appvoid commited on
Commit
1b41adc
·
verified ·
1 Parent(s): b66c70e

Qwen90M step 480000; 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 +246 -246
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 475000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368",
12
  "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
- "shard_scheduler.json": "c7547122b453ecc191b4f5a85a0fcc7e7fd9861a4ac8d453704e064cb85f7f50",
15
- "coverage.json": "e1b19a902793bcbca1bc3b1a6316d7ddff0a810806697b5e85713b565d8e41a4",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 480000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d",
12
  "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905",
13
  "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb",
14
+ "shard_scheduler.json": "99dff589984cd9ecd0f8113aa15c684b2d16a8376d52ae86de45ab20d0a7c84e",
15
+ "coverage.json": "15b555705988c6d4ac287412f0fe38ff9364ae15871a90a339f590525bb34a89",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 475000,
3
  "stats": {
4
- "climbmix/blocks": 17100000,
5
- "climbmix/targets": 28583228062,
6
- "climbmix/input_tokens": 28585039176,
7
- "ultra_qa/blocks": 6412500,
8
- "ultra_qa/targets": 10460856297,
9
- "ultra_qa/input_tokens": 10460876436,
10
- "cortex/blocks": 6412500,
11
- "cortex/targets": 243889844,
12
- "cortex/input_tokens": 10193624611,
13
- "ultra_style/blocks": 6412500,
14
- "ultra_style/targets": 10143291330,
15
- "ultra_style/input_tokens": 10143822186,
16
- "openmath/blocks": 4275000,
17
- "openmath/targets": 6605793668,
18
- "openmath/input_tokens": 6607994980,
19
- "rewrite6/blocks": 2137500,
20
- "rewrite6/targets": 3504058776,
21
- "rewrite6/input_tokens": 3507900626
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2342,
26
- "row": 1120,
27
  "start": {
28
- "file": 113,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 28,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 9593952
38
  },
39
  "rewrite6": {
40
- "page": 32,
41
- "row": 1761,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 131072
47
  }
48
  },
49
  "epoch": 26,
50
  "offset": 2048,
51
- "documents": 5473337
52
  },
53
  "openmath": {
54
- "page": 1463,
55
- "row": 3070,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 315,
61
- "row": 998
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 5995518
67
  },
68
  "ultra_qa": {
69
- "page": 632,
70
- "row": 3515,
71
  "start": {
72
- "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 7,
76
- "row": 57974
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 2592187
82
  },
83
  "ultra_style": {
84
- "page": 965,
85
- "row": 1470,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
- "row": 18773
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 4096,
96
- "documents": 3954110
97
  },
98
  "cortex": {
99
- "page": 124371,
100
- "row": 95,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 15919583
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 480000,
3
  "stats": {
4
+ "climbmix/blocks": 17280000,
5
+ "climbmix/targets": 28884017959,
6
+ "climbmix/input_tokens": 28885848126,
7
+ "ultra_qa/blocks": 6480000,
8
+ "ultra_qa/targets": 10570973360,
9
+ "ultra_qa/input_tokens": 10570993707,
10
+ "cortex/blocks": 6480000,
11
+ "cortex/targets": 246461343,
12
+ "cortex/input_tokens": 10301013526,
13
+ "ultra_style/blocks": 6480000,
14
+ "ultra_style/targets": 10249752091,
15
+ "ultra_style/input_tokens": 10250288878,
16
+ "openmath/blocks": 4320000,
17
+ "openmath/targets": 6675226184,
18
+ "openmath/input_tokens": 6677450412,
19
+ "rewrite6/blocks": 2160000,
20
+ "rewrite6/targets": 3541074887,
21
+ "rewrite6/input_tokens": 3544957180
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2366,
26
+ "row": 3875,
27
  "start": {
28
+ "file": 114,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 41,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 9695011
38
  },
39
  "rewrite6": {
40
+ "page": 46,
41
+ "row": 1940,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 188416
47
  }
48
  },
49
  "epoch": 26,
50
  "offset": 2048,
51
+ "documents": 5530860
52
  },
53
  "openmath": {
54
+ "page": 1479,
55
+ "row": 475,
56
  "start": {
57
  "file": 13,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 381,
61
+ "row": 534
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 6058459
67
  },
68
  "ultra_qa": {
69
+ "page": 639,
70
+ "row": 2127,
71
  "start": {
72
+ "file": 5,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 0,
76
+ "row": 19215
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 2619471
82
  },
83
  "ultra_style": {
84
+ "page": 975,
85
+ "row": 2846,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 6,
91
+ "row": 59733
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 4096,
96
+ "documents": 3996446
97
  },
98
  "cortex": {
99
+ "page": 125681,
100
+ "row": 34,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 16087202
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 475000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "c7547122b453ecc191b4f5a85a0fcc7e7fd9861a4ac8d453704e064cb85f7f50", "coverage.json": "e1b19a902793bcbca1bc3b1a6316d7ddff0a810806697b5e85713b565d8e41a4", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000475000"}
 
1
+ {"step": 480000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d", "training_config.json": "42b690efbefcf434fbcb690b27d17a2af5e3b65c3c6aab71efa1d83d1c6a6905", "performance.json": "58f8b6f79b191aa11b2bedb4594ca2897b6fdad2419ea7c9210688ac8022e7cb", "shard_scheduler.json": "99dff589984cd9ecd0f8113aa15c684b2d16a8376d52ae86de45ab20d0a7c84e", "coverage.json": "15b555705988c6d4ac287412f0fe38ff9364ae15871a90a339f590525bb34a89", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000480000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:703d2814f0ec23aea4d4ff5f43b6f3db780c5bd0ae33a5dfe00319ba58e59a63
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e59bfaa3ad90f7bdfba61500283a1881f70fb7357e98f55e965e5c5450a7199a
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": 75488.53687979521,
26
- "required_pages_per_second_upper_estimate": 0.005222472138881793
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": 9436.067109974401,
40
- "required_pages_per_second_upper_estimate": 0.003950684251458008
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": 18872.134219948803,
54
- "required_pages_per_second_upper_estimate": 0.003930567866367642
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": 28308.201329923202,
68
- "required_pages_per_second_upper_estimate": 0.001697033445608223
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": 28308.201329923202,
82
- "required_pages_per_second_upper_estimate": 0.002200838138782989
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": 22,
95
- "required_token_positions_per_second": 28308.201329923202,
96
- "required_pages_per_second_upper_estimate": 0.3631072116817794
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 704566,
161
- "producer_batch_seconds": 0.1093707058749942,
162
- "producer_p95_seconds": 0.32194404499750817,
163
- "gpu_update_seconds": 0.9766780897793977,
164
- "producer_batches_per_second": 9.143216110746831,
165
- "gpu_batches_per_second": 1.0238788096760418,
166
- "producer_headroom": 8.92997884548443,
167
  "sustainable": true,
168
- "buffer_seconds": 60.55404156632266,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 85,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 399042.23246710177,
174
- "upload_bytes_per_second": 5347906.198548627,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 85.62517385968961,
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": 1.1180898200000229,
184
- "bytes": 7687731,
185
- "pages": 38,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.1014808299987635,
190
- "bytes": 5892940,
191
- "pages": 37,
192
- "shards": 2
193
  },
194
  {
195
- "seconds": 0.8753045230005227,
196
- "bytes": 5867523,
197
- "pages": 28,
198
- "shards": 2
199
  },
200
  {
201
- "seconds": 1.8846877780015348,
202
- "bytes": 8419016,
203
- "pages": 36,
204
  "shards": 3
205
  },
206
  {
207
- "seconds": 0.9963944320006703,
208
- "bytes": 6337215,
209
- "pages": 37,
210
- "shards": 2
211
- },
212
- {
213
- "seconds": 1.0649026120008784,
214
- "bytes": 9997467,
215
- "pages": 31,
216
  "shards": 3
217
  },
218
  {
219
- "seconds": 1.0817448720008542,
220
- "bytes": 2541266,
221
- "pages": 37,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 1.4285129730014887,
226
- "bytes": 6270013,
227
  "pages": 36,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.485799860998668,
232
- "bytes": 7350955,
233
- "pages": 31,
234
- "shards": 3
235
- },
236
- {
237
- "seconds": 1.3855582350006443,
238
- "bytes": 11926342,
239
- "pages": 38,
240
- "shards": 3
241
  },
242
  {
243
- "seconds": 1.1727449809986865,
244
- "bytes": 5685962,
245
  "pages": 28,
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.2225738800007093,
250
- "bytes": 2613097,
251
- "pages": 37,
252
- "shards": 2
253
- },
254
- {
255
- "seconds": 1.0333214540005429,
256
- "bytes": 6358222,
257
- "pages": 38,
258
- "shards": 2
259
  },
260
  {
261
- "seconds": 1.2548035990002973,
262
- "bytes": 11160297,
263
- "pages": 33,
264
- "shards": 4
265
  },
266
  {
267
- "seconds": 1.0187819040002069,
268
- "bytes": 1294212,
269
- "pages": 35,
270
- "shards": 1
271
  },
272
  {
273
- "seconds": 1.2491005060001044,
274
- "bytes": 11541986,
275
- "pages": 30,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.6100285119973705,
280
- "bytes": 7404435,
281
- "pages": 39,
282
  "shards": 3
283
  },
284
  {
285
- "seconds": 1.136316104002617,
286
- "bytes": 5061136,
287
  "pages": 30,
288
  "shards": 2
289
  },
290
  {
291
- "seconds": 1.2786440780000703,
292
- "bytes": 7516874,
293
- "pages": 38,
294
  "shards": 3
295
  },
296
  {
297
- "seconds": 0.9931387870019535,
298
- "bytes": 6287159,
299
- "pages": 38,
300
- "shards": 2
301
- },
302
- {
303
- "seconds": 0.9191857780024293,
304
- "bytes": 859128,
305
- "pages": 26,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.4843631869989622,
310
- "bytes": 18084349,
311
- "pages": 33,
312
- "shards": 5
313
  },
314
  {
315
- "seconds": 0.9931742519984255,
316
- "bytes": 1121111,
317
  "pages": 35,
318
  "shards": 1
319
  },
320
  {
321
- "seconds": 1.0368349029995443,
322
- "bytes": 5950136,
323
- "pages": 29,
324
- "shards": 2
325
  },
326
  {
327
- "seconds": 0.8421170019973943,
328
- "bytes": 2576016,
329
- "pages": 37,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 0.9882992809980351,
334
- "bytes": 6015320,
335
- "pages": 28,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 1.0212832630022604,
340
- "bytes": 4591085,
341
- "pages": 28,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.4545433539969963,
346
- "bytes": 12931413,
347
- "pages": 25,
348
- "shards": 4
349
  },
350
  {
351
- "seconds": 1.2126570350010297,
352
- "bytes": 6557056,
353
- "pages": 45,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 0.8083091259977664,
358
- "bytes": 741591,
359
- "pages": 21,
360
  "shards": 1
361
  },
362
  {
363
- "seconds": 1.6113828489978914,
364
- "bytes": 12223061,
365
- "pages": 49,
366
- "shards": 4
367
  },
368
  {
369
- "seconds": 0.9428402260018629,
370
- "bytes": 5721024,
371
- "pages": 29,
372
  "shards": 2
373
  },
374
  {
375
- "seconds": 1.6995308779987681,
376
- "bytes": 2763568,
377
  "pages": 39,
378
- "shards": 2
 
 
 
 
 
 
379
  },
380
  {
381
- "seconds": 1.0716314189994591,
382
- "bytes": 5628326,
383
- "pages": 20,
384
  "shards": 2
385
  },
386
  {
387
- "seconds": 1.0639040940004634,
388
- "bytes": 11047745,
389
- "pages": 29,
390
  "shards": 3
391
  },
392
  {
393
- "seconds": 1.0850444019997667,
394
- "bytes": 7634525,
395
  "pages": 39,
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 4.141720985000575,
400
- "bytes": 6233393,
401
  "pages": 37,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 0.9377798560017254,
406
- "bytes": 884750,
407
- "pages": 27,
408
- "shards": 1
409
- },
410
- {
411
- "seconds": 1.367523049000738,
412
- "bytes": 11388768,
413
- "pages": 34,
414
- "shards": 4
415
  },
416
  {
417
- "seconds": 0.9475808069983032,
418
- "bytes": 1235313,
419
- "pages": 36,
420
  "shards": 1
421
  },
422
  {
423
- "seconds": 1.3809676959972421,
424
- "bytes": 13424902,
425
- "pages": 35,
426
  "shards": 4
427
  },
428
  {
429
- "seconds": 0.9754727400031697,
430
- "bytes": 994247,
431
- "pages": 29,
432
  "shards": 1
433
  },
434
  {
435
- "seconds": 1.2377208289981354,
436
- "bytes": 11067407,
437
- "pages": 30,
438
  "shards": 3
439
  },
440
  {
441
- "seconds": 1.5320398900003056,
442
- "bytes": 7616229,
443
- "pages": 35,
444
- "shards": 3
445
  },
446
  {
447
- "seconds": 0.994875273998332,
448
- "bytes": 1060442,
449
- "pages": 31,
450
- "shards": 1
451
  },
452
  {
453
- "seconds": 1.0037741340020148,
454
- "bytes": 6334013,
455
- "pages": 36,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 1.4788876900020114,
460
- "bytes": 12280922,
461
- "pages": 40,
462
- "shards": 4
463
  },
464
  {
465
- "seconds": 1.3189756050014694,
466
- "bytes": 5844934,
467
- "pages": 28,
468
  "shards": 2
469
  },
470
  {
471
- "seconds": 1.142625862001296,
472
- "bytes": 7700249,
473
- "pages": 41,
474
- "shards": 3
475
  },
476
  {
477
- "seconds": 0.8653922449957463,
478
- "bytes": 888116,
479
  "pages": 27,
480
  "shards": 1
481
  },
482
  {
483
- "seconds": 1.3053653089955333,
484
- "bytes": 10336652,
485
- "pages": 41,
486
  "shards": 3
487
  },
488
  {
489
- "seconds": 1.1407873469943297,
490
- "bytes": 7506775,
491
  "pages": 31,
492
  "shards": 3
493
  },
494
  {
495
- "seconds": 1.0540632260017446,
496
- "bytes": 6792818,
497
- "pages": 30,
 
 
 
 
 
 
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 0.9928698779986007,
502
- "bytes": 5905666,
503
- "pages": 30,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 2.588622392999241,
508
- "bytes": 7185173,
509
  "pages": 40,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.1344652970001334,
514
- "bytes": 5709671,
515
- "pages": 28,
516
- "shards": 2
517
- },
518
- {
519
- "seconds": 1.2757203309956822,
520
- "bytes": 7569977,
521
- "pages": 40,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 0.8296898909975425,
526
- "bytes": 1129072,
527
- "pages": 35,
528
  "shards": 1
529
  },
530
  {
531
- "seconds": 1.0205191320055746,
532
- "bytes": 9233133,
533
- "pages": 13,
534
  "shards": 3
535
  },
536
  {
537
- "seconds": 1.1241545290031354,
538
- "bytes": 8856234,
539
- "pages": 49,
540
  "shards": 3
541
  },
542
  {
543
- "seconds": 1.2785397469997406,
544
- "bytes": 5725754,
545
- "pages": 29,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.0718349349990604,
550
- "bytes": 7504941,
551
- "pages": 39,
552
  "shards": 3
553
  },
554
  {
555
- "seconds": 1.2024066430021776,
556
- "bytes": 5964235,
557
- "pages": 32,
 
 
 
 
 
 
 
 
 
 
 
 
558
  "shards": 2
559
  },
560
  {
561
- "seconds": 1.2837438990027294,
562
- "bytes": 5816020,
 
 
 
 
 
 
563
  "pages": 32,
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": 75948.0229487074,
26
+ "required_pages_per_second_upper_estimate": 0.005254260451283705
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": 9493.502868588424,
40
+ "required_pages_per_second_upper_estimate": 0.003974731404194673
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": 18987.00573717685,
54
+ "required_pages_per_second_upper_estimate": 0.003954492574040618
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": 28480.508605765273,
68
+ "required_pages_per_second_upper_estimate": 0.0017073630036969789
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": 28480.508605765273,
82
+ "required_pages_per_second_upper_estimate": 0.002214234271580101
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": 22,
95
+ "required_token_positions_per_second": 28480.508605765273,
96
+ "required_pages_per_second_upper_estimate": 0.36531738440714295
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": 704566,
161
+ "producer_batch_seconds": 0.14875617208599579,
162
+ "producer_p95_seconds": 0.35381016099563567,
163
+ "gpu_update_seconds": 0.970769180519594,
164
+ "producer_batches_per_second": 6.722410142564714,
165
+ "gpu_batches_per_second": 1.030110988345098,
166
+ "producer_headroom": 6.525908585214155,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.18768919221483,
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": 352449.9817351,
174
+ "upload_bytes_per_second": 5306492.160540054,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 75.62756146468877,
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": 1.3809676959972421,
184
+ "bytes": 13424902,
185
+ "pages": 35,
186
+ "shards": 4
 
 
 
 
 
 
187
  },
188
  {
189
+ "seconds": 0.9754727400031697,
190
+ "bytes": 994247,
191
+ "pages": 29,
192
+ "shards": 1
193
  },
194
  {
195
+ "seconds": 1.2377208289981354,
196
+ "bytes": 11067407,
197
+ "pages": 30,
198
  "shards": 3
199
  },
200
  {
201
+ "seconds": 1.5320398900003056,
202
+ "bytes": 7616229,
203
+ "pages": 35,
 
 
 
 
 
 
204
  "shards": 3
205
  },
206
  {
207
+ "seconds": 0.994875273998332,
208
+ "bytes": 1060442,
209
+ "pages": 31,
210
+ "shards": 1
211
  },
212
  {
213
+ "seconds": 1.0037741340020148,
214
+ "bytes": 6334013,
215
  "pages": 36,
216
  "shards": 2
217
  },
218
  {
219
+ "seconds": 1.4788876900020114,
220
+ "bytes": 12280922,
221
+ "pages": 40,
222
+ "shards": 4
 
 
 
 
 
 
223
  },
224
  {
225
+ "seconds": 1.3189756050014694,
226
+ "bytes": 5844934,
227
  "pages": 28,
228
  "shards": 2
229
  },
230
  {
231
+ "seconds": 1.142625862001296,
232
+ "bytes": 7700249,
233
+ "pages": 41,
234
+ "shards": 3
 
 
 
 
 
 
235
  },
236
  {
237
+ "seconds": 0.8653922449957463,
238
+ "bytes": 888116,
239
+ "pages": 27,
240
+ "shards": 1
241
  },
242
  {
243
+ "seconds": 1.3053653089955333,
244
+ "bytes": 10336652,
245
+ "pages": 41,
246
+ "shards": 3
247
  },
248
  {
249
+ "seconds": 1.1407873469943297,
250
+ "bytes": 7506775,
251
+ "pages": 31,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 1.0540632260017446,
256
+ "bytes": 6792818,
257
+ "pages": 30,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 0.9928698779986007,
262
+ "bytes": 5905666,
263
  "pages": 30,
264
  "shards": 2
265
  },
266
  {
267
+ "seconds": 2.588622392999241,
268
+ "bytes": 7185173,
269
+ "pages": 40,
270
  "shards": 3
271
  },
272
  {
273
+ "seconds": 1.1344652970001334,
274
+ "bytes": 5709671,
275
+ "pages": 28,
 
 
 
 
 
 
276
  "shards": 2
277
  },
278
  {
279
+ "seconds": 1.2757203309956822,
280
+ "bytes": 7569977,
281
+ "pages": 40,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 0.8296898909975425,
286
+ "bytes": 1129072,
287
  "pages": 35,
288
  "shards": 1
289
  },
290
  {
291
+ "seconds": 1.0205191320055746,
292
+ "bytes": 9233133,
293
+ "pages": 13,
294
+ "shards": 3
295
  },
296
  {
297
+ "seconds": 1.1241545290031354,
298
+ "bytes": 8856234,
299
+ "pages": 49,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 1.2785397469997406,
304
+ "bytes": 5725754,
305
+ "pages": 29,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.0718349349990604,
310
+ "bytes": 7504941,
311
+ "pages": 39,
312
+ "shards": 3
313
  },
314
  {
315
+ "seconds": 1.2024066430021776,
316
+ "bytes": 5964235,
317
+ "pages": 32,
318
+ "shards": 2
319
  },
320
  {
321
+ "seconds": 1.2837438990027294,
322
+ "bytes": 5816020,
323
+ "pages": 32,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.0380413320017396,
328
+ "bytes": 232525,
329
+ "pages": 6,
330
  "shards": 1
331
  },
332
  {
333
+ "seconds": 1.743612624995876,
334
+ "bytes": 7907888,
335
+ "pages": 23,
336
+ "shards": 3
337
  },
338
  {
339
+ "seconds": 2.2645728169954964,
340
+ "bytes": 6235357,
341
+ "pages": 44,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.3341411860019434,
346
+ "bytes": 11745696,
347
  "pages": 39,
348
+ "shards": 4
349
+ },
350
+ {
351
+ "seconds": 3.1945914269963396,
352
+ "bytes": 906980,
353
+ "pages": 27,
354
+ "shards": 1
355
  },
356
  {
357
+ "seconds": 1.100945904996479,
358
+ "bytes": 6366458,
359
+ "pages": 37,
360
  "shards": 2
361
  },
362
  {
363
+ "seconds": 1.1544355079968227,
364
+ "bytes": 7474922,
365
+ "pages": 32,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.0820681480036,
370
+ "bytes": 10987528,
371
  "pages": 39,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.2853161590028321,
376
+ "bytes": 6397328,
377
  "pages": 37,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.1332088359995396,
382
+ "bytes": 7397305,
383
+ "pages": 32,
384
+ "shards": 3
 
 
 
 
 
 
385
  },
386
  {
387
+ "seconds": 0.7730173130039475,
388
+ "bytes": 1053057,
389
+ "pages": 32,
390
  "shards": 1
391
  },
392
  {
393
+ "seconds": 1.3466375830030302,
394
+ "bytes": 11763121,
395
+ "pages": 37,
396
  "shards": 4
397
  },
398
  {
399
+ "seconds": 1.4802930930018192,
400
+ "bytes": 1138429,
401
+ "pages": 32,
402
  "shards": 1
403
  },
404
  {
405
+ "seconds": 1.8897758500024793,
406
+ "bytes": 11833830,
407
+ "pages": 33,
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.1506007540010614,
412
+ "bytes": 2440099,
413
+ "pages": 31,
414
+ "shards": 2
415
  },
416
  {
417
+ "seconds": 1.5953800970019074,
418
+ "bytes": 10985431,
419
+ "pages": 43,
420
+ "shards": 3
421
  },
422
  {
423
+ "seconds": 1.5438401100036572,
424
+ "bytes": 7453011,
425
+ "pages": 31,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.0392796939995605,
430
+ "bytes": 1282728,
431
+ "pages": 37,
432
+ "shards": 1
433
  },
434
  {
435
+ "seconds": 1.3358671570022125,
436
+ "bytes": 5881131,
437
+ "pages": 29,
438
  "shards": 2
439
  },
440
  {
441
+ "seconds": 1.5660001680007554,
442
+ "bytes": 16980575,
443
+ "pages": 44,
444
+ "shards": 6
445
  },
446
  {
447
+ "seconds": 0.8579309720007586,
448
+ "bytes": 927462,
449
  "pages": 27,
450
  "shards": 1
451
  },
452
  {
453
+ "seconds": 1.3555858160034404,
454
+ "bytes": 7535281,
455
+ "pages": 40,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 1.1513379759999225,
460
+ "bytes": 10862357,
461
  "pages": 31,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 0.9964048779947916,
466
+ "bytes": 1334873,
467
+ "pages": 36,
468
+ "shards": 1
469
+ },
470
+ {
471
+ "seconds": 1.2506730089953635,
472
+ "bytes": 7336892,
473
+ "pages": 32,
474
  "shards": 3
475
  },
476
  {
477
+ "seconds": 1.3737510540013318,
478
+ "bytes": 6860686,
479
+ "pages": 33,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.2590301210002508,
484
+ "bytes": 10720957,
485
  "pages": 40,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.187173960999644,
490
+ "bytes": 7325771,
491
+ "pages": 31,
 
 
 
 
 
 
492
  "shards": 3
493
  },
494
  {
495
+ "seconds": 0.8256457960014814,
496
+ "bytes": 1213741,
497
+ "pages": 36,
498
  "shards": 1
499
  },
500
  {
501
+ "seconds": 1.0744473489976372,
502
+ "bytes": 7318826,
503
+ "pages": 31,
504
  "shards": 3
505
  },
506
  {
507
+ "seconds": 1.3007696439963183,
508
+ "bytes": 9895316,
509
+ "pages": 41,
510
  "shards": 3
511
  },
512
  {
513
+ "seconds": 1.3112097839984926,
514
+ "bytes": 6821672,
515
+ "pages": 34,
516
  "shards": 2
517
  },
518
  {
519
+ "seconds": 1.0360545559960883,
520
+ "bytes": 7415725,
521
+ "pages": 31,
522
  "shards": 3
523
  },
524
  {
525
+ "seconds": 0.9522623859957093,
526
+ "bytes": 963026,
527
+ "pages": 30,
528
+ "shards": 1
529
+ },
530
+ {
531
+ "seconds": 1.1533947639982216,
532
+ "bytes": 11916585,
533
+ "pages": 41,
534
+ "shards": 4
535
+ },
536
+ {
537
+ "seconds": 1.1312003819984966,
538
+ "bytes": 5813649,
539
+ "pages": 29,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 1.0730638719978742,
544
+ "bytes": 1187099,
545
+ "pages": 36,
546
+ "shards": 1
547
+ },
548
+ {
549
+ "seconds": 1.1143769740010612,
550
+ "bytes": 13033863,
551
  "pages": 32,
552
+ "shards": 4
553
+ },
554
+ {
555
+ "seconds": 1.1261608329950832,
556
+ "bytes": 10613783,
557
+ "pages": 39,
558
+ "shards": 3
559
+ },
560
+ {
561
+ "seconds": 1.0124226469997666,
562
+ "bytes": 904800,
563
+ "pages": 27,
564
+ "shards": 1
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:74993c29167e02b4a918d7349a70338c13f23664ef7d0636060fee4da7f81368
3
- size 721768703
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db9a911bd1f9aa38ba6720dbd8e0a226e9b574ebcb1cd5814f64817fae1e107d
3
+ size 721770047