appvoid commited on
Commit
ba9c76f
·
verified ·
1 Parent(s): 3f6b08c

Qwen90M step 550000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +47 -47
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +249 -249
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 545000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7",
12
  "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb",
13
  "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
- "shard_scheduler.json": "9362a296b2832b4b5007ca3ec1a25ba07f94fdb1146f86932cdf2269d63a6f4c",
15
- "coverage.json": "682a00d4779d7723fea633cd98c833ad848bbb3df21fa445fa4095cd4ab9203f",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 550000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead",
12
  "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb",
13
  "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af",
14
+ "shard_scheduler.json": "0714cd433019a43dac02d272ee7a38fe3cc84979e30ed3a971db3c2b11c513a0",
15
+ "coverage.json": "e56a23246a1ce8c2d499577bf18fa4049c8262287fd5f510b2a1aa7259f1d255",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 545000,
3
  "stats": {
4
- "climbmix/blocks": 19620000,
5
- "climbmix/targets": 32797131226,
6
- "climbmix/input_tokens": 32799209443,
7
- "ultra_qa/blocks": 7357500,
8
- "ultra_qa/targets": 12002847915,
9
- "ultra_qa/input_tokens": 12002871095,
10
- "cortex/blocks": 7357500,
11
- "cortex/targets": 279804176,
12
- "cortex/input_tokens": 11695457470,
13
- "ultra_style/blocks": 7357500,
14
- "ultra_style/targets": 11636722703,
15
- "ultra_style/input_tokens": 11637332832,
16
- "openmath/blocks": 4905000,
17
- "openmath/targets": 7578975444,
18
- "openmath/input_tokens": 7581499711,
19
- "rewrite6/blocks": 2452500,
20
- "rewrite6/targets": 4020473427,
21
- "rewrite6/input_tokens": 4024881624
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2687,
26
- "row": 1151,
27
  "start": {
28
- "file": 130,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 3,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 11007103
38
  },
39
  "rewrite6": {
40
- "page": 28,
41
- "row": 3287,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 114688
47
  }
48
  },
49
  "epoch": 30,
50
  "offset": 0,
51
- "documents": 6280095
52
  },
53
  "openmath": {
54
- "page": 1679,
55
- "row": 155,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 327,
61
- "row": 434
62
  }
63
  },
64
  "epoch": 0,
65
- "offset": 0,
66
- "documents": 6877339
67
  },
68
  "ultra_qa": {
69
- "page": 726,
70
- "row": 497,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 5,
76
- "row": 53043
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
- "documents": 2974193
82
  },
83
  "ultra_style": {
84
- "page": 1108,
85
- "row": 980,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
- "row": 2462
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 2048,
96
- "documents": 4539348
97
  },
98
  "cortex": {
99
- "page": 142692,
100
- "row": 98,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 18264674
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 550000,
3
  "stats": {
4
+ "climbmix/blocks": 19800000,
5
+ "climbmix/targets": 33097835451,
6
+ "climbmix/input_tokens": 33099932886,
7
+ "ultra_qa/blocks": 7425000,
8
+ "ultra_qa/targets": 12112854779,
9
+ "ultra_qa/input_tokens": 12112878132,
10
+ "cortex/blocks": 7425000,
11
+ "cortex/targets": 282367717,
12
+ "cortex/input_tokens": 11802740868,
13
+ "ultra_style/blocks": 7425000,
14
+ "ultra_style/targets": 11743946324,
15
+ "ultra_style/input_tokens": 11744561998,
16
+ "openmath/blocks": 4950000,
17
+ "openmath/targets": 7648567050,
18
+ "openmath/input_tokens": 7651114390,
19
+ "rewrite6/blocks": 2475000,
20
+ "rewrite6/targets": 4057422202,
21
+ "rewrite6/input_tokens": 4061870651
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2711,
26
+ "row": 4062,
27
  "start": {
28
+ "file": 131,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 16,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 11108318
38
  },
39
  "rewrite6": {
40
+ "page": 42,
41
+ "row": 3348,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 172032
47
  }
48
  },
49
  "epoch": 30,
50
  "offset": 0,
51
+ "documents": 6337500
52
  },
53
  "openmath": {
54
+ "page": 1694,
55
+ "row": 1687,
56
  "start": {
57
  "file": 15,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 388,
61
+ "row": 874
62
  }
63
  },
64
  "epoch": 0,
65
+ "offset": 2048,
66
+ "documents": 6940311
67
  },
68
  "ultra_qa": {
69
+ "page": 732,
70
+ "row": 3118,
71
  "start": {
72
  "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 6,
76
+ "row": 13034
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 4096,
81
+ "documents": 3001390
82
  },
83
  "ultra_style": {
84
+ "page": 1118,
85
+ "row": 1245,
86
  "start": {
87
  "file": 6,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 5,
91
+ "row": 43422
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 4096,
96
+ "documents": 4580573
97
  },
98
  "cortex": {
99
+ "page": 144001,
100
+ "row": 84,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 18432212
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 545000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7", "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "9362a296b2832b4b5007ca3ec1a25ba07f94fdb1146f86932cdf2269d63a6f4c", "coverage.json": "682a00d4779d7723fea633cd98c833ad848bbb3df21fa445fa4095cd4ab9203f", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000545000"}
 
1
+ {"step": 550000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead", "training_config.json": "eb51b7316ffa2242848fb50069dd8c6e0ecf28534fa6202f12a9fcfdcf6b16eb", "performance.json": "3070ebbdf9b2b0fd6a0a237819043aa2c2a35c74a97edfa223e5c324471af4af", "shard_scheduler.json": "0714cd433019a43dac02d272ee7a38fe3cc84979e30ed3a971db3c2b11c513a0", "coverage.json": "e56a23246a1ce8c2d499577bf18fa4049c8262287fd5f510b2a1aa7259f1d255", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000550000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a7beda1249d2d8fd1d0789e4fbf74cf1401e6a10abd39cdf9f1651fdfbec6e4b
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2004ee887235fd3517efbc4fa42c721dc73f7135207f4c6dd87736e218e909a7
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": 77724.66568028349,
26
- "required_pages_per_second_upper_estimate": 0.00537717271783323
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": 9715.583210035436,
40
- "required_pages_per_second_upper_estimate": 0.004067711805593652
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": 19431.166420070873,
54
- "required_pages_per_second_upper_estimate": 0.00404699953098255
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": 29146.74963010631,
68
- "required_pages_per_second_upper_estimate": 0.0017473031358150829
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": 29146.74963010631,
82
- "required_pages_per_second_upper_estimate": 0.002266031580738054
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": 29146.74963010631,
96
- "required_pages_per_second_upper_estimate": 0.37386320891351205
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.10329342065628566,
162
- "producer_p95_seconds": 0.2151632269997208,
163
- "gpu_update_seconds": 0.9485791846731952,
164
- "producer_batches_per_second": 9.681158718981271,
165
- "gpu_batches_per_second": 1.0542082476166923,
166
- "producer_headroom": 9.18334564434305,
167
  "sustainable": true,
168
- "buffer_seconds": 60.709067819084495,
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": 297407.2945755357,
174
- "upload_bytes_per_second": 5461795.829792473,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 63.81668212842515,
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": 1.1747847249989718,
184
- "bytes": 10716254,
185
- "pages": 34,
 
 
 
 
 
 
186
  "shards": 3
187
  },
188
  {
189
- "seconds": 1.0711626960001013,
190
- "bytes": 2631436,
191
- "pages": 37,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 1.137612402000741,
196
- "bytes": 5722821,
197
- "pages": 28,
198
  "shards": 2
199
  },
200
  {
201
- "seconds": 1.4303736129986646,
202
- "bytes": 11919616,
203
- "pages": 42,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 1.0499967649993778,
208
- "bytes": 7310677,
209
- "pages": 35,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.0670534790006059,
214
- "bytes": 5802765,
215
- "pages": 28,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 0.9111343770000531,
220
- "bytes": 2649643,
221
- "pages": 36,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 1.2196053769985156,
226
- "bytes": 6057143,
227
- "pages": 37,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.2347321990000637,
232
- "bytes": 9643472,
233
- "pages": 35,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.1648727040010272,
238
- "bytes": 8469052,
239
  "pages": 35,
240
- "shards": 3
241
- },
242
- {
243
- "seconds": 1.1223673130007228,
244
- "bytes": 6068911,
245
- "pages": 36,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.1270621679996111,
250
- "bytes": 5842224,
251
  "pages": 28,
252
- "shards": 2
253
  },
254
  {
255
- "seconds": 1.3927226420000807,
256
- "bytes": 7163586,
257
- "pages": 36,
258
- "shards": 4
259
  },
260
  {
261
- "seconds": 1.230830958000297,
262
- "bytes": 6020552,
263
  "pages": 37,
264
- "shards": 2
265
- },
266
- {
267
- "seconds": 1.2729082169989852,
268
- "bytes": 7700586,
269
- "pages": 39,
270
- "shards": 3
271
  },
272
  {
273
- "seconds": 0.9891643420014589,
274
- "bytes": 6883017,
275
  "pages": 33,
276
- "shards": 2
277
  },
278
  {
279
- "seconds": 1.2779713299987634,
280
- "bytes": 9455856,
281
  "pages": 30,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.159703276000073,
286
- "bytes": 2570411,
287
- "pages": 35,
288
- "shards": 2
289
- },
290
- {
291
- "seconds": 1.2064628600001015,
292
- "bytes": 6189523,
293
- "pages": 36,
294
- "shards": 2
295
  },
296
  {
297
- "seconds": 1.4404870660000597,
298
- "bytes": 11907122,
299
- "pages": 40,
300
  "shards": 4
301
  },
302
  {
303
- "seconds": 0.842472636999446,
304
- "bytes": 857193,
305
- "pages": 26,
306
- "shards": 1
307
  },
308
  {
309
- "seconds": 1.0425618379995285,
310
- "bytes": 11991809,
311
- "pages": 35,
312
- "shards": 3
313
  },
314
  {
315
- "seconds": 1.543527111000003,
316
- "bytes": 7770797,
317
- "pages": 39,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.0592705330000172,
322
- "bytes": 5143840,
323
  "pages": 36,
324
- "shards": 2
325
  },
326
  {
327
- "seconds": 0.9495372719993611,
328
- "bytes": 5544847,
329
- "pages": 18,
330
  "shards": 2
331
  },
332
  {
333
- "seconds": 1.068828384999506,
334
- "bytes": 2455135,
335
- "pages": 29,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.2464688429990929,
340
- "bytes": 6118587,
341
- "pages": 36,
342
  "shards": 2
343
  },
344
  {
345
- "seconds": 0.8108600380001008,
346
- "bytes": 790179,
347
- "pages": 24,
348
- "shards": 1
349
- },
350
- {
351
- "seconds": 2.05223027700049,
352
- "bytes": 18642145,
353
- "pages": 45,
354
- "shards": 5
355
- },
356
- {
357
- "seconds": 1.3427877739995893,
358
- "bytes": 6044460,
359
- "pages": 36,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.2501696180006547,
364
- "bytes": 2425262,
365
  "pages": 29,
366
  "shards": 2
367
  },
368
  {
369
- "seconds": 1.180065475000447,
370
- "bytes": 6094205,
371
- "pages": 36,
372
- "shards": 2
373
  },
374
  {
375
- "seconds": 1.2719860980032536,
376
- "bytes": 10158142,
377
- "pages": 41,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 0.8902074819998234,
382
- "bytes": 2207572,
383
- "pages": 24,
384
- "shards": 2
385
  },
386
  {
387
- "seconds": 1.290630097999383,
388
- "bytes": 12177821,
389
- "pages": 42,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 3.9100393570006418,
394
- "bytes": 2234803,
395
- "pages": 29,
396
- "shards": 2
397
- },
398
- {
399
- "seconds": 0.9967152519966476,
400
- "bytes": 11168227,
401
  "pages": 38,
402
  "shards": 3
403
  },
404
  {
405
- "seconds": 1.0345216240020818,
406
- "bytes": 6241760,
407
- "pages": 36,
408
- "shards": 2
409
  },
410
  {
411
- "seconds": 1.262870265996753,
412
- "bytes": 2380360,
413
- "pages": 28,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 0.9234551270019438,
418
- "bytes": 5872826,
419
- "pages": 30,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 1.268712075001531,
424
- "bytes": 10759229,
425
- "pages": 40,
426
- "shards": 3
427
  },
428
  {
429
- "seconds": 1.2720383849991777,
430
- "bytes": 7578314,
431
  "pages": 39,
432
- "shards": 3
433
  },
434
  {
435
- "seconds": 1.001895314999274,
436
- "bytes": 6152237,
437
  "pages": 36,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 0.8757686259996262,
442
- "bytes": 2112854,
443
- "pages": 23,
444
  "shards": 3
445
  },
446
  {
447
- "seconds": 1.275124282001343,
448
- "bytes": 10806597,
449
- "pages": 46,
450
- "shards": 3
451
  },
452
  {
453
- "seconds": 0.989914507001231,
454
- "bytes": 672610,
455
- "pages": 18,
456
  "shards": 1
457
  },
458
  {
459
- "seconds": 1.4889224110011128,
460
- "bytes": 13780761,
461
- "pages": 49,
462
  "shards": 4
463
  },
464
  {
465
- "seconds": 0.993486828996538,
466
- "bytes": 5717549,
467
- "pages": 19,
468
- "shards": 2
469
  },
470
  {
471
- "seconds": 1.089157542999601,
472
- "bytes": 7603521,
473
- "pages": 46,
474
- "shards": 3
475
  },
476
  {
477
- "seconds": 1.0148978149991308,
478
- "bytes": 6571585,
479
- "pages": 35,
480
  "shards": 2
481
  },
482
  {
483
- "seconds": 1.1422953030014469,
484
- "bytes": 877918,
485
- "pages": 28,
486
- "shards": 1
487
  },
488
  {
489
- "seconds": 1.1798037480002677,
490
- "bytes": 7431767,
491
- "pages": 29,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.270509606001724,
496
- "bytes": 15692090,
497
  "pages": 37,
 
 
 
 
 
 
498
  "shards": 4
499
  },
500
  {
501
- "seconds": 1.250901432998944,
502
- "bytes": 1087271,
503
- "pages": 33,
 
 
 
 
 
 
504
  "shards": 1
505
  },
506
  {
507
- "seconds": 1.5110034520002955,
508
- "bytes": 7289248,
509
- "pages": 30,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.0708622599995579,
514
- "bytes": 1033241,
515
- "pages": 32,
 
 
 
 
 
 
516
  "shards": 1
517
  },
518
  {
519
- "seconds": 1.2405112279993773,
520
- "bytes": 12607883,
521
- "pages": 39,
522
  "shards": 4
523
  },
524
  {
525
- "seconds": 1.1709360159984499,
526
- "bytes": 6311701,
527
- "pages": 35,
528
- "shards": 2
529
  },
530
  {
531
- "seconds": 1.1361314889982168,
532
- "bytes": 6780795,
533
- "pages": 32,
534
  "shards": 2
535
  },
536
  {
537
- "seconds": 1.0505489870010933,
538
- "bytes": 7688939,
539
  "pages": 38,
540
- "shards": 3
541
  },
542
  {
543
- "seconds": 1.4415530749975005,
544
- "bytes": 9558241,
545
- "pages": 36,
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 2.3017471879975346,
550
- "bytes": 2579509,
551
- "pages": 37,
552
  "shards": 2
553
  },
554
  {
555
- "seconds": 1.0516134489989781,
556
- "bytes": 6165714,
557
- "pages": 27,
558
- "shards": 2
559
  },
560
  {
561
- "seconds": 0.908222386999114,
562
- "bytes": 5648395,
563
- "pages": 35,
564
- "shards": 2
 
 
 
 
 
 
 
 
 
 
 
 
565
  }
566
  ],
567
  "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": 77488.5880998427,
26
+ "required_pages_per_second_upper_estimate": 0.005360840323042881
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": 9686.073512480338,
40
+ "required_pages_per_second_upper_estimate": 0.004055356711459891
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": 19372.147024960675,
54
+ "required_pages_per_second_upper_estimate": 0.004034707347427211
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": 29058.220537441015,
68
+ "required_pages_per_second_upper_estimate": 0.001741995951885896
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": 29058.220537441015,
82
+ "required_pages_per_second_upper_estimate": 0.002259148833181653
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": 29058.220537441015,
96
+ "required_pages_per_second_upper_estimate": 0.37272765276793546
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
158
  "prefetch_batches": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
+ "producer_batch_seconds": 0.32388174496861666,
162
+ "producer_p95_seconds": 0.31020041200099513,
163
+ "gpu_update_seconds": 0.9514691363972556,
164
+ "producer_batches_per_second": 3.0875466602691595,
165
+ "gpu_batches_per_second": 1.051006240503509,
166
+ "producer_headroom": 2.937705354432528,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.89402472942436,
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": 458187.6926747766,
174
+ "upload_bytes_per_second": 3886870.9094786476,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 98.31641278440927,
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": 0.9234551270019438,
184
+ "bytes": 5872826,
185
+ "pages": 30,
186
+ "shards": 2
187
+ },
188
+ {
189
+ "seconds": 1.268712075001531,
190
+ "bytes": 10759229,
191
+ "pages": 40,
192
  "shards": 3
193
  },
194
  {
195
+ "seconds": 1.2720383849991777,
196
+ "bytes": 7578314,
197
+ "pages": 39,
198
+ "shards": 3
199
  },
200
  {
201
+ "seconds": 1.001895314999274,
202
+ "bytes": 6152237,
203
+ "pages": 36,
204
  "shards": 2
205
  },
206
  {
207
+ "seconds": 0.8757686259996262,
208
+ "bytes": 2112854,
209
+ "pages": 23,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 1.275124282001343,
214
+ "bytes": 10806597,
215
+ "pages": 46,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 0.989914507001231,
220
+ "bytes": 672610,
221
+ "pages": 18,
222
+ "shards": 1
223
  },
224
  {
225
+ "seconds": 1.4889224110011128,
226
+ "bytes": 13780761,
227
+ "pages": 49,
228
+ "shards": 4
229
  },
230
  {
231
+ "seconds": 0.993486828996538,
232
+ "bytes": 5717549,
233
+ "pages": 19,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.089157542999601,
238
+ "bytes": 7603521,
239
+ "pages": 46,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.0148978149991308,
244
+ "bytes": 6571585,
245
  "pages": 35,
 
 
 
 
 
 
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.1422953030014469,
250
+ "bytes": 877918,
251
  "pages": 28,
252
+ "shards": 1
253
  },
254
  {
255
+ "seconds": 1.1798037480002677,
256
+ "bytes": 7431767,
257
+ "pages": 29,
258
+ "shards": 3
259
  },
260
  {
261
+ "seconds": 1.270509606001724,
262
+ "bytes": 15692090,
263
  "pages": 37,
264
+ "shards": 4
 
 
 
 
 
 
265
  },
266
  {
267
+ "seconds": 1.250901432998944,
268
+ "bytes": 1087271,
269
  "pages": 33,
270
+ "shards": 1
271
  },
272
  {
273
+ "seconds": 1.5110034520002955,
274
+ "bytes": 7289248,
275
  "pages": 30,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.0708622599995579,
280
+ "bytes": 1033241,
281
+ "pages": 32,
282
+ "shards": 1
 
 
 
 
 
 
283
  },
284
  {
285
+ "seconds": 1.2405112279993773,
286
+ "bytes": 12607883,
287
+ "pages": 39,
288
  "shards": 4
289
  },
290
  {
291
+ "seconds": 1.1709360159984499,
292
+ "bytes": 6311701,
293
+ "pages": 35,
294
+ "shards": 2
295
  },
296
  {
297
+ "seconds": 1.1361314889982168,
298
+ "bytes": 6780795,
299
+ "pages": 32,
300
+ "shards": 2
301
  },
302
  {
303
+ "seconds": 1.0505489870010933,
304
+ "bytes": 7688939,
305
+ "pages": 38,
306
  "shards": 3
307
  },
308
  {
309
+ "seconds": 1.4415530749975005,
310
+ "bytes": 9558241,
311
  "pages": 36,
312
+ "shards": 3
313
  },
314
  {
315
+ "seconds": 2.3017471879975346,
316
+ "bytes": 2579509,
317
+ "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.0516134489989781,
322
+ "bytes": 6165714,
323
+ "pages": 27,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 0.908222386999114,
328
+ "bytes": 5648395,
329
+ "pages": 35,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.3117666090001876,
334
+ "bytes": 13264605,
335
+ "pages": 39,
336
+ "shards": 4
 
 
 
 
 
 
 
 
 
 
 
 
337
  },
338
  {
339
+ "seconds": 1.0446096289997513,
340
+ "bytes": 5930135,
341
  "pages": 29,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 0.9939750769990496,
346
+ "bytes": 1211971,
347
+ "pages": 34,
348
+ "shards": 1
349
  },
350
  {
351
+ "seconds": 1.65895906500009,
352
+ "bytes": 12493410,
353
+ "pages": 36,
354
+ "shards": 4
355
  },
356
  {
357
+ "seconds": 0.8486537320022762,
358
+ "bytes": 1121895,
359
+ "pages": 35,
360
+ "shards": 1
361
  },
362
  {
363
+ "seconds": 1.125895612000022,
364
+ "bytes": 7606699,
365
+ "pages": 33,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.2285171910007193,
370
+ "bytes": 12124317,
 
 
 
 
 
 
371
  "pages": 38,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 0.8537721579996287,
376
+ "bytes": 1128490,
377
+ "pages": 34,
378
+ "shards": 1
379
  },
380
  {
381
+ "seconds": 1.7903945020007086,
382
+ "bytes": 11075879,
383
+ "pages": 34,
384
+ "shards": 4
385
  },
386
  {
387
+ "seconds": 0.9962719150025805,
388
+ "bytes": 6075379,
389
+ "pages": 37,
390
+ "shards": 3
391
  },
392
  {
393
+ "seconds": 0.8558462190012506,
394
+ "bytes": 876425,
395
+ "pages": 25,
396
+ "shards": 1
397
  },
398
  {
399
+ "seconds": 1.5038929210022616,
400
+ "bytes": 12535029,
401
  "pages": 39,
402
+ "shards": 4
403
  },
404
  {
405
+ "seconds": 1.2274471419987094,
406
+ "bytes": 6988556,
407
  "pages": 36,
408
  "shards": 2
409
  },
410
  {
411
+ "seconds": 1.103131982999912,
412
+ "bytes": 7144691,
413
+ "pages": 29,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 0.9400397869976587,
418
+ "bytes": 6119958,
419
+ "pages": 37,
420
+ "shards": 2
421
  },
422
  {
423
+ "seconds": 0.879637284000637,
424
+ "bytes": 1034061,
425
+ "pages": 29,
426
  "shards": 1
427
  },
428
  {
429
+ "seconds": 1.2181368159981503,
430
+ "bytes": 12003212,
431
+ "pages": 41,
432
  "shards": 4
433
  },
434
  {
435
+ "seconds": 0.9120353730031638,
436
+ "bytes": 952497,
437
+ "pages": 31,
438
+ "shards": 1
439
  },
440
  {
441
+ "seconds": 1.6672031760026584,
442
+ "bytes": 13248105,
443
+ "pages": 37,
444
+ "shards": 4
445
  },
446
  {
447
+ "seconds": 1.4338108570009354,
448
+ "bytes": 5962058,
449
+ "pages": 28,
450
  "shards": 2
451
  },
452
  {
453
+ "seconds": 1.2595222179988923,
454
+ "bytes": 5231936,
455
+ "pages": 41,
456
+ "shards": 2
457
  },
458
  {
459
+ "seconds": 1.146318651000911,
460
+ "bytes": 7213324,
461
+ "pages": 30,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.1155308010020235,
466
+ "bytes": 6292434,
467
  "pages": 37,
468
+ "shards": 2
469
+ },
470
+ {
471
+ "seconds": 1.3688695919991005,
472
+ "bytes": 12940523,
473
+ "pages": 39,
474
  "shards": 4
475
  },
476
  {
477
+ "seconds": 1.308492041000136,
478
+ "bytes": 6108009,
479
+ "pages": 32,
480
+ "shards": 2
481
+ },
482
+ {
483
+ "seconds": 0.922122173000389,
484
+ "bytes": 1095644,
485
+ "pages": 31,
486
  "shards": 1
487
  },
488
  {
489
+ "seconds": 1.0743960049985617,
490
+ "bytes": 7509044,
491
+ "pages": 39,
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 1.3776889159998973,
496
+ "bytes": 10102958,
497
+ "pages": 36,
498
+ "shards": 3
499
+ },
500
+ {
501
+ "seconds": 4.997759090001637,
502
+ "bytes": 838514,
503
+ "pages": 26,
504
  "shards": 1
505
  },
506
  {
507
+ "seconds": 34.2643265609986,
508
+ "bytes": 13380891,
509
+ "pages": 44,
510
  "shards": 4
511
  },
512
  {
513
+ "seconds": 1.1768819699973392,
514
+ "bytes": 10254247,
515
+ "pages": 45,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 2.0107650110003306,
520
+ "bytes": 2402302,
521
+ "pages": 28,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.039769940001861,
526
+ "bytes": 6252309,
527
  "pages": 38,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 1.272125319999759,
532
+ "bytes": 7866146,
533
+ "pages": 39,
534
  "shards": 3
535
  },
536
  {
537
+ "seconds": 1.004612441000063,
538
+ "bytes": 5431394,
539
+ "pages": 26,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.1393026930018095,
544
+ "bytes": 12031348,
545
+ "pages": 42,
546
+ "shards": 3
547
  },
548
  {
549
+ "seconds": 1.1874672249978175,
550
+ "bytes": 7293811,
551
+ "pages": 30,
552
+ "shards": 3
553
+ },
554
+ {
555
+ "seconds": 0.9514401040032681,
556
+ "bytes": 1189191,
557
+ "pages": 36,
558
+ "shards": 1
559
+ },
560
+ {
561
+ "seconds": 1.3702037240000209,
562
+ "bytes": 10453891,
563
+ "pages": 36,
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:abf7465cb001fce37e180b53d58dac3a443134a6c6fe31d634de0b34d4e174e7
3
- size 721774207
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76e34ffa1366844a13a3350fa07375abf43216d9786a0f72de9472864bdc1ead
3
+ size 721775551