appvoid commited on
Commit
39f8db7
·
verified ·
1 Parent(s): e41af20

Qwen90M step 775000; 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 +259 -187
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 770000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d",
14
- "shard_scheduler.json": "d4f4b8859106eed29ba6de07b880ae2a733ce84e3ecc5289a12c47202ed328b5",
15
- "coverage.json": "0b5c9661fa1c47243f8665233277a741ea19c000bbb01a96a0cb91fec8b18598",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 775000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "62d59c2938f0295d3b0f8368ee314c8fc85d4acb69d47b1b11a15d84b82da2a7",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "3b0cdf7569ff3f756bba96adbbd77bced2a88ef4f965a14cbe15078a1fd95894",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d",
14
+ "shard_scheduler.json": "9a616e2bb05ee9b5cb075c6eccfe164bbcf57cf736739b2ec6ff6e8eead85217",
15
+ "coverage.json": "df852170d83f78171ae60b5d208710eaa74621666ecccf6bcd6b78262cb0b924",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 770000,
3
  "stats": {
4
- "climbmix/blocks": 27720000,
5
- "climbmix/targets": 46338195596,
6
- "climbmix/input_tokens": 46341132368,
7
- "ultra_qa/blocks": 10395000,
8
- "ultra_qa/targets": 16958238643,
9
- "ultra_qa/input_tokens": 16958271297,
10
- "cortex/blocks": 10395000,
11
- "cortex/targets": 395254793,
12
- "cortex/input_tokens": 16523352599,
13
- "ultra_style/blocks": 10395000,
14
- "ultra_style/targets": 16442103528,
15
- "ultra_style/input_tokens": 16442965847,
16
- "openmath/blocks": 6930000,
17
- "openmath/targets": 10708073683,
18
- "openmath/input_tokens": 10711641641,
19
- "rewrite6/blocks": 3465000,
20
- "rewrite6/targets": 5680408647,
21
- "rewrite6/input_tokens": 5686636248
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 3796,
26
- "row": 2700,
27
  "start": {
28
- "file": 183,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 61,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 2048,
37
- "documents": 15551116
38
  },
39
  "rewrite6": {
40
- "page": 9,
41
- "row": 2929,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 36864
47
  }
48
  },
49
  "epoch": 43,
50
  "offset": 0,
51
- "documents": 8872165
52
  },
53
  "openmath": {
54
- "page": 2372,
55
- "row": 2873,
56
  "start": {
57
  "file": 22,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 109,
61
- "row": 412
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 9718585
67
  },
68
  "ultra_qa": {
69
- "page": 1025,
70
- "row": 3695,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
- "row_group": 0,
76
- "row": 41389
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 0,
81
- "documents": 4202095
82
  },
83
  "ultra_style": {
84
- "page": 1565,
85
- "row": 3159,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 2,
91
- "row": 75510
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 6413399
97
  },
98
  "cortex": {
99
- "page": 201598,
100
- "row": 95,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 25804639
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 775000,
3
  "stats": {
4
+ "climbmix/blocks": 27900000,
5
+ "climbmix/targets": 46638920719,
6
+ "climbmix/input_tokens": 46641876738,
7
+ "ultra_qa/blocks": 10462500,
8
+ "ultra_qa/targets": 17068384083,
9
+ "ultra_qa/input_tokens": 17068416928,
10
+ "cortex/blocks": 10462500,
11
+ "cortex/targets": 397833446,
12
+ "cortex/input_tokens": 16630731446,
13
+ "ultra_style/blocks": 10462500,
14
+ "ultra_style/targets": 16549240635,
15
+ "ultra_style/input_tokens": 16550108681,
16
+ "openmath/blocks": 6975000,
17
+ "openmath/targets": 10777571114,
18
+ "openmath/input_tokens": 10781162121,
19
+ "rewrite6/blocks": 3487500,
20
+ "rewrite6/targets": 5717285134,
21
+ "rewrite6/input_tokens": 5723553756
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 3821,
26
+ "row": 1836,
27
  "start": {
28
+ "file": 184,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 78,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 0,
37
+ "documents": 15652652
38
  },
39
  "rewrite6": {
40
+ "page": 23,
41
+ "row": 3880,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 94208
47
  }
48
  },
49
  "epoch": 43,
50
  "offset": 0,
51
+ "documents": 8930460
52
  },
53
  "openmath": {
54
+ "page": 2388,
55
+ "row": 276,
56
  "start": {
57
  "file": 22,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 174,
61
+ "row": 948
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 9781524
67
  },
68
  "ultra_qa": {
69
+ "page": 1032,
70
+ "row": 2254,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
+ "row_group": 1,
76
+ "row": 5540
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 0,
81
+ "documents": 4229326
82
  },
83
  "ultra_style": {
84
+ "page": 1575,
85
+ "row": 3932,
86
  "start": {
87
  "file": 9,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 3,
91
+ "row": 31988
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 6455132
97
  },
98
  "cortex": {
99
+ "page": 202908,
100
+ "row": 41,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 25972265
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 770000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "d4f4b8859106eed29ba6de07b880ae2a733ce84e3ecc5289a12c47202ed328b5", "coverage.json": "0b5c9661fa1c47243f8665233277a741ea19c000bbb01a96a0cb91fec8b18598", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000770000"}
 
1
+ {"step": 775000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "62d59c2938f0295d3b0f8368ee314c8fc85d4acb69d47b1b11a15d84b82da2a7", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3b0cdf7569ff3f756bba96adbbd77bced2a88ef4f965a14cbe15078a1fd95894", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "9a616e2bb05ee9b5cb075c6eccfe164bbcf57cf736739b2ec6ff6e8eead85217", "coverage.json": "df852170d83f78171ae60b5d208710eaa74621666ecccf6bcd6b78262cb0b924", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000775000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62d59c2938f0295d3b0f8368ee314c8fc85d4acb69d47b1b11a15d84b82da2a7
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": 76400.702468659,
26
- "required_pages_per_second_upper_estimate": 0.005285577870835155
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": 9550.087808582375,
40
- "required_pages_per_second_upper_estimate": 0.003998422336942226
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": 19100.17561716475,
54
- "required_pages_per_second_upper_estimate": 0.003978062875551665
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": 28650.263425747125,
68
- "required_pages_per_second_upper_estimate": 0.0017175395459542889
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": 28650.263425747125,
82
- "required_pages_per_second_upper_estimate": 0.002227431962161148
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": 28650.263425747125,
96
- "required_pages_per_second_upper_estimate": 0.3674948169693453
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,338 +158,410 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.0745995468515872,
162
- "producer_p95_seconds": 0.19039384700045048,
163
- "gpu_update_seconds": 0.9650173050469608,
164
- "producer_batches_per_second": 13.40490716370516,
165
- "gpu_batches_per_second": 1.036250847285414,
166
- "producer_headroom": 12.935967385523451,
167
  "sustainable": true,
168
- "buffer_seconds": 60.79609021795853,
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": 277550.1790140697,
174
- "upload_bytes_per_second": 5458592.488142992,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 59.555807378926936,
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": 0.9713157099999989,
184
- "bytes": 23380,
185
- "pages": 1,
186
  "shards": 1
187
  },
188
  {
189
- "seconds": 1.497052147999966,
190
- "bytes": 11817101,
191
- "pages": 34,
192
- "shards": 4
193
  },
194
  {
195
- "seconds": 1.0985135669998272,
196
- "bytes": 6249223,
197
- "pages": 36,
198
- "shards": 2
199
  },
200
  {
201
- "seconds": 1.3139188940003805,
202
- "bytes": 2690437,
203
- "pages": 38,
204
- "shards": 3
205
  },
206
  {
207
- "seconds": 1.4740382569998474,
208
- "bytes": 11539917,
209
- "pages": 32,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.530777683999986,
214
- "bytes": 10014604,
215
- "pages": 34,
216
  "shards": 3
217
  },
218
  {
219
- "seconds": 1.4347790020001412,
220
- "bytes": 2652487,
221
- "pages": 37,
222
  "shards": 2
223
  },
224
  {
225
- "seconds": 1.0014089890000832,
226
- "bytes": 6005067,
 
 
 
 
 
 
227
  "pages": 36,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 0.9762548340004287,
232
- "bytes": 2290276,
233
- "pages": 28,
 
 
 
 
 
 
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.4867014489991561,
238
- "bytes": 11031744,
239
- "pages": 35,
 
 
 
 
 
 
240
  "shards": 3
241
  },
242
  {
243
- "seconds": 2.084590124000897,
244
- "bytes": 12097526,
245
  "pages": 37,
246
- "shards": 3
247
  },
248
  {
249
- "seconds": 0.8632086769994203,
250
- "bytes": 2602865,
251
- "pages": 34,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.1280111480000414,
256
- "bytes": 298657,
257
- "pages": 9,
 
 
 
 
 
 
258
  "shards": 1
259
  },
260
  {
261
- "seconds": 1.3505803890002426,
262
- "bytes": 6226670,
263
- "pages": 36,
 
 
 
 
 
 
264
  "shards": 2
265
  },
266
  {
267
- "seconds": 1.6309672779998436,
268
- "bytes": 11574403,
269
- "pages": 34,
270
  "shards": 4
271
  },
272
  {
273
- "seconds": 1.1955943939992721,
274
- "bytes": 921426,
275
- "pages": 30,
276
  "shards": 1
277
  },
278
  {
279
- "seconds": 1.0380561770007262,
280
- "bytes": 6122044,
 
 
 
 
 
 
 
 
 
 
 
 
281
  "pages": 36,
282
  "shards": 2
283
  },
284
  {
285
- "seconds": 1.4052477490004094,
286
- "bytes": 13455700,
287
- "pages": 38,
288
- "shards": 4
 
 
 
 
 
 
 
 
 
 
 
 
289
  },
290
  {
291
- "seconds": 0.9797629530003178,
292
- "bytes": 4961107,
293
  "pages": 36,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.0490742649999447,
298
- "bytes": 7239835,
299
  "pages": 30,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 1.0533643100006884,
304
- "bytes": 1118487,
305
  "pages": 34,
306
  "shards": 1
307
  },
308
  {
309
- "seconds": 1.5156246349997673,
310
- "bytes": 10630930,
311
- "pages": 34,
312
  "shards": 3
313
  },
314
  {
315
- "seconds": 1.1701987970000118,
316
- "bytes": 7151255,
317
  "pages": 34,
318
  "shards": 3
319
  },
320
  {
321
- "seconds": 1.1764370720002262,
322
- "bytes": 6881233,
323
- "pages": 35,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.185791879000135,
328
- "bytes": 7588973,
329
- "pages": 38,
330
- "shards": 3
331
  },
332
  {
333
- "seconds": 0.8400346700000227,
334
- "bytes": 806070,
335
- "pages": 24,
336
  "shards": 1
337
  },
338
  {
339
- "seconds": 1.260675972000172,
340
- "bytes": 10735591,
341
- "pages": 41,
342
- "shards": 3
343
  },
344
  {
345
- "seconds": 1.1190538959999685,
346
- "bytes": 7646532,
347
- "pages": 38,
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.0387209649998113,
352
- "bytes": 1064900,
353
  "pages": 32,
354
- "shards": 1
355
- },
356
- {
357
- "seconds": 1.1239516590003404,
358
- "bytes": 11388095,
359
- "pages": 24,
360
  "shards": 3
361
  },
362
  {
363
- "seconds": 1.270762184999512,
364
- "bytes": 12213784,
365
- "pages": 48,
366
  "shards": 4
367
  },
368
  {
369
- "seconds": 1.305295600000136,
370
- "bytes": 1053111,
371
- "pages": 30,
372
  "shards": 1
373
  },
374
  {
375
- "seconds": 1.13043969499995,
376
- "bytes": 7715241,
 
 
 
 
 
 
377
  "pages": 34,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 1.1387970109999515,
382
- "bytes": 10453428,
383
- "pages": 40,
384
- "shards": 3
385
  },
386
  {
387
- "seconds": 1.3777619700003925,
388
- "bytes": 819264,
389
- "pages": 26,
390
  "shards": 2
391
  },
392
  {
393
- "seconds": 1.2043475289992784,
394
- "bytes": 13122126,
395
- "pages": 40,
396
  "shards": 4
397
  },
398
  {
399
- "seconds": 1.1971378379994349,
400
- "bytes": 5971358,
401
- "pages": 36,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 0.9979249650004931,
406
- "bytes": 6639528,
407
- "pages": 32,
408
  "shards": 3
409
  },
410
  {
411
- "seconds": 1.0738694510000641,
412
- "bytes": 6116466,
 
 
 
 
 
 
413
  "pages": 37,
414
  "shards": 2
415
  },
416
  {
417
- "seconds": 0.993601779000528,
418
- "bytes": 1093689,
419
- "pages": 34,
420
- "shards": 1
421
  },
422
  {
423
- "seconds": 1.0732885580000584,
424
- "bytes": 7199351,
425
- "pages": 29,
426
  "shards": 3
427
  },
428
  {
429
- "seconds": 1.3326483809996716,
430
- "bytes": 16444186,
431
  "pages": 37,
432
- "shards": 4
433
  },
434
  {
435
- "seconds": 0.9647154859994771,
436
- "bytes": 2573191,
437
  "pages": 37,
438
  "shards": 2
439
  },
440
  {
441
- "seconds": 1.2671719289992325,
442
- "bytes": 6070484,
443
- "pages": 36,
 
 
 
 
 
 
444
  "shards": 2
445
  },
446
  {
447
- "seconds": 0.9431054610013234,
448
- "bytes": 842654,
449
- "pages": 26,
450
- "shards": 1
451
  },
452
  {
453
- "seconds": 1.3556453979999787,
454
- "bytes": 11128099,
455
- "pages": 32,
456
- "shards": 4
457
  },
458
  {
459
- "seconds": 1.1991763389996777,
460
- "bytes": 5989028,
461
- "pages": 39,
 
 
 
 
 
 
462
  "shards": 2
463
  },
464
  {
465
- "seconds": 1.483354228999815,
466
- "bytes": 13516314,
467
- "pages": 40,
468
- "shards": 4
469
  },
470
  {
471
- "seconds": 0.855230520999612,
472
- "bytes": 838138,
473
- "pages": 29,
474
  "shards": 1
475
  },
476
  {
477
- "seconds": 1.1489447340009065,
478
- "bytes": 10403004,
479
- "pages": 36,
480
- "shards": 3
481
  },
482
  {
483
- "seconds": 0.9511843500004034,
484
- "bytes": 2374318,
485
- "pages": 29,
486
- "shards": 2
487
  },
488
  {
489
- "seconds": 1.678345239999544,
490
- "bytes": 6141170,
491
- "pages": 36,
492
- "shards": 2
493
  }
494
  ],
495
  "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": 77130.99093735527,
26
+ "required_pages_per_second_upper_estimate": 0.0053361009216022675
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": 9641.373867169408,
40
+ "required_pages_per_second_upper_estimate": 0.004036641903402943
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": 19282.747734338816,
54
+ "required_pages_per_second_upper_estimate": 0.004016087832808515
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": 28924.121601508225,
68
+ "required_pages_per_second_upper_estimate": 0.0017339569254339446
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": 28924.121601508225,
82
+ "required_pages_per_second_upper_estimate": 0.0022487232307517617
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": 28924.121601508225,
96
+ "required_pages_per_second_upper_estimate": 0.37100757560200903
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": 0.07639025929684351,
162
+ "producer_p95_seconds": 0.21046823000142467,
163
+ "gpu_update_seconds": 0.9558803679817995,
164
+ "producer_batches_per_second": 13.090674245706095,
165
+ "gpu_batches_per_second": 1.0461560185730696,
166
+ "producer_headroom": 12.513118515115409,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.22046318285337,
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": 462776.96354896243,
174
+ "upload_bytes_per_second": 5621658.650253022,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 99.30116348125128,
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": 1.0387209649998113,
184
+ "bytes": 1064900,
185
+ "pages": 32,
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 1.1239516590003404,
190
+ "bytes": 11388095,
191
+ "pages": 24,
192
+ "shards": 3
193
  },
194
  {
195
+ "seconds": 1.270762184999512,
196
+ "bytes": 12213784,
197
+ "pages": 48,
198
+ "shards": 4
199
  },
200
  {
201
+ "seconds": 1.305295600000136,
202
+ "bytes": 1053111,
203
+ "pages": 30,
204
+ "shards": 1
205
  },
206
  {
207
+ "seconds": 1.13043969499995,
208
+ "bytes": 7715241,
209
+ "pages": 34,
210
  "shards": 3
211
  },
212
  {
213
+ "seconds": 1.1387970109999515,
214
+ "bytes": 10453428,
215
+ "pages": 40,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.3777619700003925,
220
+ "bytes": 819264,
221
+ "pages": 26,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.2043475289992784,
226
+ "bytes": 13122126,
227
+ "pages": 40,
228
+ "shards": 4
229
+ },
230
+ {
231
+ "seconds": 1.1971378379994349,
232
+ "bytes": 5971358,
233
  "pages": 36,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 0.9979249650004931,
238
+ "bytes": 6639528,
239
+ "pages": 32,
240
+ "shards": 3
241
+ },
242
+ {
243
+ "seconds": 1.0738694510000641,
244
+ "bytes": 6116466,
245
+ "pages": 37,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 0.993601779000528,
250
+ "bytes": 1093689,
251
+ "pages": 34,
252
+ "shards": 1
253
+ },
254
+ {
255
+ "seconds": 1.0732885580000584,
256
+ "bytes": 7199351,
257
+ "pages": 29,
258
  "shards": 3
259
  },
260
  {
261
+ "seconds": 1.3326483809996716,
262
+ "bytes": 16444186,
263
  "pages": 37,
264
+ "shards": 4
265
  },
266
  {
267
+ "seconds": 0.9647154859994771,
268
+ "bytes": 2573191,
269
+ "pages": 37,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.2671719289992325,
274
+ "bytes": 6070484,
275
+ "pages": 36,
276
+ "shards": 2
277
+ },
278
+ {
279
+ "seconds": 0.9431054610013234,
280
+ "bytes": 842654,
281
+ "pages": 26,
282
  "shards": 1
283
  },
284
  {
285
+ "seconds": 1.3556453979999787,
286
+ "bytes": 11128099,
287
+ "pages": 32,
288
+ "shards": 4
289
+ },
290
+ {
291
+ "seconds": 1.1991763389996777,
292
+ "bytes": 5989028,
293
+ "pages": 39,
294
  "shards": 2
295
  },
296
  {
297
+ "seconds": 1.483354228999815,
298
+ "bytes": 13516314,
299
+ "pages": 40,
300
  "shards": 4
301
  },
302
  {
303
+ "seconds": 0.855230520999612,
304
+ "bytes": 838138,
305
+ "pages": 29,
306
  "shards": 1
307
  },
308
  {
309
+ "seconds": 1.1489447340009065,
310
+ "bytes": 10403004,
311
+ "pages": 36,
312
+ "shards": 3
313
+ },
314
+ {
315
+ "seconds": 0.9511843500004034,
316
+ "bytes": 2374318,
317
+ "pages": 29,
318
+ "shards": 2
319
+ },
320
+ {
321
+ "seconds": 1.678345239999544,
322
+ "bytes": 6141170,
323
  "pages": 36,
324
  "shards": 2
325
  },
326
  {
327
+ "seconds": 1.0750775129999965,
328
+ "bytes": 1235188,
329
+ "pages": 35,
330
+ "shards": 1
331
+ },
332
+ {
333
+ "seconds": 1.2133236099998612,
334
+ "bytes": 11087663,
335
+ "pages": 9,
336
+ "shards": 3
337
+ },
338
+ {
339
+ "seconds": 1.060693833998812,
340
+ "bytes": 7602919,
341
+ "pages": 45,
342
+ "shards": 3
343
  },
344
  {
345
+ "seconds": 1.202748789000907,
346
+ "bytes": 6204281,
347
  "pages": 36,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 1.4798138580008526,
352
+ "bytes": 7393031,
353
  "pages": 30,
354
  "shards": 3
355
  },
356
  {
357
+ "seconds": 0.8912479129994608,
358
+ "bytes": 1116344,
359
  "pages": 34,
360
  "shards": 1
361
  },
362
  {
363
+ "seconds": 1.331704049000109,
364
+ "bytes": 10313766,
365
+ "pages": 37,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.5044031500001438,
370
+ "bytes": 8199304,
371
  "pages": 34,
372
  "shards": 3
373
  },
374
  {
375
+ "seconds": 1.4808365860008053,
376
+ "bytes": 6083362,
377
+ "pages": 37,
378
  "shards": 2
379
  },
380
  {
381
+ "seconds": 1.2456822759995703,
382
+ "bytes": 12102575,
383
+ "pages": 32,
384
+ "shards": 4
385
  },
386
  {
387
+ "seconds": 0.872640274999867,
388
+ "bytes": 1185922,
389
+ "pages": 35,
390
  "shards": 1
391
  },
392
  {
393
+ "seconds": 1.1524197270009608,
394
+ "bytes": 6081827,
395
+ "pages": 36,
396
+ "shards": 2
397
  },
398
  {
399
+ "seconds": 1.1447629150006833,
400
+ "bytes": 7321171,
401
+ "pages": 30,
402
  "shards": 3
403
  },
404
  {
405
+ "seconds": 1.0708932610014017,
406
+ "bytes": 6812845,
407
  "pages": 32,
 
 
 
 
 
 
408
  "shards": 3
409
  },
410
  {
411
+ "seconds": 1.235942547000377,
412
+ "bytes": 11771801,
413
+ "pages": 40,
414
  "shards": 4
415
  },
416
  {
417
+ "seconds": 1.195207675000347,
418
+ "bytes": 1134238,
419
+ "pages": 34,
420
  "shards": 1
421
  },
422
  {
423
+ "seconds": 1.0769621990002634,
424
+ "bytes": 6468284,
425
+ "pages": 36,
426
+ "shards": 2
427
+ },
428
+ {
429
+ "seconds": 2.3685029429998394,
430
+ "bytes": 11410659,
431
  "pages": 34,
432
+ "shards": 4
433
  },
434
  {
435
+ "seconds": 0.8315555410008528,
436
+ "bytes": 1017212,
437
+ "pages": 30,
438
+ "shards": 1
439
  },
440
  {
441
+ "seconds": 1.2100102580006933,
442
+ "bytes": 5596380,
443
+ "pages": 21,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.2646026780003012,
448
+ "bytes": 13750693,
449
+ "pages": 48,
450
  "shards": 4
451
  },
452
  {
453
+ "seconds": 1.3428173240008618,
454
+ "bytes": 5558514,
455
+ "pages": 31,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 1.273476110000047,
460
+ "bytes": 7479055,
461
+ "pages": 38,
462
  "shards": 3
463
  },
464
  {
465
+ "seconds": 1.084104961999401,
466
+ "bytes": 947822,
467
+ "pages": 26,
468
+ "shards": 1
469
+ },
470
+ {
471
+ "seconds": 1.1121349749992078,
472
+ "bytes": 6275461,
473
  "pages": 37,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.2975029250010266,
478
+ "bytes": 6264701,
479
+ "pages": 26,
480
+ "shards": 3
481
  },
482
  {
483
+ "seconds": 1.4136529820007127,
484
+ "bytes": 11858220,
485
+ "pages": 37,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.0470604990005086,
490
+ "bytes": 6246532,
491
  "pages": 37,
492
+ "shards": 2
493
  },
494
  {
495
+ "seconds": 1.0089250809996884,
496
+ "bytes": 2580280,
497
  "pages": 37,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.1785594149987446,
502
+ "bytes": 10228291,
503
+ "pages": 34,
504
+ "shards": 3
505
+ },
506
+ {
507
+ "seconds": 0.9172461540001677,
508
+ "bytes": 2354642,
509
+ "pages": 28,
510
  "shards": 2
511
  },
512
  {
513
+ "seconds": 1.1485085579988663,
514
+ "bytes": 12167168,
515
+ "pages": 40,
516
+ "shards": 3
517
  },
518
  {
519
+ "seconds": 1.0151017729986052,
520
+ "bytes": 6037243,
521
+ "pages": 35,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.2390309599995817,
526
+ "bytes": 6886865,
527
+ "pages": 35,
528
+ "shards": 3
529
+ },
530
+ {
531
+ "seconds": 1.3525706070013257,
532
+ "bytes": 5675921,
533
+ "pages": 28,
534
  "shards": 2
535
  },
536
  {
537
+ "seconds": 1.2551854249995813,
538
+ "bytes": 7559192,
539
+ "pages": 38,
540
+ "shards": 3
541
  },
542
  {
543
+ "seconds": 0.825367109000581,
544
+ "bytes": 1141341,
545
+ "pages": 33,
546
  "shards": 1
547
  },
548
  {
549
+ "seconds": 1.5828043500005151,
550
+ "bytes": 15814359,
551
+ "pages": 33,
552
+ "shards": 4
553
  },
554
  {
555
+ "seconds": 1.3585308149995399,
556
+ "bytes": 7486305,
557
+ "pages": 37,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 0.8041142309994029,
562
+ "bytes": 1148050,
563
+ "pages": 34,
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:a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e
3
- size 721763327
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b0cdf7569ff3f756bba96adbbd77bced2a88ef4f965a14cbe15078a1fd95894
3
+ size 721780543