appvoid commited on
Commit
5aa37e6
·
verified ·
1 Parent(s): 3e28ed5

Qwen90M step 840000; complete resumable checkpoint

Browse files
Files changed (6) hide show
  1. complete.json +5 -5
  2. coverage.json +48 -48
  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": 835000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
- "shard_scheduler.json": "15ec84a118722aa0da30b1eedddde0fa6f0d20effd89f99c89d67b1b955f9b9e",
15
- "coverage.json": "2268abb420311a79a47ad9499884d787f6b2b1069f26033253f7830b5909fc05",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
 
1
  {
2
+ "step": 840000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "3e144501faba5106e3f43a2e36d221a8630b9cdfb6837a8f976f58b534e86cfa",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "c1114d95865c877e56d4ca2db61e9232398961a900ec723d9fe48b310b10c038",
12
  "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
13
  "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7",
14
+ "shard_scheduler.json": "5d089facc2fe4bc290dfa5a25481443cf9a74f9c2860a1b57795e6bdc0906c9e",
15
+ "coverage.json": "abfd3dc053aa9107e89eeb9bfe2735dc1cd78f0a1640546271b969e4322c5588",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 835000,
3
  "stats": {
4
- "climbmix/blocks": 30060000,
5
- "climbmix/targets": 50250362065,
6
- "climbmix/input_tokens": 50253546093,
7
- "ultra_qa/blocks": 11272500,
8
- "ultra_qa/targets": 18390084306,
9
- "ultra_qa/input_tokens": 18390119687,
10
- "cortex/blocks": 11272500,
11
- "cortex/targets": 428641870,
12
- "cortex/input_tokens": 17918505142,
13
- "ultra_style/blocks": 11272500,
14
- "ultra_style/targets": 17829475019,
15
- "ultra_style/input_tokens": 17830410407,
16
- "openmath/blocks": 7515000,
17
- "openmath/targets": 11612035358,
18
- "openmath/input_tokens": 11615904568,
19
- "rewrite6/blocks": 3757500,
20
- "rewrite6/targets": 6159911854,
21
- "rewrite6/input_tokens": 6166665342
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 4116,
26
- "row": 3220,
27
  "start": {
28
- "file": 199,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 20,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
- "offset": 0,
37
- "documents": 16862356
38
  },
39
  "rewrite6": {
40
- "page": 42,
41
- "row": 790,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 172032
47
  }
48
  },
49
- "epoch": 46,
50
  "offset": 0,
51
- "documents": 9621406
52
  },
53
  "openmath": {
54
- "page": 2572,
55
- "row": 4072,
56
  "start": {
57
  "file": 24,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 55,
61
- "row": 313
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 10538984
67
  },
68
  "ultra_qa": {
69
- "page": 1112,
70
- "row": 1914,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 6,
76
- "row": 10507
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 4556666
82
  },
83
  "ultra_style": {
84
- "page": 1698,
85
- "row": 306,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 1,
91
- "row": 20264
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 6955314
97
  },
98
  "cortex": {
99
- "page": 218617,
100
- "row": 105,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 27983081
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 840000,
3
  "stats": {
4
+ "climbmix/blocks": 30240000,
5
+ "climbmix/targets": 50551290330,
6
+ "climbmix/input_tokens": 50554493320,
7
+ "ultra_qa/blocks": 11340000,
8
+ "ultra_qa/targets": 18500378600,
9
+ "ultra_qa/input_tokens": 18500414184,
10
+ "cortex/blocks": 11340000,
11
+ "cortex/targets": 431213356,
12
+ "cortex/input_tokens": 18025811942,
13
+ "ultra_style/blocks": 11340000,
14
+ "ultra_style/targets": 17936310265,
15
+ "ultra_style/input_tokens": 17937251230,
16
+ "openmath/blocks": 7560000,
17
+ "openmath/targets": 11681504015,
18
+ "openmath/input_tokens": 11685396540,
19
+ "rewrite6/blocks": 3780000,
20
+ "rewrite6/targets": 6196813702,
21
+ "rewrite6/input_tokens": 6203607570
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 4141,
26
+ "row": 1870,
27
  "start": {
28
+ "file": 200,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 37,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
+ "offset": 2048,
37
+ "documents": 16963406
38
  },
39
  "rewrite6": {
40
+ "page": 6,
41
+ "row": 281,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 24576
47
  }
48
  },
49
+ "epoch": 47,
50
  "offset": 0,
51
+ "documents": 9678845
52
  },
53
  "openmath": {
54
+ "page": 2588,
55
+ "row": 1820,
56
  "start": {
57
  "file": 24,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 120,
61
+ "row": 849
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 10602268
67
  },
68
  "ultra_qa": {
69
+ "page": 1119,
70
+ "row": 459,
71
  "start": {
72
  "file": 8,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 6,
76
+ "row": 39179
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 0,
81
+ "documents": 4583883
82
  },
83
  "ultra_style": {
84
+ "page": 1708,
85
+ "row": 803,
86
  "start": {
87
  "file": 10,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 1,
91
+ "row": 61224
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 6996771
97
  },
98
  "cortex": {
99
+ "page": 219925,
100
+ "row": 120,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 28150520
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 835000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "15ec84a118722aa0da30b1eedddde0fa6f0d20effd89f99c89d67b1b955f9b9e", "coverage.json": "2268abb420311a79a47ad9499884d787f6b2b1069f26033253f7830b5909fc05", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000835000"}
 
1
+ {"step": 840000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "3e144501faba5106e3f43a2e36d221a8630b9cdfb6837a8f976f58b534e86cfa", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "c1114d95865c877e56d4ca2db61e9232398961a900ec723d9fe48b310b10c038", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "eb45b2d9ecb051b6b7e592c18f2a0bfebf4278147b783ba131f7a9a4453a9ec7", "shard_scheduler.json": "5d089facc2fe4bc290dfa5a25481443cf9a74f9c2860a1b57795e6bdc0906c9e", "coverage.json": "abfd3dc053aa9107e89eeb9bfe2735dc1cd78f0a1640546271b969e4322c5588", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000840000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:434bc5e66416d00f38daf4d1c20bd1d7d620b0fc06f97108edac543c845dc140
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e144501faba5106e3f43a2e36d221a8630b9cdfb6837a8f976f58b534e86cfa
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": 76407.9111220439,
26
- "required_pages_per_second_upper_estimate": 0.0052860765822027025
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": 9550.988890255487,
40
- "required_pages_per_second_upper_estimate": 0.003998799601022032
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": 19101.977780510973,
54
- "required_pages_per_second_upper_estimate": 0.00397843821865044
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": 28652.966670766462,
68
- "required_pages_per_second_upper_estimate": 0.001717701601365582
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": 28652.966670766462,
82
- "required_pages_per_second_upper_estimate": 0.0022276421275710828
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": 28652.966670766462,
96
- "required_pages_per_second_upper_estimate": 0.3675294912939349
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,410 @@
158
  "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.10879093923438177,
162
- "producer_p95_seconds": 0.32612680799502414,
163
- "gpu_update_seconds": 0.964926261133309,
164
- "producer_batches_per_second": 9.191941967203505,
165
- "gpu_batches_per_second": 1.0363486209044583,
166
- "producer_headroom": 8.86954619496803,
167
  "sustainable": true,
168
- "buffer_seconds": 60.79035445139847,
169
  "buffer_drain_seconds": null,
170
- "shard_target_mib": 32,
171
- "upload_pages_per_commit": 165,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 666855.3792614836,
174
- "upload_bytes_per_second": 3927796.2312513823,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 71.54582039539042,
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": 1.5374675419989217,
184
- "bytes": 6254376,
185
  "pages": 36,
 
 
 
 
 
 
186
  "shards": 2
187
  },
188
  {
189
- "seconds": 1.5695442620017275,
190
- "bytes": 11600439,
191
- "pages": 29,
192
- "shards": 3
193
  },
194
  {
195
- "seconds": 1.7552413789999264,
196
- "bytes": 6467958,
197
- "pages": 38,
198
- "shards": 3
199
  },
200
  {
201
- "seconds": 1.6743759019991558,
202
- "bytes": 6275115,
203
- "pages": 37,
204
  "shards": 2
205
  },
206
  {
207
- "seconds": 2.272818253000878,
208
- "bytes": 7589463,
209
- "pages": 39,
210
  "shards": 3
211
  },
212
  {
213
- "seconds": 1.5132595840004797,
214
- "bytes": 5701329,
215
- "pages": 28,
216
- "shards": 2
217
  },
218
  {
219
- "seconds": 1.5889989900024375,
220
- "bytes": 6430899,
221
- "pages": 37,
222
- "shards": 2
223
  },
224
  {
225
- "seconds": 1.6179743659995438,
226
- "bytes": 8234542,
227
- "pages": 31,
228
- "shards": 3
229
  },
230
  {
231
- "seconds": 1.6133322549976583,
232
- "bytes": 6059251,
233
- "pages": 37,
234
  "shards": 2
235
  },
236
  {
237
- "seconds": 1.6795359120005742,
238
- "bytes": 5921876,
239
- "pages": 24,
240
  "shards": 2
241
  },
242
  {
243
- "seconds": 1.7577662710027653,
244
- "bytes": 6866333,
245
- "pages": 46,
246
  "shards": 3
247
  },
248
  {
249
- "seconds": 1.93685000899859,
250
- "bytes": 6023833,
251
- "pages": 33,
252
  "shards": 2
253
  },
254
  {
255
- "seconds": 1.8011142590003146,
256
- "bytes": 7499768,
257
- "pages": 34,
258
- "shards": 3
259
- },
260
- {
261
- "seconds": 1.6402012090002245,
262
- "bytes": 10865294,
263
- "pages": 22,
264
  "shards": 3
265
  },
266
  {
267
- "seconds": 1.7843240430011065,
268
- "bytes": 6397585,
269
- "pages": 42,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.4411041519997525,
274
- "bytes": 2727332,
275
- "pages": 38,
276
  "shards": 3
277
  },
278
  {
279
- "seconds": 1.913498541998706,
280
- "bytes": 6177731,
281
- "pages": 28,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 2.387397704002069,
286
- "bytes": 11768966,
287
  "pages": 40,
288
- "shards": 4
289
  },
290
  {
291
- "seconds": 1.479657814001257,
292
- "bytes": 836473,
293
- "pages": 27,
294
  "shards": 1
295
  },
296
  {
297
- "seconds": 1.9911568649986293,
298
- "bytes": 11961950,
299
- "pages": 42,
300
  "shards": 3
301
  },
302
  {
303
- "seconds": 1.958790611999575,
304
- "bytes": 2439502,
305
- "pages": 29,
306
- "shards": 2
307
  },
308
  {
309
- "seconds": 1.8466758049980854,
310
- "bytes": 10680095,
311
- "pages": 40,
312
- "shards": 3
313
  },
314
  {
315
- "seconds": 1.7479712950007524,
316
- "bytes": 6009903,
317
- "pages": 28,
318
- "shards": 2
319
  },
320
  {
321
- "seconds": 2.2005035470028815,
322
- "bytes": 2644353,
323
  "pages": 37,
324
  "shards": 2
325
  },
326
  {
327
- "seconds": 1.5131899659972987,
328
- "bytes": 5828306,
329
- "pages": 28,
330
- "shards": 2
331
  },
332
  {
333
- "seconds": 1.8049468739991426,
334
- "bytes": 4697774,
335
- "pages": 26,
336
  "shards": 2
337
  },
338
  {
339
- "seconds": 2.496297884001251,
340
- "bytes": 13693480,
341
- "pages": 42,
342
- "shards": 4
343
  },
344
  {
345
- "seconds": 1.8265930059969833,
346
- "bytes": 5855746,
347
- "pages": 36,
348
  "shards": 2
349
  },
350
  {
351
- "seconds": 1.3942693540011533,
352
- "bytes": 2663289,
353
- "pages": 34,
354
  "shards": 2
355
  },
356
  {
357
- "seconds": 1.5531271910003852,
358
- "bytes": 10487346,
359
- "pages": 37,
360
- "shards": 3
361
  },
362
  {
363
- "seconds": 1.5084796250012005,
364
- "bytes": 1063958,
365
- "pages": 32,
366
- "shards": 1
367
  },
368
  {
369
- "seconds": 1.9720517980022123,
370
- "bytes": 7314841,
371
  "pages": 27,
372
- "shards": 3
373
  },
374
  {
375
- "seconds": 1.793492888999026,
376
- "bytes": 11753421,
377
  "pages": 38,
378
- "shards": 3
379
  },
380
  {
381
- "seconds": 1.5141362069989555,
382
- "bytes": 6622032,
383
- "pages": 34,
384
  "shards": 3
385
  },
386
  {
387
- "seconds": 1.5094903019999037,
388
- "bytes": 6206747,
389
- "pages": 38,
390
- "shards": 2
391
  },
392
  {
393
- "seconds": 1.7292475489994104,
394
- "bytes": 6419421,
395
- "pages": 37,
396
  "shards": 2
397
  },
398
  {
399
- "seconds": 1.6914732460027153,
400
- "bytes": 2366813,
401
- "pages": 27,
402
  "shards": 2
403
  },
404
  {
405
- "seconds": 1.8487111699978414,
406
- "bytes": 10763047,
407
- "pages": 31,
408
- "shards": 3
409
- },
410
- {
411
- "seconds": 1.605599512000481,
412
- "bytes": 8445757,
413
  "pages": 39,
414
- "shards": 3
415
  },
416
  {
417
- "seconds": 1.6673135730015929,
418
- "bytes": 6027005,
419
- "pages": 28,
420
- "shards": 2
421
  },
422
  {
423
- "seconds": 1.4991452859976562,
424
- "bytes": 1121499,
425
  "pages": 36,
426
  "shards": 1
427
  },
428
  {
429
- "seconds": 2.2660662760026753,
430
- "bytes": 11276737,
431
- "pages": 36,
432
- "shards": 4
433
  },
434
  {
435
- "seconds": 1.8010207819970674,
436
- "bytes": 5858104,
437
- "pages": 32,
438
- "shards": 2
439
  },
440
  {
441
- "seconds": 1.3410764899963397,
442
- "bytes": 1019830,
443
- "pages": 31,
444
- "shards": 1
445
  },
446
  {
447
- "seconds": 1.8211984559966368,
448
- "bytes": 13136869,
449
- "pages": 32,
450
- "shards": 4
451
  },
452
  {
453
- "seconds": 1.4420257410019985,
454
- "bytes": 5442386,
455
- "pages": 33,
456
- "shards": 2
457
  },
458
  {
459
- "seconds": 2.6930293979967246,
460
- "bytes": 5991867,
461
- "pages": 29,
462
  "shards": 3
463
  },
464
  {
465
- "seconds": 1.7729905640007928,
466
- "bytes": 7619842,
467
- "pages": 39,
468
  "shards": 3
469
  },
470
  {
471
- "seconds": 1.2987675089971162,
472
- "bytes": 1223490,
473
- "pages": 35,
474
- "shards": 1
475
  },
476
  {
477
- "seconds": 1.546362834997126,
478
- "bytes": 11822887,
479
- "pages": 33,
480
  "shards": 4
481
  },
482
  {
483
- "seconds": 1.5667999549987144,
484
- "bytes": 6923893,
485
- "pages": 35,
486
- "shards": 2
487
  },
488
  {
489
- "seconds": 1.4826895270016394,
490
- "bytes": 6163509,
491
- "pages": 32,
492
  "shards": 2
493
  },
494
  {
495
- "seconds": 1.7263165549957193,
496
- "bytes": 7465725,
497
- "pages": 39,
498
  "shards": 3
499
  },
500
  {
501
- "seconds": 2.012805174999812,
502
- "bytes": 4595806,
503
- "pages": 30,
504
  "shards": 2
505
  },
506
  {
507
- "seconds": 1.6076512380022905,
508
- "bytes": 7510724,
509
- "pages": 40,
510
  "shards": 3
511
  },
512
  {
513
- "seconds": 1.5883049809999648,
514
- "bytes": 5665128,
515
- "pages": 27,
516
  "shards": 2
517
  },
518
  {
519
- "seconds": 2.0745964650050155,
520
- "bytes": 11457524,
 
 
 
 
 
 
521
  "pages": 35,
522
  "shards": 3
523
  },
524
  {
525
- "seconds": 1.6211695970050641,
526
- "bytes": 7634724,
527
- "pages": 40,
528
- "shards": 3
529
  },
530
  {
531
- "seconds": 1.4634708860030514,
532
- "bytes": 987981,
533
  "pages": 29,
534
- "shards": 1
535
  },
536
  {
537
- "seconds": 1.6518487119974452,
538
- "bytes": 7534957,
539
- "pages": 37,
540
- "shards": 3
541
  },
542
  {
543
- "seconds": 1.5817108970004483,
544
- "bytes": 10450177,
545
- "pages": 33,
 
 
 
 
 
 
546
  "shards": 3
547
  },
548
  {
549
- "seconds": 1.328966004999529,
550
- "bytes": 1184587,
551
- "pages": 36,
552
  "shards": 1
553
  },
554
  {
555
- "seconds": 2.0095185120007955,
556
- "bytes": 12961722,
557
- "pages": 36,
558
- "shards": 4
559
  },
560
  {
561
- "seconds": 1.4560003849983332,
562
- "bytes": 6473170,
563
- "pages": 37,
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": 77053.7277135633,
26
+ "required_pages_per_second_upper_estimate": 0.005330755672504961
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": 9631.715964195413,
40
+ "required_pages_per_second_upper_estimate": 0.0040325983411076795
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": 19263.431928390826,
54
+ "required_pages_per_second_upper_estimate": 0.004012064859821606
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": 28895.147892586243,
68
+ "required_pages_per_second_upper_estimate": 0.0017322199958243636
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": 28895.147892586243,
82
+ "required_pages_per_second_upper_estimate": 0.0022464706523249544
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": 28895.147892586243,
96
+ "required_pages_per_second_upper_estimate": 0.3706359319735027
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": 0.15130693903114434,
162
+ "producer_p95_seconds": 0.37083707300189417,
163
+ "gpu_update_seconds": 0.956838847227142,
164
+ "producer_batches_per_second": 6.609082216607161,
165
+ "gpu_batches_per_second": 1.0451080690316203,
166
+ "producer_headroom": 6.3238266093678,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.280847375309946,
169
  "buffer_drain_seconds": null,
170
+ "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 82,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 369362.01326889056,
174
+ "upload_bytes_per_second": 4183866.9114862867,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 79.25648973989523,
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": 2.2660662760026753,
184
+ "bytes": 11276737,
185
  "pages": 36,
186
+ "shards": 4
187
+ },
188
+ {
189
+ "seconds": 1.8010207819970674,
190
+ "bytes": 5858104,
191
+ "pages": 32,
192
  "shards": 2
193
  },
194
  {
195
+ "seconds": 1.3410764899963397,
196
+ "bytes": 1019830,
197
+ "pages": 31,
198
+ "shards": 1
199
  },
200
  {
201
+ "seconds": 1.8211984559966368,
202
+ "bytes": 13136869,
203
+ "pages": 32,
204
+ "shards": 4
205
  },
206
  {
207
+ "seconds": 1.4420257410019985,
208
+ "bytes": 5442386,
209
+ "pages": 33,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 2.6930293979967246,
214
+ "bytes": 5991867,
215
+ "pages": 29,
216
  "shards": 3
217
  },
218
  {
219
+ "seconds": 1.7729905640007928,
220
+ "bytes": 7619842,
221
+ "pages": 39,
222
+ "shards": 3
223
  },
224
  {
225
+ "seconds": 1.2987675089971162,
226
+ "bytes": 1223490,
227
+ "pages": 35,
228
+ "shards": 1
229
  },
230
  {
231
+ "seconds": 1.546362834997126,
232
+ "bytes": 11822887,
233
+ "pages": 33,
234
+ "shards": 4
235
  },
236
  {
237
+ "seconds": 1.5667999549987144,
238
+ "bytes": 6923893,
239
+ "pages": 35,
240
  "shards": 2
241
  },
242
  {
243
+ "seconds": 1.4826895270016394,
244
+ "bytes": 6163509,
245
+ "pages": 32,
246
  "shards": 2
247
  },
248
  {
249
+ "seconds": 1.7263165549957193,
250
+ "bytes": 7465725,
251
+ "pages": 39,
252
  "shards": 3
253
  },
254
  {
255
+ "seconds": 2.012805174999812,
256
+ "bytes": 4595806,
257
+ "pages": 30,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.6076512380022905,
262
+ "bytes": 7510724,
263
+ "pages": 40,
 
 
 
 
 
 
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.5883049809999648,
268
+ "bytes": 5665128,
269
+ "pages": 27,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 2.0745964650050155,
274
+ "bytes": 11457524,
275
+ "pages": 35,
276
  "shards": 3
277
  },
278
  {
279
+ "seconds": 1.6211695970050641,
280
+ "bytes": 7634724,
 
 
 
 
 
 
281
  "pages": 40,
282
+ "shards": 3
283
  },
284
  {
285
+ "seconds": 1.4634708860030514,
286
+ "bytes": 987981,
287
+ "pages": 29,
288
  "shards": 1
289
  },
290
  {
291
+ "seconds": 1.6518487119974452,
292
+ "bytes": 7534957,
293
+ "pages": 37,
294
  "shards": 3
295
  },
296
  {
297
+ "seconds": 1.5817108970004483,
298
+ "bytes": 10450177,
299
+ "pages": 33,
300
+ "shards": 3
301
  },
302
  {
303
+ "seconds": 1.328966004999529,
304
+ "bytes": 1184587,
305
+ "pages": 36,
306
+ "shards": 1
307
  },
308
  {
309
+ "seconds": 2.0095185120007955,
310
+ "bytes": 12961722,
311
+ "pages": 36,
312
+ "shards": 4
313
  },
314
  {
315
+ "seconds": 1.4560003849983332,
316
+ "bytes": 6473170,
317
  "pages": 37,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.6064382549957372,
322
+ "bytes": 6604109,
323
+ "pages": 34,
324
+ "shards": 3
325
  },
326
  {
327
+ "seconds": 1.684055683996121,
328
+ "bytes": 5176791,
329
+ "pages": 11,
330
  "shards": 2
331
  },
332
  {
333
+ "seconds": 1.8713450549985282,
334
+ "bytes": 942860,
335
+ "pages": 27,
336
+ "shards": 1
337
  },
338
  {
339
+ "seconds": 1.5147400999994716,
340
+ "bytes": 6664610,
341
+ "pages": 37,
342
  "shards": 2
343
  },
344
  {
345
+ "seconds": 1.4935083919990575,
346
+ "bytes": 2459570,
347
+ "pages": 28,
348
  "shards": 2
349
  },
350
  {
351
+ "seconds": 2.1903198990039527,
352
+ "bytes": 16888519,
353
+ "pages": 41,
354
+ "shards": 4
355
  },
356
  {
357
+ "seconds": 2.158359011998982,
358
+ "bytes": 7618380,
359
+ "pages": 40,
360
+ "shards": 3
361
  },
362
  {
363
+ "seconds": 1.546953492994362,
364
+ "bytes": 982772,
365
  "pages": 27,
366
+ "shards": 1
367
  },
368
  {
369
+ "seconds": 1.5179005070021958,
370
+ "bytes": 6221682,
371
  "pages": 38,
372
+ "shards": 2
373
  },
374
  {
375
+ "seconds": 1.4849864080024417,
376
+ "bytes": 6331011,
377
+ "pages": 36,
378
  "shards": 3
379
  },
380
  {
381
+ "seconds": 1.7072867230017437,
382
+ "bytes": 11821281,
383
+ "pages": 36,
384
+ "shards": 3
385
  },
386
  {
387
+ "seconds": 1.688430515001528,
388
+ "bytes": 5778338,
389
+ "pages": 29,
390
  "shards": 2
391
  },
392
  {
393
+ "seconds": 1.6583789540018188,
394
+ "bytes": 2527741,
395
+ "pages": 35,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.606587074995332,
400
+ "bytes": 10744249,
 
 
 
 
 
 
401
  "pages": 39,
402
+ "shards": 4
403
  },
404
  {
405
+ "seconds": 1.0940176939984667,
406
+ "bytes": 7184499,
407
+ "pages": 30,
408
+ "shards": 3
409
  },
410
  {
411
+ "seconds": 1.0425019949980197,
412
+ "bytes": 1083508,
413
  "pages": 36,
414
  "shards": 1
415
  },
416
  {
417
+ "seconds": 1.5655531079974025,
418
+ "bytes": 12067871,
419
+ "pages": 35,
420
+ "shards": 3
421
  },
422
  {
423
+ "seconds": 1.9511204809969058,
424
+ "bytes": 6917854,
425
+ "pages": 33,
426
+ "shards": 3
427
  },
428
  {
429
+ "seconds": 1.1665768739985651,
430
+ "bytes": 5996687,
431
+ "pages": 35,
432
+ "shards": 2
433
  },
434
  {
435
+ "seconds": 1.6431592789958813,
436
+ "bytes": 7717315,
437
+ "pages": 37,
438
+ "shards": 3
439
  },
440
  {
441
+ "seconds": 1.3486441710047075,
442
+ "bytes": 976930,
443
+ "pages": 30,
444
+ "shards": 1
445
  },
446
  {
447
+ "seconds": 1.5532393190005678,
448
+ "bytes": 10410458,
449
+ "pages": 39,
450
  "shards": 3
451
  },
452
  {
453
+ "seconds": 1.5896227589983027,
454
+ "bytes": 8024141,
455
+ "pages": 26,
456
  "shards": 3
457
  },
458
  {
459
+ "seconds": 1.570418814997538,
460
+ "bytes": 6148151,
461
+ "pages": 41,
462
+ "shards": 2
463
  },
464
  {
465
+ "seconds": 1.876006619000691,
466
+ "bytes": 11510011,
467
+ "pages": 30,
468
  "shards": 4
469
  },
470
  {
471
+ "seconds": 1.1747664850045112,
472
+ "bytes": 1216277,
473
+ "pages": 36,
474
+ "shards": 1
475
  },
476
  {
477
+ "seconds": 1.0712760539972805,
478
+ "bytes": 6301436,
479
+ "pages": 37,
480
  "shards": 2
481
  },
482
  {
483
+ "seconds": 1.2888189970035455,
484
+ "bytes": 7385771,
485
+ "pages": 32,
486
  "shards": 3
487
  },
488
  {
489
+ "seconds": 1.5881777600006899,
490
+ "bytes": 5494719,
491
+ "pages": 25,
492
  "shards": 2
493
  },
494
  {
495
+ "seconds": 1.394116926996503,
496
+ "bytes": 11997464,
497
+ "pages": 39,
498
  "shards": 3
499
  },
500
  {
501
+ "seconds": 1.7342578780007898,
502
+ "bytes": 2696654,
503
+ "pages": 39,
504
  "shards": 2
505
  },
506
  {
507
+ "seconds": 1.611772108997684,
508
+ "bytes": 5903925,
509
+ "pages": 29,
510
+ "shards": 2
511
+ },
512
+ {
513
+ "seconds": 1.512940188003995,
514
+ "bytes": 7372289,
515
  "pages": 35,
516
  "shards": 3
517
  },
518
  {
519
+ "seconds": 0.9355933799961349,
520
+ "bytes": 5063118,
521
+ "pages": 34,
522
+ "shards": 2
523
  },
524
  {
525
+ "seconds": 1.774848651002685,
526
+ "bytes": 5953016,
527
  "pages": 29,
528
+ "shards": 2
529
  },
530
  {
531
+ "seconds": 1.877187440994021,
532
+ "bytes": 13513556,
533
+ "pages": 41,
534
+ "shards": 4
535
  },
536
  {
537
+ "seconds": 1.8802343610004755,
538
+ "bytes": 5656226,
539
+ "pages": 32,
540
+ "shards": 2
541
+ },
542
+ {
543
+ "seconds": 1.6476056760002393,
544
+ "bytes": 7839870,
545
+ "pages": 38,
546
  "shards": 3
547
  },
548
  {
549
+ "seconds": 1.3661578879982699,
550
+ "bytes": 1247022,
551
+ "pages": 35,
552
  "shards": 1
553
  },
554
  {
555
+ "seconds": 1.4103239969990682,
556
+ "bytes": 5948663,
557
+ "pages": 28,
558
+ "shards": 2
559
  },
560
  {
561
+ "seconds": 1.2626624109980185,
562
+ "bytes": 7374458,
563
+ "pages": 32,
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 sha256:8fc02097e06a5d8dde4c92534118b36500d860341974727dbdf1235db028d4da
3
- size 721778239
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1114d95865c877e56d4ca2db61e9232398961a900ec723d9fe48b310b10c038
3
+ size 721765631