appvoid commited on
Commit
087bf40
·
verified ·
1 Parent(s): 00d2ebc

Qwen90M step 280000; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +47 -47
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. performance.json +3 -3
  6. shard_scheduler.json +32 -404
  7. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 275000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "2af807d46a42c19f19c7b32b800e214d63c1433a37862b054d1767b88ee00e92",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "8f706fccdd6b87209ff5215abd019c5c04c63cfa5c3c1d339b8f5d3a1bdf7fe4",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
- "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7",
14
- "shard_scheduler.json": "d4146d4949443885a89e71741ab422c603360627ea40e296539a0be0a218caee",
15
- "coverage.json": "8b908eef3d6b156adf6b83ba5c2f7263f4abb751ab7df3899af5bff2595659e7",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 280000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "b202beb2e692c902003e4624a58ed61b492689af5528e336e78027d172cdc6a5",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "086c04c0e606c24ef7dab8710d39cde6eff9e0a7ec67fc90f9930de401842100",
12
  "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
+ "performance.json": "7ab839f3b8798329051e29b39e08bd08f8678f49ed79d09e3b70b9de78a4765c",
14
+ "shard_scheduler.json": "bb41f80e95e1faea1ad1243a8e7d9ee991b7eb64bf0e6b9b463722b029efa994",
15
+ "coverage.json": "c7677efe249bc40dea701da173ea31fe609c61d934d175645935757dc9bffe1c",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 275000,
3
  "stats": {
4
- "climbmix/blocks": 9900000,
5
- "climbmix/targets": 16549266942,
6
- "climbmix/input_tokens": 16550315706,
7
- "ultra_qa/blocks": 3712500,
8
- "ultra_qa/targets": 6056418591,
9
- "ultra_qa/input_tokens": 6056430195,
10
- "cortex/blocks": 3712500,
11
- "cortex/targets": 141226525,
12
- "cortex/input_tokens": 5901276792,
13
- "ultra_style/blocks": 3712500,
14
- "ultra_style/targets": 5871811169,
15
- "ultra_style/input_tokens": 5872119620,
16
- "openmath/blocks": 2475000,
17
- "openmath/targets": 3824405251,
18
- "openmath/input_tokens": 3825680165,
19
- "rewrite6/blocks": 1237500,
20
- "rewrite6/targets": 2028675231,
21
- "rewrite6/input_tokens": 2030899715
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1355,
26
- "row": 3394,
27
  "start": {
28
- "file": 65,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 48,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 5553474
38
  },
39
  "rewrite6": {
40
- "page": 21,
41
- "row": 1930,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 86016
47
  }
48
  },
49
  "epoch": 15,
50
  "offset": 0,
51
- "documents": 3169006
52
  },
53
  "openmath": {
54
- "page": 847,
55
- "row": 2438,
56
  "start": {
57
- "file": 7,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 412,
61
- "row": 762
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 3471750
67
  },
68
  "ultra_qa": {
69
- "page": 366,
70
- "row": 1530,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
- "row": 7993
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 1500666
82
  },
83
  "ultra_style": {
84
- "page": 559,
85
- "row": 2401,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 2,
91
- "row": 63989
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 2292065
97
  },
98
  "cortex": {
99
- "page": 71999,
100
- "row": 4,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 9215876
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 280000,
3
  "stats": {
4
+ "climbmix/blocks": 10080000,
5
+ "climbmix/targets": 16850702549,
6
+ "climbmix/input_tokens": 16851770299,
7
+ "ultra_qa/blocks": 3780000,
8
+ "ultra_qa/targets": 6166762492,
9
+ "ultra_qa/input_tokens": 6166774323,
10
+ "cortex/blocks": 3780000,
11
+ "cortex/targets": 143784713,
12
+ "cortex/input_tokens": 6008633757,
13
+ "ultra_style/blocks": 3780000,
14
+ "ultra_style/targets": 5978866890,
15
+ "ultra_style/input_tokens": 5979181245,
16
+ "openmath/blocks": 2520000,
17
+ "openmath/targets": 3893894891,
18
+ "openmath/input_tokens": 3895193089,
19
+ "rewrite6/blocks": 1260000,
20
+ "rewrite6/targets": 2065551785,
21
+ "rewrite6/input_tokens": 2067816586
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1380,
26
+ "row": 1360,
27
  "start": {
28
+ "file": 66,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 66,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 5653840
38
  },
39
  "rewrite6": {
40
+ "page": 35,
41
+ "row": 2118,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 143360
47
  }
48
  },
49
  "epoch": 15,
50
  "offset": 0,
51
+ "documents": 3226538
52
  },
53
  "openmath": {
54
+ "page": 863,
55
+ "row": 65,
56
  "start": {
57
+ "file": 8,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 41,
61
+ "row": 648
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 3534913
67
  },
68
  "ultra_qa": {
69
+ "page": 373,
70
+ "row": 122,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 7,
76
+ "row": 36665
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 1527930
82
  },
83
  "ultra_style": {
84
+ "page": 569,
85
+ "row": 3492,
86
  "start": {
87
  "file": 3,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 3,
91
+ "row": 18427
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 2334116
97
  },
98
  "cortex": {
99
+ "page": 73309,
100
+ "row": 69,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 9383621
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 275000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2af807d46a42c19f19c7b32b800e214d63c1433a37862b054d1767b88ee00e92", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8f706fccdd6b87209ff5215abd019c5c04c63cfa5c3c1d339b8f5d3a1bdf7fe4", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "ac15a69eb79c2ac9ef8406edc12d9ae8abafb561b0f74b54b23d5920ee4fe3f7", "shard_scheduler.json": "d4146d4949443885a89e71741ab422c603360627ea40e296539a0be0a218caee", "coverage.json": "8b908eef3d6b156adf6b83ba5c2f7263f4abb751ab7df3899af5bff2595659e7", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000275000"}
 
1
+ {"step": 280000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "b202beb2e692c902003e4624a58ed61b492689af5528e336e78027d172cdc6a5", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "086c04c0e606c24ef7dab8710d39cde6eff9e0a7ec67fc90f9930de401842100", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7ab839f3b8798329051e29b39e08bd08f8678f49ed79d09e3b70b9de78a4765c", "shard_scheduler.json": "bb41f80e95e1faea1ad1243a8e7d9ee991b7eb64bf0e6b9b463722b029efa994", "coverage.json": "c7677efe249bc40dea701da173ea31fe609c61d934d175645935757dc9bffe1c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000280000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2af807d46a42c19f19c7b32b800e214d63c1433a37862b054d1767b88ee00e92
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b202beb2e692c902003e4624a58ed61b492689af5528e336e78027d172cdc6a5
3
  size 360622744
performance.json CHANGED
@@ -49,11 +49,11 @@
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
- "microbatch_seconds": 0.8988038599999868,
53
  "peak_gb": 85.39913082122803,
54
  "fits_budget": true,
55
- "estimated_update_seconds": 0.9552281219999941,
56
- "input_tokens_per_second": 195958.21495470946
57
  },
58
  "measurements": [
59
  {
 
49
  "selected": {
50
  "microbatch": 86,
51
  "mode": "compiled",
52
+ "microbatch_seconds": 0.8973423759999832,
53
  "peak_gb": 85.39913082122803,
54
  "fits_budget": true,
55
+ "estimated_update_seconds": 0.9528963840000415,
56
+ "input_tokens_per_second": 196277.36827175462
57
  },
58
  "measurements": [
59
  {
shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 75917.17339470465,
26
- "required_pages_per_second_upper_estimate": 0.006580308179536131
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9489.64667433808,
40
- "required_pages_per_second_upper_estimate": 0.0037855619412550186
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 18979.29334867616,
54
- "required_pages_per_second_upper_estimate": 0.004390525492061635
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28468.940023014242,
68
- "required_pages_per_second_upper_estimate": 0.001733609082416801
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28468.940023014242,
82
- "required_pages_per_second_upper_estimate": 0.002191617810041497
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28468.940023014242,
96
- "required_pages_per_second_upper_estimate": 0.35348895567272487
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,410 +158,38 @@
158
  "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 690670,
161
- "producer_batch_seconds": 0.10167211079664185,
162
- "producer_p95_seconds": 0.38399030399887124,
163
- "gpu_update_seconds": 0.9711636603838922,
164
- "producer_batches_per_second": 9.835538892274373,
165
- "gpu_batches_per_second": 1.0296925644898092,
166
- "producer_headroom": 9.551917952469312,
167
  "sustainable": true,
168
- "buffer_seconds": 60.212146943801315,
169
  "buffer_drain_seconds": null,
170
- "shard_target_mib": 16,
171
- "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 298328.0392779603,
174
- "upload_bytes_per_second": 5292592.716316272,
175
- "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 64.01425250772579,
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": 1.5463005599995086,
184
- "bytes": 13315083,
185
- "pages": 40,
186
- "shards": 4
187
- },
188
- {
189
- "seconds": 1.3599293949991988,
190
- "bytes": 4987866,
191
- "pages": 33,
192
- "shards": 3
193
- },
194
- {
195
- "seconds": 1.145182983000268,
196
- "bytes": 6203180,
197
- "pages": 32,
198
- "shards": 2
199
- },
200
- {
201
- "seconds": 0.9317472719994839,
202
- "bytes": 2642711,
203
- "pages": 36,
204
- "shards": 2
205
- },
206
- {
207
- "seconds": 0.9964947309999843,
208
- "bytes": 5697842,
209
- "pages": 28,
210
- "shards": 2
211
- },
212
- {
213
- "seconds": 1.1368190850007522,
214
- "bytes": 10318857,
215
- "pages": 37,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 1.0222451929985255,
220
- "bytes": 8375215,
221
- "pages": 35,
222
- "shards": 3
223
- },
224
- {
225
- "seconds": 0.9998172479972709,
226
- "bytes": 6024653,
227
- "pages": 36,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 0.9568293169977551,
232
- "bytes": 2417450,
233
- "pages": 29,
234
- "shards": 2
235
- },
236
- {
237
- "seconds": 1.2115873000002466,
238
- "bytes": 10372111,
239
- "pages": 33,
240
- "shards": 3
241
- },
242
- {
243
- "seconds": 1.1448558080010116,
244
- "bytes": 6260573,
245
- "pages": 36,
246
- "shards": 2
247
- },
248
- {
249
- "seconds": 1.1358172180007386,
250
- "bytes": 2620953,
251
- "pages": 36,
252
- "shards": 2
253
- },
254
- {
255
- "seconds": 1.4711359560023993,
256
- "bytes": 11859463,
257
- "pages": 29,
258
- "shards": 3
259
- },
260
- {
261
- "seconds": 1.2465269989988883,
262
- "bytes": 5535585,
263
- "pages": 33,
264
- "shards": 2
265
- },
266
- {
267
- "seconds": 1.2969893279987446,
268
- "bytes": 7578437,
269
- "pages": 39,
270
- "shards": 3
271
- },
272
- {
273
- "seconds": 1.5002685959989321,
274
- "bytes": 6193769,
275
- "pages": 27,
276
- "shards": 2
277
- },
278
- {
279
- "seconds": 1.1753300150012365,
280
- "bytes": 2646901,
281
- "pages": 37,
282
- "shards": 2
283
- },
284
- {
285
- "seconds": 1.4061265680029464,
286
- "bytes": 9691299,
287
- "pages": 28,
288
- "shards": 3
289
- },
290
- {
291
- "seconds": 1.0368577580011333,
292
- "bytes": 6834427,
293
- "pages": 31,
294
- "shards": 2
295
- },
296
- {
297
- "seconds": 1.4560243180021644,
298
- "bytes": 7277270,
299
- "pages": 30,
300
- "shards": 3
301
- },
302
- {
303
- "seconds": 1.4158115079990239,
304
- "bytes": 6180346,
305
- "pages": 38,
306
- "shards": 2
307
- },
308
- {
309
- "seconds": 1.5717087800003355,
310
- "bytes": 560796,
311
- "pages": 18,
312
- "shards": 1
313
- },
314
- {
315
- "seconds": 4.375384323997423,
316
- "bytes": 11988381,
317
- "pages": 43,
318
- "shards": 4
319
- },
320
- {
321
- "seconds": 0.9212507589982124,
322
- "bytes": 677495,
323
- "pages": 18,
324
- "shards": 1
325
- },
326
- {
327
- "seconds": 0.8941036469987012,
328
- "bytes": 796711,
329
- "pages": 25,
330
  "shards": 1
331
  },
332
  {
333
- "seconds": 1.6776708719989983,
334
- "bytes": 13189069,
335
- "pages": 31,
336
- "shards": 4
337
- },
338
- {
339
- "seconds": 1.216554687001917,
340
- "bytes": 10409429,
341
- "pages": 33,
342
- "shards": 3
343
- },
344
- {
345
- "seconds": 1.1135386019959697,
346
- "bytes": 1273006,
347
- "pages": 34,
348
- "shards": 1
349
- },
350
- {
351
- "seconds": 1.2612176139955409,
352
- "bytes": 7427910,
353
- "pages": 31,
354
- "shards": 3
355
- },
356
- {
357
- "seconds": 0.7726261329953559,
358
- "bytes": 1134173,
359
- "pages": 35,
360
- "shards": 1
361
- },
362
- {
363
- "seconds": 1.5038225289972615,
364
- "bytes": 10256540,
365
- "pages": 38,
366
- "shards": 3
367
- },
368
- {
369
- "seconds": 1.727812716999324,
370
- "bytes": 13149585,
371
- "pages": 36,
372
- "shards": 4
373
- },
374
- {
375
- "seconds": 0.9765169189995504,
376
- "bytes": 821604,
377
  "pages": 27,
378
  "shards": 1
379
- },
380
- {
381
- "seconds": 1.1198572450011852,
382
- "bytes": 7695575,
383
- "pages": 38,
384
- "shards": 4
385
- },
386
- {
387
- "seconds": 1.1846314020003774,
388
- "bytes": 10317171,
389
- "pages": 36,
390
- "shards": 3
391
- },
392
- {
393
- "seconds": 1.0667465840015211,
394
- "bytes": 820608,
395
- "pages": 26,
396
- "shards": 1
397
- },
398
- {
399
- "seconds": 1.5008300169938593,
400
- "bytes": 7497655,
401
- "pages": 39,
402
- "shards": 3
403
- },
404
- {
405
- "seconds": 1.085039866004081,
406
- "bytes": 745159,
407
- "pages": 23,
408
- "shards": 1
409
- },
410
- {
411
- "seconds": 2.1224419440040947,
412
- "bytes": 16141114,
413
- "pages": 38,
414
- "shards": 4
415
- },
416
- {
417
- "seconds": 1.301262590000988,
418
- "bytes": 7677650,
419
- "pages": 39,
420
- "shards": 3
421
- },
422
- {
423
- "seconds": 0.976449344001594,
424
- "bytes": 1139759,
425
- "pages": 35,
426
- "shards": 1
427
- },
428
- {
429
- "seconds": 1.2381579889988643,
430
- "bytes": 7253317,
431
- "pages": 31,
432
- "shards": 3
433
- },
434
- {
435
- "seconds": 1.0107035210021422,
436
- "bytes": 4758637,
437
- "pages": 36,
438
- "shards": 2
439
- },
440
- {
441
- "seconds": 1.040413394002826,
442
- "bytes": 5884125,
443
- "pages": 26,
444
- "shards": 2
445
- },
446
- {
447
- "seconds": 1.2434632910008077,
448
- "bytes": 13606244,
449
- "pages": 40,
450
- "shards": 4
451
- },
452
- {
453
- "seconds": 0.7609897279980942,
454
- "bytes": 952904,
455
- "pages": 26,
456
- "shards": 1
457
- },
458
- {
459
- "seconds": 1.0394860849992256,
460
- "bytes": 10827416,
461
- "pages": 40,
462
- "shards": 3
463
- },
464
- {
465
- "seconds": 0.9459326559954206,
466
- "bytes": 2314805,
467
- "pages": 28,
468
- "shards": 2
469
- },
470
- {
471
- "seconds": 1.2841032439973787,
472
- "bytes": 6126814,
473
- "pages": 38,
474
- "shards": 2
475
- },
476
- {
477
- "seconds": 1.3457263079981203,
478
- "bytes": 7394502,
479
- "pages": 29,
480
- "shards": 3
481
- },
482
- {
483
- "seconds": 1.2961700559972087,
484
- "bytes": 10630748,
485
- "pages": 31,
486
- "shards": 3
487
- },
488
- {
489
- "seconds": 1.0071994079989963,
490
- "bytes": 6082322,
491
- "pages": 37,
492
- "shards": 2
493
- },
494
- {
495
- "seconds": 0.977721251998446,
496
- "bytes": 2308590,
497
- "pages": 28,
498
- "shards": 2
499
- },
500
- {
501
- "seconds": 1.0036859399988316,
502
- "bytes": 6116236,
503
- "pages": 36,
504
- "shards": 2
505
- },
506
- {
507
- "seconds": 0.9474672509968514,
508
- "bytes": 5614078,
509
- "pages": 31,
510
- "shards": 2
511
- },
512
- {
513
- "seconds": 1.196396844003175,
514
- "bytes": 7831133,
515
- "pages": 38,
516
- "shards": 3
517
- },
518
- {
519
- "seconds": 1.1607214789983118,
520
- "bytes": 11917502,
521
- "pages": 33,
522
- "shards": 3
523
- },
524
- {
525
- "seconds": 1.2743855959997745,
526
- "bytes": 2697194,
527
- "pages": 36,
528
- "shards": 2
529
- },
530
- {
531
- "seconds": 1.4709027479984798,
532
- "bytes": 10099926,
533
- "pages": 34,
534
- "shards": 3
535
- },
536
- {
537
- "seconds": 0.8219810020018485,
538
- "bytes": 940002,
539
- "pages": 29,
540
- "shards": 1
541
- },
542
- {
543
- "seconds": 1.2411883530003252,
544
- "bytes": 7506339,
545
- "pages": 35,
546
- "shards": 3
547
- },
548
- {
549
- "seconds": 1.1381002670023008,
550
- "bytes": 6063351,
551
- "pages": 37,
552
- "shards": 2
553
- },
554
- {
555
- "seconds": 1.1012217039969983,
556
- "bytes": 10763585,
557
- "pages": 27,
558
- "shards": 3
559
- },
560
- {
561
- "seconds": 1.2329660859977594,
562
- "bytes": 7594797,
563
- "pages": 31,
564
- "shards": 3
565
  }
566
  ],
567
  "cache_key": {
@@ -606,7 +234,7 @@
606
  },
607
  "cache_reused": true,
608
  "resume_probe_seconds": [
609
- 2.844574209999962,
610
- 0.2503580399999805
611
  ]
612
  }
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 75842.57462698415,
26
+ "required_pages_per_second_upper_estimate": 0.006573842147418966
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9480.321828373018,
40
+ "required_pages_per_second_upper_estimate": 0.0037818421207806837
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 18960.643656746037,
54
+ "required_pages_per_second_upper_estimate": 0.0043862112140569684
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28440.965485119053,
68
+ "required_pages_per_second_upper_estimate": 0.0017319055798300411
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28440.965485119053,
82
+ "required_pages_per_second_upper_estimate": 0.002189464252675848
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28440.965485119053,
96
+ "required_pages_per_second_upper_estimate": 0.3531416055368222
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": 690670,
161
+ "producer_batch_seconds": 0.080577971249987,
162
+ "producer_p95_seconds": 0.18257642799926543,
163
+ "gpu_update_seconds": 0.9721188971051651,
164
+ "producer_batches_per_second": 12.410339755236285,
165
+ "gpu_batches_per_second": 1.028680753946725,
166
+ "producer_headroom": 12.06432579556068,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.271371620520235,
169
  "buffer_drain_seconds": null,
170
+ "shard_target_mib": 32,
171
+ "upload_pages_per_commit": 160,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 905156.4916824569,
174
+ "upload_bytes_per_second": 607793.4295288165,
175
+ "upload_sustainable": false,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 97.11275607516899,
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": 0.8291290090000984,
184
+ "bytes": 27860,
185
+ "pages": 1,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
186
  "shards": 1
187
  },
188
  {
189
+ "seconds": 0.7566531879992908,
190
+ "bytes": 935968,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
191
  "pages": 27,
192
  "shards": 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
193
  }
194
  ],
195
  "cache_key": {
 
234
  },
235
  "cache_reused": true,
236
  "resume_probe_seconds": [
237
+ 3.215140269000017,
238
+ 0.014588057999958437
239
  ]
240
  }
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8f706fccdd6b87209ff5215abd019c5c04c63cfa5c3c1d339b8f5d3a1bdf7fe4
3
- size 721785471
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:086c04c0e606c24ef7dab8710d39cde6eff9e0a7ec67fc90f9930de401842100
3
+ size 721763967