
85
- "row": 3626,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
- "row": 10908
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 2666026
97
  },
98
  "cortex": {
99
- "page": 83788,
100
- "row": 117,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 10724981
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 325000,
3
  "stats": {
4
+ "climbmix/blocks": 11700000,
5
+ "climbmix/targets": 19557876823,
6
+ "climbmix/input_tokens": 19559116697,
7
+ "ultra_qa/blocks": 4387500,
8
+ "ultra_qa/targets": 7157752008,
9
+ "ultra_qa/input_tokens": 7157765811,
10
+ "cortex/blocks": 4387500,
11
+ "cortex/targets": 166932432,
12
+ "cortex/input_tokens": 6974752559,
13
+ "ultra_style/blocks": 4387500,
14
+ "ultra_style/targets": 6940030981,
15
+ "ultra_style/input_tokens": 6940394926,
16
+ "openmath/blocks": 2925000,
17
+ "openmath/targets": 4519854927,
18
+ "openmath/input_tokens": 4521361389,
19
+ "rewrite6/blocks": 1462500,
20
+ "rewrite6/targets": 2397565289,
21
+ "rewrite6/input_tokens": 2400193805
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1602,
26
+ "row": 3042,
27
  "start": {
28
+ "file": 77,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 44,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 6564834
38
  },
39
  "rewrite6": {
40
+ "page": 11,
41
+ "row": 2354,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 45056
47
  }
48
  },
49
+ "epoch": 18,
50
  "offset": 2048,
51
+ "documents": 3744682
52
  },
53
  "openmath": {
54
+ "page": 1001,
55
+ "row": 2216,
56
  "start": {
57
  "file": 9,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 170,
61
+ "row": 246
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 4102312
67
  },
68
  "ultra_qa": {
69
+ "page": 432,
70
+ "row": 4084,
71
  "start": {
72
  "file": 3,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 3,
76
+ "row": 16720
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 2048,
81
+ "documents": 1773556
82
  },
83
  "ultra_style": {
84
+ "page": 660,
85
+ "row": 3969,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 7,
91
+ "row": 51868
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 4096,
96
+ "documents": 2707329
97
  },
98
  "cortex": {
99
+ "page": 85096,
100
+ "row": 43,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 10892331
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 320000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "7af0a447c456f6760d4f6e84edc350c5e69416f8c3a7a61824547f910318f805", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "38c2f886ccf49b55cd34e215a6a6c1aa8127d46138bf70445d4ca70032231d6c", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "d5343011e97e001a66563c4b9f78ab277480b08424f02858b0f8f0948b4c217f", "coverage.json": "1f8b82802b7d23b4ac70b19b7e71ccea4fd98cb1ced9c8e431325fe81b59768d", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000320000"}
 
1
+ {"step": 325000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "90ad39d684ac792f235cbfbeef498d4c81f09977d8f89fb8c3bdd41e515f347a", "shard_scheduler.json": "fcc8218f92e7a2b7ab38bc12e80b76d8e5163a9d89f4285f1044c2f3f6d27145", "coverage.json": "877bcc297c777298fe977b7e68fa8d3e9711497b231150a19f8f8e0f118f5cab", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000325000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7af0a447c456f6760d4f6e84edc350c5e69416f8c3a7a61824547f910318f805
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8ad3e84656fa5ff64e4509de8f55d1d215edb737955024a56bff4c9c42a4dde
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": 75458.65267281262,
26
- "required_pages_per_second_upper_estimate": 0.006540564765472674
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": 9432.331584101577,
40
- "required_pages_per_second_upper_estimate": 0.003762698094822713
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": 18864.663168203155,
54
- "required_pages_per_second_upper_estimate": 0.004364007817231477
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": 28296.994752304734,
68
- "required_pages_per_second_upper_estimate": 0.0017231385175577071
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": 28296.994752304734,
82
- "required_pages_per_second_upper_estimate": 0.002178380987127306
83
  },
84
  {
85
  "source": "cortex",
@@ -91,9 +91,9 @@
91
  "read_seconds": 0.042780299000014566,
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": 28296.994752304734,
96
- "required_pages_per_second_upper_estimate": 0.35135397087431536
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,138 +155,30 @@
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
157
  "selected": {
158
- "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
- "producer_batch_seconds": 0.0511901649687303,
162
- "producer_p95_seconds": 0.18669305799994618,
163
- "gpu_update_seconds": 0.9770648877032472,
164
- "producer_batches_per_second": 19.535002487506215,
165
- "gpu_batches_per_second": 1.0234734791776885,
166
- "producer_headroom": 19.086965011737917,
167
  "sustainable": true,
168
- "buffer_seconds": 60.57802303760133,
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": 395163.83161593793,
174
- "upload_bytes_per_second": 5372382.916989888,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 84.79295932158092,
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": 1.0731404800000064,
184
- "bytes": 44464,
185
- "pages": 1,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 1.131805757000052,
190
- "bytes": 882364,
191
- "pages": 25,
192
- "shards": 1
193
- },
194
- {
195
- "seconds": 1.1303505340000584,
196
- "bytes": 11613053,
197
- "pages": 29,
198
- "shards": 3
199
- },
200
- {
201
- "seconds": 0.9671371199999612,
202
- "bytes": 2609325,
203
- "pages": 34,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 0.9158913300000222,
208
- "bytes": 6129878,
209
- "pages": 32,
210
- "shards": 2
211
- },
212
- {
213
- "seconds": 1.5951903079999283,
214
- "bytes": 6553135,
215
- "pages": 34,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 0.9309580040001038,
220
- "bytes": 6008731,
221
- "pages": 35,
222
- "shards": 2
223
- },
224
- {
225
- "seconds": 1.3192515309999635,
226
- "bytes": 6183122,
227
- "pages": 35,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 1.2959069240000645,
232
- "bytes": 8294236,
233
- "pages": 30,
234
- "shards": 3
235
- },
236
- {
237
- "seconds": 1.3743943120000495,
238
- "bytes": 10663078,
239
- "pages": 35,
240
- "shards": 3
241
- },
242
- {
243
- "seconds": 0.8003151779998916,
244
- "bytes": 1078242,
245
- "pages": 33,
246
- "shards": 1
247
- },
248
- {
249
- "seconds": 1.0970220180001888,
250
- "bytes": 7456296,
251
- "pages": 36,
252
- "shards": 3
253
- },
254
- {
255
- "seconds": 1.0635986559996127,
256
- "bytes": 6186061,
257
- "pages": 35,
258
- "shards": 2
259
- },
260
- {
261
- "seconds": 1.1765199270002995,
262
- "bytes": 6893561,
263
- "pages": 28,
264
- "shards": 3
265
- },
266
- {
267
- "seconds": 3.6869929439999396,
268
- "bytes": 11547756,
269
- "pages": 29,
270
- "shards": 3
271
- },
272
- {
273
- "seconds": 0.8728801329998532,
274
- "bytes": 1038671,
275
- "pages": 33,
276
- "shards": 1
277
- },
278
- {
279
- "seconds": 1.359243317000164,
280
- "bytes": 7794752,
281
- "pages": 37,
282
- "shards": 4
283
- },
284
- {
285
- "seconds": 0.911286476999976,
286
- "bytes": 856023,
287
- "pages": 24,
288
- "shards": 1
289
- },
290
  {
291
  "seconds": 1.8453115359998264,
292
  "bytes": 9879354,
@@ -418,6 +310,258 @@
418
  "bytes": 11609605,
419
  "pages": 32,
420
  "shards": 3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
421
  }
422
  ],
423
  "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": 74569.75015202582,
26
+ "required_pages_per_second_upper_estimate": 0.006463516947873717
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": 9321.218769003228,
40
+ "required_pages_per_second_upper_estimate": 0.0037183735315953516
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": 18642.437538006456,
54
+ "required_pages_per_second_upper_estimate": 0.004312599828722984
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": 27963.65630700968,
68
+ "required_pages_per_second_upper_estimate": 0.0017028399551308995
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": 27963.65630700968,
82
+ "required_pages_per_second_upper_estimate": 0.0021527196708686185
83
  },
84
  {
85
  "source": "cortex",
 
91
  "read_seconds": 0.042780299000014566,
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
+ "estimated_pages_for_buffer": 21,
95
+ "required_token_positions_per_second": 27963.65630700968,
96
+ "required_pages_per_second_upper_estimate": 0.34721502299576196
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
157
  "selected": {
158
+ "prefetch_batches": 61,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
+ "producer_batch_seconds": 0.074832562304735,
162
+ "producer_p95_seconds": 0.19214319799903024,
163
+ "gpu_update_seconds": 0.988711908645131,
164
+ "producer_batches_per_second": 13.363166637643321,
165
+ "gpu_batches_per_second": 1.011416967122746,
166
+ "producer_headroom": 13.212321991847265,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.31142642735299,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 86,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 299588.0660923766,
174
+ "upload_bytes_per_second": 4058653.93467418,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 64.28462493017648,
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": 1.8453115359998264,
184
  "bytes": 9879354,
 
310
  "bytes": 11609605,
311
  "pages": 32,
312
  "shards": 3
313
+ },
314
+ {
315
+ "seconds": 1.5608764529997643,
316
+ "bytes": 2529116,
317
+ "pages": 35,
318
+ "shards": 2
319
+ },
320
+ {
321
+ "seconds": 1.0529914669996288,
322
+ "bytes": 5481895,
323
+ "pages": 19,
324
+ "shards": 2
325
+ },
326
+ {
327
+ "seconds": 1.0121831689993996,
328
+ "bytes": 5661667,
329
+ "pages": 37,
330
+ "shards": 2
331
+ },
332
+ {
333
+ "seconds": 1.0493596160004017,
334
+ "bytes": 7456331,
335
+ "pages": 30,
336
+ "shards": 3
337
+ },
338
+ {
339
+ "seconds": 0.9663659530006043,
340
+ "bytes": 5966016,
341
+ "pages": 35,
342
+ "shards": 2
343
+ },
344
+ {
345
+ "seconds": 0.7928353870001956,
346
+ "bytes": 647903,
347
+ "pages": 20,
348
+ "shards": 1
349
+ },
350
+ {
351
+ "seconds": 1.4357455670005947,
352
+ "bytes": 17610080,
353
+ "pages": 43,
354
+ "shards": 5
355
+ },
356
+ {
357
+ "seconds": 0.884776743000657,
358
+ "bytes": 1180635,
359
+ "pages": 33,
360
+ "shards": 1
361
+ },
362
+ {
363
+ "seconds": 1.2576917689993934,
364
+ "bytes": 6183813,
365
+ "pages": 36,
366
+ "shards": 3
367
+ },
368
+ {
369
+ "seconds": 1.3994414539993159,
370
+ "bytes": 7586010,
371
+ "pages": 36,
372
+ "shards": 3
373
+ },
374
+ {
375
+ "seconds": 1.0743332829997598,
376
+ "bytes": 4951926,
377
+ "pages": 31,
378
+ "shards": 2
379
+ },
380
+ {
381
+ "seconds": 1.3130787809996036,
382
+ "bytes": 11524150,
383
+ "pages": 30,
384
+ "shards": 3
385
+ },
386
+ {
387
+ "seconds": 1.0262837310001487,
388
+ "bytes": 2640097,
389
+ "pages": 35,
390
+ "shards": 2
391
+ },
392
+ {
393
+ "seconds": 1.0240621299999475,
394
+ "bytes": 6061414,
395
+ "pages": 35,
396
+ "shards": 2
397
+ },
398
+ {
399
+ "seconds": 1.345420446999924,
400
+ "bytes": 12056021,
401
+ "pages": 35,
402
+ "shards": 4
403
+ },
404
+ {
405
+ "seconds": 1.0655096850005066,
406
+ "bytes": 1134459,
407
+ "pages": 33,
408
+ "shards": 1
409
+ },
410
+ {
411
+ "seconds": 1.1734728310002538,
412
+ "bytes": 6054797,
413
+ "pages": 35,
414
+ "shards": 2
415
+ },
416
+ {
417
+ "seconds": 1.2160326559996975,
418
+ "bytes": 2369678,
419
+ "pages": 28,
420
+ "shards": 2
421
+ },
422
+ {
423
+ "seconds": 1.405440059999819,
424
+ "bytes": 15806866,
425
+ "pages": 36,
426
+ "shards": 4
427
+ },
428
+ {
429
+ "seconds": 1.6009747379994224,
430
+ "bytes": 7396060,
431
+ "pages": 37,
432
+ "shards": 3
433
+ },
434
+ {
435
+ "seconds": 0.9060076790001403,
436
+ "bytes": 1026778,
437
+ "pages": 33,
438
+ "shards": 1
439
+ },
440
+ {
441
+ "seconds": 1.30779888699999,
442
+ "bytes": 10925777,
443
+ "pages": 35,
444
+ "shards": 3
445
+ },
446
+ {
447
+ "seconds": 0.9117075719987042,
448
+ "bytes": 2359249,
449
+ "pages": 28,
450
+ "shards": 2
451
+ },
452
+ {
453
+ "seconds": 0.9851995050012192,
454
+ "bytes": 5930447,
455
+ "pages": 34,
456
+ "shards": 2
457
+ },
458
+ {
459
+ "seconds": 1.2710061699999642,
460
+ "bytes": 13577169,
461
+ "pages": 42,
462
+ "shards": 4
463
+ },
464
+ {
465
+ "seconds": 0.8450371549988631,
466
+ "bytes": 895470,
467
+ "pages": 25,
468
+ "shards": 1
469
+ },
470
+ {
471
+ "seconds": 32.774606008000774,
472
+ "bytes": 10411787,
473
+ "pages": 35,
474
+ "shards": 3
475
+ },
476
+ {
477
+ "seconds": 1.2439632269997674,
478
+ "bytes": 7901436,
479
+ "pages": 46,
480
+ "shards": 3
481
+ },
482
+ {
483
+ "seconds": 0.8556265790011821,
484
+ "bytes": 1138175,
485
+ "pages": 33,
486
+ "shards": 1
487
+ },
488
+ {
489
+ "seconds": 1.1632403840012557,
490
+ "bytes": 11705717,
491
+ "pages": 30,
492
+ "shards": 3
493
+ },
494
+ {
495
+ "seconds": 1.197896118999779,
496
+ "bytes": 12261236,
497
+ "pages": 40,
498
+ "shards": 4
499
+ },
500
+ {
501
+ "seconds": 0.8173304430001735,
502
+ "bytes": 1136030,
503
+ "pages": 33,
504
+ "shards": 1
505
+ },
506
+ {
507
+ "seconds": 1.1651587139986077,
508
+ "bytes": 5801842,
509
+ "pages": 20,
510
+ "shards": 2
511
+ },
512
+ {
513
+ "seconds": 1.0791701340003783,
514
+ "bytes": 2256184,
515
+ "pages": 26,
516
+ "shards": 2
517
+ },
518
+ {
519
+ "seconds": 1.5194503539987636,
520
+ "bytes": 9946744,
521
+ "pages": 38,
522
+ "shards": 3
523
+ },
524
+ {
525
+ "seconds": 1.188283532999776,
526
+ "bytes": 2401103,
527
+ "pages": 29,
528
+ "shards": 2
529
+ },
530
+ {
531
+ "seconds": 1.182281362998765,
532
+ "bytes": 11896636,
533
+ "pages": 38,
534
+ "shards": 3
535
+ },
536
+ {
537
+ "seconds": 0.8353424750002887,
538
+ "bytes": 975468,
539
+ "pages": 29,
540
+ "shards": 1
541
+ },
542
+ {
543
+ "seconds": 1.1824829600009252,
544
+ "bytes": 12495213,
545
+ "pages": 37,
546
+ "shards": 4
547
+ },
548
+ {
549
+ "seconds": 1.254413848000695,
550
+ "bytes": 6063292,
551
+ "pages": 35,
552
+ "shards": 3
553
+ },
554
+ {
555
+ "seconds": 0.9384880159996101,
556
+ "bytes": 1115458,
557
+ "pages": 34,
558
+ "shards": 1
559
+ },
560
+ {
561
+ "seconds": 1.1859958829991228,
562
+ "bytes": 7375397,
563
+ "pages": 29,
564
+ "shards": 3
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:38c2f886ccf49b55cd34e215a6a6c1aa8127d46138bf70445d4ca70032231d6c
3
- size 721783871
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e8357e00b06e5d5d31d5c5c4045de3cc1e94f50d29f778b0036c1689949df64
3
+ size 721767807
training_config.json CHANGED
@@ -46,8 +46,8 @@
46
  "cpu_threads": 4,
47
  "page_rows": 4096,
48
  "cortex_page_rows": 128,
49
- "prefetch_batches": 62,
50
- "prefetch_low_water": 31,
51
  "memory_pages": 12,
52
  "cache_gb": 20,
53
  "force_rebenchmark": false,
 
46
  "cpu_threads": 4,
47
  "page_rows": 4096,
48
  "cortex_page_rows": 128,
49
+ "prefetch_batches": 61,
50
+ "prefetch_low_water": 30,
51
  "memory_pages": 12,
52
  "cache_gb": 20,
53
  "force_rebenchmark": false,