appvoid commited on
Commit
d9dc184
·
verified ·
1 Parent(s): 723a73f

Qwen90M step 615000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +46 -46
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +263 -161
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 610000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "8b25da00cb0b2a963e4cb081606d08666f7bba3d0bf0aee4e1acf0666e72cf45",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "478e3feddd5e715e333af4d01c7c276cb17b2faf46c01877e15a16c415e9bb60",
12
  "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
13
  "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e",
14
- "shard_scheduler.json": "d2dd9104d6ecd28878b172aff3036091ad951fcff01930fc538e5377b692dfb4",
15
- "coverage.json": "fdc409abc87a90ebef70a4935f32045cc301b48f97054e4d2a9d47db721dea80",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"
 
1
  {
2
+ "step": 615000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "5dafb78e58d2434052a2f73226b28c4135549c3ad84f91855ab29a59004932b9",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "8a7925c71cd513cd3d5f53037b00418d6c7a7f8c7b424e1f49846780b2108ebe",
12
  "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb",
13
  "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e",
14
+ "shard_scheduler.json": "9c4ffe74747c4cccb687d247aca424092ee0f5baab8450b599ecbfb62af1ef3d",
15
+ "coverage.json": "caa1ddfdddd2441adccd3fe63c7891dea469bcdb5d7199604e60c94b258433d9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 610000,
3
  "stats": {
4
- "climbmix/blocks": 21960000,
5
- "climbmix/targets": 36709591885,
6
- "climbmix/input_tokens": 36711918536,
7
- "ultra_qa/blocks": 8235000,
8
- "ultra_qa/targets": 13434677003,
9
- "ultra_qa/input_tokens": 13434702938,
10
- "cortex/blocks": 8235000,
11
- "cortex/targets": 313126567,
12
- "cortex/input_tokens": 13090163320,
13
- "ultra_style/blocks": 8235000,
14
- "ultra_style/targets": 13025170521,
15
- "ultra_style/input_tokens": 13025854292,
16
- "openmath/blocks": 5490000,
17
- "openmath/targets": 8483160554,
18
- "openmath/input_tokens": 8485986310,
19
- "rewrite6/blocks": 2745000,
20
- "rewrite6/targets": 4500059266,
21
- "rewrite6/input_tokens": 4504992845
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3007,
26
- "row": 2901,
27
  "start": {
28
- "file": 145,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 44,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
- "documents": 12319573
38
  },
39
  "rewrite6": {
40
- "page": 10,
41
- "row": 3924,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 40960
47
  }
48
  },
49
  "epoch": 34,
50
  "offset": 0,
51
- "documents": 7028620
52
  },
53
  "openmath": {
54
- "page": 1879,
55
- "row": 1544,
56
  "start": {
57
  "file": 17,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 273,
61
- "row": 334
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 7697928
67
  },
68
  "ultra_qa": {
69
- "page": 812,
70
- "row": 2805,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
- "row": 14647
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 3328757
82
  },
83
  "ultra_style": {
84
- "page": 1240,
85
- "row": 2528,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
- "row": 27699
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 5081568
97
  },
98
  "cortex": {
99
- "page": 159712,
100
- "row": 53,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 20443189
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 615000,
3
  "stats": {
4
+ "climbmix/blocks": 22140000,
5
+ "climbmix/targets": 37010537230,
6
+ "climbmix/input_tokens": 37012882752,
7
+ "ultra_qa/blocks": 8302500,
8
+ "ultra_qa/targets": 13544842671,
9
+ "ultra_qa/input_tokens": 13544868823,
10
+ "cortex/blocks": 8302500,
11
+ "cortex/targets": 315671192,
12
+ "cortex/input_tokens": 13197429410,
13
+ "ultra_style/blocks": 8302500,
14
+ "ultra_style/targets": 13131878532,
15
+ "ultra_style/input_tokens": 13132567915,
16
+ "openmath/blocks": 5535000,
17
+ "openmath/targets": 8552607269,
18
+ "openmath/input_tokens": 8555456133,
19
+ "rewrite6/blocks": 2767500,
20
+ "rewrite6/targets": 4536948321,
21
+ "rewrite6/input_tokens": 4541922859
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3032,
26
+ "row": 1301,
27
  "start": {
28
+ "file": 146,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 62,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 0,
37
+ "documents": 12420373
38
  },
39
  "rewrite6": {
40
+ "page": 25,
41
+ "row": 724,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 102400
47
  }
48
  },
49
  "epoch": 34,
50
  "offset": 0,
51
+ "documents": 7086860
52
  },
53
  "openmath": {
54
+ "page": 1894,
55
+ "row": 3308,
56
  "start": {
57
  "file": 17,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 334,
61
+ "row": 774
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 7761132
67
  },
68
  "ultra_qa": {
69
+ "page": 819,
70
+ "row": 1438,
71
  "start": {
72
  "file": 6,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
+ "row": 43319
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 0,
81
+ "documents": 3356062
82
  },
83
  "ultra_style": {
84
+ "page": 1250,
85
+ "row": 3208,
86
  "start": {
87
  "file": 7,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 3,
91
+ "row": 68659
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 5123208
97
  },
98
  "cortex": {
99
+ "page": 161021,
100
+ "row": 86,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 20610774
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 610000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "8b25da00cb0b2a963e4cb081606d08666f7bba3d0bf0aee4e1acf0666e72cf45", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "478e3feddd5e715e333af4d01c7c276cb17b2faf46c01877e15a16c415e9bb60", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "d2dd9104d6ecd28878b172aff3036091ad951fcff01930fc538e5377b692dfb4", "coverage.json": "fdc409abc87a90ebef70a4935f32045cc301b48f97054e4d2a9d47db721dea80", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000610000"}
 
1
+ {"step": 615000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "5dafb78e58d2434052a2f73226b28c4135549c3ad84f91855ab29a59004932b9", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8a7925c71cd513cd3d5f53037b00418d6c7a7f8c7b424e1f49846780b2108ebe", "training_config.json": "6c3c3ccff89dabb5098a22bc207c2f3abaee8b3bffec3f6cf6ad361a63b1a7fb", "performance.json": "539995c1e4203016bfc6259363445c7c25dfb51604d74cd18d2cc7ad51bfb71e", "shard_scheduler.json": "9c4ffe74747c4cccb687d247aca424092ee0f5baab8450b599ecbfb62af1ef3d", "coverage.json": "caa1ddfdddd2441adccd3fe63c7891dea469bcdb5d7199604e60c94b258433d9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "33a459803c8cffa8929a90def998b4ad2c7c52fd1245365a9ee946b5552d845a", "folder": "step-000000615000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8b25da00cb0b2a963e4cb081606d08666f7bba3d0bf0aee4e1acf0666e72cf45
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5dafb78e58d2434052a2f73226b28c4135549c3ad84f91855ab29a59004932b9
3
  size 360622744
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 77013.47401247967,
26
- "required_pages_per_second_upper_estimate": 0.00532797082806254
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9626.684251559958,
40
- "required_pages_per_second_upper_estimate": 0.004030491668101323
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19253.368503119917,
54
- "required_pages_per_second_upper_estimate": 0.004009968913728047
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28880.052754679873,
68
- "required_pages_per_second_upper_estimate": 0.0017313150653558177
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28880.052754679873,
82
- "required_pages_per_second_upper_estimate": 0.002245297071749229
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 28880.052754679873,
96
- "required_pages_per_second_upper_estimate": 0.37044230775233605
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,165 +158,27 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.05319347438283728,
162
- "producer_p95_seconds": 0.16769830500015814,
163
- "gpu_update_seconds": 0.9573389714643016,
164
- "producer_batches_per_second": 18.799298440311073,
165
- "gpu_batches_per_second": 1.0445620932682247,
166
- "producer_headroom": 17.997301033097852,
167
  "sustainable": true,
168
- "buffer_seconds": 60.312355202251,
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": 489298.1172003974,
174
- "upload_bytes_per_second": 5419125.96073942,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 104.9919856739897,
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": 1.386034201999962,
184
- "bytes": 47132,
185
- "pages": 1,
186
- "shards": 1
187
- },
188
- {
189
- "seconds": 1.0035069749999934,
190
- "bytes": 155991,
191
- "pages": 4,
192
- "shards": 1
193
- },
194
- {
195
- "seconds": 0.9961135309999918,
196
- "bytes": 5836820,
197
- "pages": 28,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 1.1866138010000213,
202
- "bytes": 7420176,
203
- "pages": 30,
204
- "shards": 3
205
- },
206
- {
207
- "seconds": 1.0025722640000367,
208
- "bytes": 6023360,
209
- "pages": 35,
210
- "shards": 2
211
- },
212
- {
213
- "seconds": 1.4614262610000424,
214
- "bytes": 6002577,
215
- "pages": 28,
216
- "shards": 2
217
- },
218
- {
219
- "seconds": 1.2785650519999763,
220
- "bytes": 7789202,
221
- "pages": 19,
222
- "shards": 3
223
- },
224
- {
225
- "seconds": 1.1231035699997847,
226
- "bytes": 6063645,
227
- "pages": 36,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 0.8495643409999047,
232
- "bytes": 825792,
233
- "pages": 27,
234
- "shards": 1
235
- },
236
- {
237
- "seconds": 1.5516937659999712,
238
- "bytes": 10821331,
239
- "pages": 34,
240
- "shards": 4
241
- },
242
- {
243
- "seconds": 1.0081327930001862,
244
- "bytes": 1129039,
245
- "pages": 34,
246
- "shards": 1
247
- },
248
- {
249
- "seconds": 1.1534891419998985,
250
- "bytes": 7764027,
251
- "pages": 37,
252
- "shards": 3
253
- },
254
- {
255
- "seconds": 1.4264037609998468,
256
- "bytes": 6430937,
257
- "pages": 21,
258
- "shards": 2
259
- },
260
- {
261
- "seconds": 1.233521140999983,
262
- "bytes": 10999051,
263
- "pages": 45,
264
- "shards": 3
265
- },
266
- {
267
- "seconds": 1.1022731230000318,
268
- "bytes": 7378325,
269
- "pages": 29,
270
- "shards": 3
271
- },
272
- {
273
- "seconds": 0.7817219419998764,
274
- "bytes": 1157795,
275
- "pages": 35,
276
- "shards": 1
277
- },
278
- {
279
- "seconds": 1.0034240229997522,
280
- "bytes": 6082079,
281
- "pages": 28,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.0907009550001021,
286
- "bytes": 7267703,
287
- "pages": 37,
288
- "shards": 3
289
- },
290
- {
291
- "seconds": 1.059233770999981,
292
- "bytes": 6001877,
293
- "pages": 28,
294
- "shards": 2
295
- },
296
- {
297
- "seconds": 1.2980492899996534,
298
- "bytes": 11801808,
299
- "pages": 36,
300
- "shards": 3
301
- },
302
- {
303
- "seconds": 1.0429953859998022,
304
- "bytes": 2436103,
305
- "pages": 28,
306
- "shards": 2
307
- },
308
- {
309
- "seconds": 1.247582260999934,
310
- "bytes": 10464030,
311
- "pages": 39,
312
- "shards": 3
313
- },
314
- {
315
- "seconds": 1.128213264000351,
316
- "bytes": 2552181,
317
- "pages": 37,
318
- "shards": 2
319
- },
320
  {
321
  "seconds": 1.4455456679997951,
322
  "bytes": 5746642,
@@ -460,6 +322,246 @@
460
  "bytes": 875600,
461
  "pages": 26,
462
  "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
463
  }
464
  ],
465
  "cache_key": {
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 76847.5007153862,
26
+ "required_pages_per_second_upper_estimate": 0.005316488410258506
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9605.937589423274,
40
+ "required_pages_per_second_upper_estimate": 0.004021805473904264
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19211.87517884655,
54
+ "required_pages_per_second_upper_estimate": 0.004001326948534451
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28817.812768269825,
68
+ "required_pages_per_second_upper_estimate": 0.0017275838732054918
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28817.812768269825,
82
+ "required_pages_per_second_upper_estimate": 0.0022404581865706156
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
+ "required_token_positions_per_second": 28817.812768269825,
96
+ "required_pages_per_second_upper_estimate": 0.36964396003475875
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.11640952138283467,
162
+ "producer_p95_seconds": 0.20131237399982638,
163
+ "gpu_update_seconds": 0.9594066080699275,
164
+ "producer_batches_per_second": 8.590362610557529,
165
+ "gpu_batches_per_second": 1.042310936352352,
166
+ "producer_headroom": 8.241650654285726,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.442616308405434,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 81,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 310821.5374057803,
174
+ "upload_bytes_per_second": 5657201.142948646,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 66.69506637220437,
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": 1.4455456679997951,
184
  "bytes": 5746642,
 
322
  "bytes": 875600,
323
  "pages": 26,
324
  "shards": 1
325
+ },
326
+ {
327
+ "seconds": 0.8441450039999836,
328
+ "bytes": 578644,
329
+ "pages": 17,
330
+ "shards": 1
331
+ },
332
+ {
333
+ "seconds": 1.364240728000368,
334
+ "bytes": 10856561,
335
+ "pages": 35,
336
+ "shards": 3
337
+ },
338
+ {
339
+ "seconds": 1.2959960469997895,
340
+ "bytes": 13446052,
341
+ "pages": 36,
342
+ "shards": 4
343
+ },
344
+ {
345
+ "seconds": 0.8787024550001661,
346
+ "bytes": 837425,
347
+ "pages": 26,
348
+ "shards": 1
349
+ },
350
+ {
351
+ "seconds": 1.2931989250000697,
352
+ "bytes": 7658868,
353
+ "pages": 39,
354
+ "shards": 3
355
+ },
356
+ {
357
+ "seconds": 1.2650147490003292,
358
+ "bytes": 6116564,
359
+ "pages": 36,
360
+ "shards": 2
361
+ },
362
+ {
363
+ "seconds": 1.0610393269998895,
364
+ "bytes": 5051210,
365
+ "pages": 35,
366
+ "shards": 2
367
+ },
368
+ {
369
+ "seconds": 1.4254050959998494,
370
+ "bytes": 7441828,
371
+ "pages": 29,
372
+ "shards": 3
373
+ },
374
+ {
375
+ "seconds": 1.1509858810004516,
376
+ "bytes": 6964137,
377
+ "pages": 33,
378
+ "shards": 2
379
+ },
380
+ {
381
+ "seconds": 1.1934967530005451,
382
+ "bytes": 5866491,
383
+ "pages": 30,
384
+ "shards": 3
385
+ },
386
+ {
387
+ "seconds": 1.2554875369996807,
388
+ "bytes": 11949834,
389
+ "pages": 44,
390
+ "shards": 4
391
+ },
392
+ {
393
+ "seconds": 0.8095207870001104,
394
+ "bytes": 1155021,
395
+ "pages": 34,
396
+ "shards": 1
397
+ },
398
+ {
399
+ "seconds": 1.4986395109999648,
400
+ "bytes": 7663876,
401
+ "pages": 32,
402
+ "shards": 3
403
+ },
404
+ {
405
+ "seconds": 0.9853112940008941,
406
+ "bytes": 5162186,
407
+ "pages": 34,
408
+ "shards": 2
409
+ },
410
+ {
411
+ "seconds": 1.2030387599997994,
412
+ "bytes": 11763818,
413
+ "pages": 37,
414
+ "shards": 3
415
+ },
416
+ {
417
+ "seconds": 1.0410187759998735,
418
+ "bytes": 7475284,
419
+ "pages": 38,
420
+ "shards": 3
421
+ },
422
+ {
423
+ "seconds": 1.175038504999975,
424
+ "bytes": 877584,
425
+ "pages": 26,
426
+ "shards": 1
427
+ },
428
+ {
429
+ "seconds": 1.2262302330000239,
430
+ "bytes": 12002858,
431
+ "pages": 41,
432
+ "shards": 4
433
+ },
434
+ {
435
+ "seconds": 0.976691177000248,
436
+ "bytes": 6244802,
437
+ "pages": 36,
438
+ "shards": 2
439
+ },
440
+ {
441
+ "seconds": 1.5793915419999394,
442
+ "bytes": 781594,
443
+ "pages": 26,
444
+ "shards": 1
445
+ },
446
+ {
447
+ "seconds": 1.4397608880008193,
448
+ "bytes": 13440043,
449
+ "pages": 39,
450
+ "shards": 4
451
+ },
452
+ {
453
+ "seconds": 1.215214173998902,
454
+ "bytes": 10529948,
455
+ "pages": 39,
456
+ "shards": 3
457
+ },
458
+ {
459
+ "seconds": 1.3451746110004024,
460
+ "bytes": 2254637,
461
+ "pages": 28,
462
+ "shards": 2
463
+ },
464
+ {
465
+ "seconds": 1.1764542460005032,
466
+ "bytes": 6191157,
467
+ "pages": 36,
468
+ "shards": 2
469
+ },
470
+ {
471
+ "seconds": 1.0847412589992018,
472
+ "bytes": 1126127,
473
+ "pages": 34,
474
+ "shards": 1
475
+ },
476
+ {
477
+ "seconds": 1.2304532270009076,
478
+ "bytes": 11321849,
479
+ "pages": 33,
480
+ "shards": 4
481
+ },
482
+ {
483
+ "seconds": 1.4404141169998184,
484
+ "bytes": 12120623,
485
+ "pages": 39,
486
+ "shards": 3
487
+ },
488
+ {
489
+ "seconds": 0.8672846369991021,
490
+ "bytes": 2691618,
491
+ "pages": 37,
492
+ "shards": 2
493
+ },
494
+ {
495
+ "seconds": 0.9980244999987917,
496
+ "bytes": 5788373,
497
+ "pages": 22,
498
+ "shards": 2
499
+ },
500
+ {
501
+ "seconds": 3.3476856620000035,
502
+ "bytes": 5904691,
503
+ "pages": 43,
504
+ "shards": 2
505
+ },
506
+ {
507
+ "seconds": 1.2678156700003456,
508
+ "bytes": 7637773,
509
+ "pages": 38,
510
+ "shards": 3
511
+ },
512
+ {
513
+ "seconds": 1.1486175610007194,
514
+ "bytes": 5966085,
515
+ "pages": 28,
516
+ "shards": 2
517
+ },
518
+ {
519
+ "seconds": 1.0020495059998211,
520
+ "bytes": 6673858,
521
+ "pages": 32,
522
+ "shards": 2
523
+ },
524
+ {
525
+ "seconds": 1.1948771370007307,
526
+ "bytes": 12132833,
527
+ "pages": 38,
528
+ "shards": 4
529
+ },
530
+ {
531
+ "seconds": 1.2424271460004093,
532
+ "bytes": 1081337,
533
+ "pages": 34,
534
+ "shards": 1
535
+ },
536
+ {
537
+ "seconds": 1.3408090630000515,
538
+ "bytes": 7658166,
539
+ "pages": 39,
540
+ "shards": 3
541
+ },
542
+ {
543
+ "seconds": 0.8867853440006002,
544
+ "bytes": 5785267,
545
+ "pages": 27,
546
+ "shards": 2
547
+ },
548
+ {
549
+ "seconds": 1.081288327000948,
550
+ "bytes": 5787210,
551
+ "pages": 36,
552
+ "shards": 2
553
+ },
554
+ {
555
+ "seconds": 1.1759565290012688,
556
+ "bytes": 13132586,
557
+ "pages": 31,
558
+ "shards": 4
559
+ },
560
+ {
561
+ "seconds": 1.004212905998429,
562
+ "bytes": 6157452,
563
+ "pages": 38,
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 sha256:478e3feddd5e715e333af4d01c7c276cb17b2faf46c01877e15a16c415e9bb60
3
- size 721778367
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a7925c71cd513cd3d5f53037b00418d6c7a7f8c7b424e1f49846780b2108ebe
3
+ size 721771711