appvoid commited on
Commit
190e37e
·
verified ·
1 Parent(s): 1b6df39

Qwen90M step 440000; 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 +247 -247
  6. trainer.pt +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 435000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "bf19dfbdafa63e001aa9aafbc90b1803081ed56a7003a3405b4bd251b7f12fb6",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "ef0d375c6273462fad4461943f3ee97eb93fad33803aa4f396c84603803639c6",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
- "shard_scheduler.json": "8aeff81a2ed0c0c9f85d391f6a9d59dff3e5b2336e287e0f785392dac9c5cb4d",
15
- "coverage.json": "142a67e23a41b166cd9863617f189163392272c3e95892f505038ddd8f6114c9",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 440000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "fe214f4ab1fb9964cf107eac88ddf367929c1be6f58e91da736242cfa8027d31",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "8bb0a7558188a5dfb0a78d6ae95be1812ae85f92993c239c65c6dc7be7f1be7d",
12
  "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6",
13
  "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18",
14
+ "shard_scheduler.json": "6724984473eb995a25775ee3318e6a2730c21a7c2160f6311680bbf3c31a3993",
15
+ "coverage.json": "3bc0c290712aa33916e32c37c9e507b57b125caed3a356f046223639dd1963d8",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 435000,
3
  "stats": {
4
- "climbmix/blocks": 15660000,
5
- "climbmix/targets": 26178177994,
6
- "climbmix/input_tokens": 26179836274,
7
- "ultra_qa/blocks": 5872500,
8
- "ultra_qa/targets": 9579803583,
9
- "ultra_qa/input_tokens": 9579821954,
10
- "cortex/blocks": 5872500,
11
- "cortex/targets": 223336398,
12
- "cortex/input_tokens": 9335114674,
13
- "ultra_style/blocks": 5872500,
14
- "ultra_style/targets": 9288626928,
15
- "ultra_style/input_tokens": 9289113965,
16
- "openmath/blocks": 3915000,
17
- "openmath/targets": 6049599492,
18
- "openmath/input_tokens": 6051615394,
19
- "rewrite6/blocks": 1957500,
20
- "rewrite6/targets": 3209011309,
21
- "rewrite6/input_tokens": 3212529771
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 2144,
26
- "row": 2831,
27
  "start": {
28
- "file": 103,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 65,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 8784655
38
  },
39
  "rewrite6": {
40
- "page": 20,
41
- "row": 892,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 81920
47
  }
48
  },
49
  "epoch": 24,
50
- "offset": 2048,
51
- "documents": 5012508
52
  },
53
  "openmath": {
54
- "page": 1340,
55
- "row": 1711,
56
  "start": {
57
  "file": 12,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 248,
61
- "row": 840
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 5490351
67
  },
68
  "ultra_qa": {
69
- "page": 579,
70
- "row": 2175,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
- "row": 34798
77
  }
78
  },
79
  "epoch": 0,
80
- "offset": 2048,
81
- "documents": 2373759
82
  },
83
  "ultra_style": {
84
- "page": 884,
85
- "row": 2600,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
- "row": 25169
92
  }
93
  },
94
  "epoch": 0,
95
- "offset": 0,
96
- "documents": 3623464
97
  },
98
  "cortex": {
99
- "page": 113898,
100
- "row": 103,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 14579047
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 440000,
3
  "stats": {
4
+ "climbmix/blocks": 15840000,
5
+ "climbmix/targets": 26479322066,
6
+ "climbmix/input_tokens": 26480999183,
7
+ "ultra_qa/blocks": 5940000,
8
+ "ultra_qa/targets": 9689929802,
9
+ "ultra_qa/input_tokens": 9689948398,
10
+ "cortex/blocks": 5940000,
11
+ "cortex/targets": 225912648,
12
+ "cortex/input_tokens": 9442402042,
13
+ "ultra_style/blocks": 5940000,
14
+ "ultra_style/targets": 9395030681,
15
+ "ultra_style/input_tokens": 9395523198,
16
+ "openmath/blocks": 3960000,
17
+ "openmath/targets": 6119068769,
18
+ "openmath/input_tokens": 6121107744,
19
+ "rewrite6/blocks": 1980000,
20
+ "rewrite6/targets": 3245883112,
21
+ "rewrite6/input_tokens": 3249442073
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 2169,
26
+ "row": 730,
27
  "start": {
28
+ "file": 105,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 0,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 8884954
38
  },
39
  "rewrite6": {
40
+ "page": 34,
41
+ "row": 1296,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 139264
47
  }
48
  },
49
  "epoch": 24,
50
+ "offset": 0,
51
+ "documents": 5070256
52
  },
53
  "openmath": {
54
+ "page": 1355,
55
+ "row": 3306,
56
  "start": {
57
  "file": 12,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 310,
61
+ "row": 280
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 5553386
67
  },
68
  "ultra_qa": {
69
+ "page": 586,
70
+ "row": 858,
71
  "start": {
72
  "file": 4,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 4,
76
+ "row": 63470
77
  }
78
  },
79
  "epoch": 0,
80
+ "offset": 4096,
81
+ "documents": 2401114
82
  },
83
  "ultra_style": {
84
+ "page": 894,
85
+ "row": 3292,
86
  "start": {
87
  "file": 5,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
  "row_group": 2,
91
+ "row": 66129
92
  }
93
  },
94
  "epoch": 0,
95
+ "offset": 2048,
96
+ "documents": 3665116
97
  },
98
  "cortex": {
99
+ "page": 115207,
100
+ "row": 41,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 14746537
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 435000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "bf19dfbdafa63e001aa9aafbc90b1803081ed56a7003a3405b4bd251b7f12fb6", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "ef0d375c6273462fad4461943f3ee97eb93fad33803aa4f396c84603803639c6", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "8aeff81a2ed0c0c9f85d391f6a9d59dff3e5b2336e287e0f785392dac9c5cb4d", "coverage.json": "142a67e23a41b166cd9863617f189163392272c3e95892f505038ddd8f6114c9", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000435000"}
 
1
+ {"step": 440000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "fe214f4ab1fb9964cf107eac88ddf367929c1be6f58e91da736242cfa8027d31", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "8bb0a7558188a5dfb0a78d6ae95be1812ae85f92993c239c65c6dc7be7f1be7d", "training_config.json": "c8bdf052510c2d9092e5a6824b119dd6584dc7697f8c4a8e9b7ecf3057c5bbb6", "performance.json": "a6ae17d5ca4d76a29c6e069dddb352125dfbf6f18e59e9553ab94c741a962b18", "shard_scheduler.json": "6724984473eb995a25775ee3318e6a2730c21a7c2160f6311680bbf3c31a3993", "coverage.json": "3bc0c290712aa33916e32c37c9e507b57b125caed3a356f046223639dd1963d8", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000440000"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bf19dfbdafa63e001aa9aafbc90b1803081ed56a7003a3405b4bd251b7f12fb6
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe214f4ab1fb9964cf107eac88ddf367929c1be6f58e91da736242cfa8027d31
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": 77090.63733768974,
26
- "required_pages_per_second_upper_estimate": 0.005333309165944155
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": 9636.329667211217,
40
- "required_pages_per_second_upper_estimate": 0.004034530002215322
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": 19272.659334422435,
54
- "required_pages_per_second_upper_estimate": 0.004013986685154929
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": 28908.989001633654,
68
- "required_pages_per_second_upper_estimate": 0.0017330497491776062
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": 28908.989001633654,
82
- "required_pages_per_second_upper_estimate": 0.002247546737672786
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": 28908.989001633654,
96
- "required_pages_per_second_upper_estimate": 0.37081347085893784
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": 704566,
161
- "producer_batch_seconds": 0.14530138408622406,
162
- "producer_p95_seconds": 0.3642332669987809,
163
- "gpu_update_seconds": 0.9563807298289679,
164
- "producer_batches_per_second": 6.882246898670867,
165
- "gpu_batches_per_second": 1.0456086878484394,
166
- "producer_headroom": 6.5820483118139945,
167
  "sustainable": true,
168
- "buffer_seconds": 60.25198597922498,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 434186.3837777914,
174
- "upload_bytes_per_second": 5398305.177243957,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 93.16629061699206,
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": 1.1164300489981542,
184
- "bytes": 9840833,
185
  "pages": 29,
186
- "shards": 3
187
  },
188
  {
189
- "seconds": 1.1696946029987885,
190
- "bytes": 13219787,
191
- "pages": 36,
192
- "shards": 4
193
  },
194
  {
195
- "seconds": 0.8621246889997565,
196
- "bytes": 1265428,
197
- "pages": 35,
198
- "shards": 1
199
  },
200
  {
201
- "seconds": 1.4639245990001655,
202
- "bytes": 10067426,
203
- "pages": 30,
204
- "shards": 3
205
  },
206
  {
207
- "seconds": 0.8352084129983268,
208
- "bytes": 2708130,
209
- "pages": 39,
210
  "shards": 2
211
  },
212
  {
213
- "seconds": 0.9732167729998764,
214
- "bytes": 6207933,
215
- "pages": 37,
 
 
 
 
 
 
216
  "shards": 2
217
  },
218
  {
219
- "seconds": 0.9747246619990619,
220
- "bytes": 7297014,
221
- "pages": 31,
222
- "shards": 4
223
  },
224
  {
225
- "seconds": 1.0759720290006953,
226
- "bytes": 6305836,
227
- "pages": 19,
228
  "shards": 2
229
  },
230
  {
231
- "seconds": 1.0651401020004414,
232
- "bytes": 11486452,
233
- "pages": 47,
234
  "shards": 3
235
  },
236
  {
237
- "seconds": 1.0763056159994449,
238
- "bytes": 2441249,
239
- "pages": 29,
240
- "shards": 2
241
  },
242
  {
243
- "seconds": 4.102204398001049,
244
- "bytes": 6262944,
245
- "pages": 38,
 
 
 
 
 
 
246
  "shards": 2
247
  },
248
  {
249
- "seconds": 1.1364409210000304,
250
- "bytes": 7535674,
251
- "pages": 31,
252
  "shards": 3
253
  },
254
  {
255
- "seconds": 0.9002165780002542,
256
- "bytes": 5078039,
257
- "pages": 38,
258
  "shards": 2
259
  },
260
  {
261
- "seconds": 1.1330570440004522,
262
- "bytes": 12010212,
263
  "pages": 32,
264
- "shards": 3
265
  },
266
  {
267
- "seconds": 1.1789866869985417,
268
- "bytes": 2676190,
269
  "pages": 38,
270
  "shards": 2
271
  },
272
  {
273
- "seconds": 1.2169118900019384,
274
- "bytes": 10000231,
275
- "pages": 33,
276
- "shards": 3
277
  },
278
  {
279
- "seconds": 1.287032372001704,
280
- "bytes": 6056411,
281
- "pages": 32,
282
- "shards": 2
283
  },
284
  {
285
- "seconds": 1.0081800109983305,
286
- "bytes": 2487942,
287
- "pages": 34,
288
- "shards": 2
289
  },
290
  {
291
- "seconds": 1.1110187749982288,
292
- "bytes": 6282739,
293
- "pages": 38,
294
  "shards": 2
295
  },
296
  {
297
- "seconds": 1.3052810369990766,
298
- "bytes": 12708434,
299
- "pages": 30,
300
  "shards": 4
301
  },
302
  {
303
- "seconds": 1.1760405560016807,
304
- "bytes": 6277751,
305
- "pages": 38,
306
  "shards": 2
307
  },
308
  {
309
- "seconds": 1.038220692997129,
310
- "bytes": 6391937,
311
  "pages": 35,
312
- "shards": 2
313
  },
314
  {
315
- "seconds": 0.970617956998467,
316
- "bytes": 2311552,
317
- "pages": 30,
318
- "shards": 2
319
  },
320
  {
321
- "seconds": 1.171514493998984,
322
- "bytes": 9915176,
323
- "pages": 34,
324
  "shards": 3
325
  },
326
  {
327
- "seconds": 3.413111547000881,
328
- "bytes": 7579762,
329
- "pages": 40,
330
  "shards": 3
331
  },
332
  {
333
- "seconds": 1.0360288759984542,
334
- "bytes": 6839680,
335
- "pages": 32,
336
- "shards": 2
337
  },
338
  {
339
- "seconds": 1.1501934409971,
340
- "bytes": 5967653,
341
- "pages": 29,
342
- "shards": 2
343
  },
344
  {
345
- "seconds": 1.0980839240000932,
346
- "bytes": 6693759,
347
  "pages": 35,
 
 
 
 
 
 
348
  "shards": 3
349
  },
350
  {
351
- "seconds": 1.258701953003765,
352
- "bytes": 6452551,
353
- "pages": 37,
354
- "shards": 2
355
  },
356
  {
357
- "seconds": 3.7520576210008585,
358
- "bytes": 2441375,
359
  "pages": 28,
360
- "shards": 2
361
  },
362
  {
363
- "seconds": 1.106896608005627,
364
- "bytes": 5997924,
365
- "pages": 35,
366
- "shards": 2
367
  },
368
  {
369
- "seconds": 1.6541905339981895,
370
- "bytes": 16581039,
371
- "pages": 37,
372
- "shards": 4
373
  },
374
  {
375
- "seconds": 1.211611481005093,
376
- "bytes": 2609258,
377
- "pages": 36,
378
  "shards": 2
379
  },
380
  {
381
- "seconds": 1.0486429719967418,
382
- "bytes": 6039932,
383
- "pages": 37,
384
- "shards": 2
385
  },
386
  {
387
- "seconds": 1.093474820998381,
388
- "bytes": 5981075,
389
- "pages": 28,
390
- "shards": 2
391
  },
392
  {
393
- "seconds": 1.4037001169999712,
394
- "bytes": 6358641,
395
- "pages": 36,
396
  "shards": 3
397
  },
398
  {
399
- "seconds": 1.1129142319987295,
400
- "bytes": 6226935,
401
  "pages": 37,
402
- "shards": 2
403
  },
404
  {
405
- "seconds": 1.6370596409979044,
406
- "bytes": 13312554,
 
 
 
 
 
 
407
  "pages": 36,
408
- "shards": 5
409
  },
410
  {
411
- "seconds": 0.9763937130046543,
412
- "bytes": 4930611,
413
- "pages": 30,
414
- "shards": 2
415
  },
416
  {
417
- "seconds": 1.5403500989996246,
418
- "bytes": 6045634,
419
- "pages": 38,
420
  "shards": 2
421
  },
422
  {
423
- "seconds": 0.8855937030020868,
424
- "bytes": 2335409,
425
- "pages": 29,
426
- "shards": 2
427
  },
428
  {
429
- "seconds": 1.0506481749980594,
430
- "bytes": 6177700,
431
- "pages": 37,
432
  "shards": 2
433
  },
434
  {
435
- "seconds": 1.1661089699991862,
436
- "bytes": 11581137,
437
- "pages": 34,
438
  "shards": 4
439
  },
440
  {
441
- "seconds": 1.0408890239996254,
442
- "bytes": 7081099,
443
- "pages": 36,
444
- "shards": 2
445
  },
446
  {
447
- "seconds": 1.0998482240029261,
448
- "bytes": 5731633,
449
- "pages": 29,
450
  "shards": 2
451
  },
452
  {
453
- "seconds": 1.2273955099954037,
454
- "bytes": 7571144,
455
- "pages": 41,
456
- "shards": 3
457
  },
458
  {
459
- "seconds": 1.115527177003969,
460
- "bytes": 5389343,
461
- "pages": 30,
462
- "shards": 2
463
  },
464
  {
465
- "seconds": 1.2195735799978138,
466
- "bytes": 6166768,
467
- "pages": 38,
468
- "shards": 2
469
  },
470
  {
471
- "seconds": 0.954921390999516,
472
- "bytes": 2330284,
473
- "pages": 28,
474
  "shards": 2
475
  },
476
  {
477
- "seconds": 1.4671854679982061,
478
- "bytes": 11708465,
479
- "pages": 38,
480
  "shards": 3
481
  },
482
  {
483
- "seconds": 1.1728415899997344,
484
- "bytes": 11022627,
485
- "pages": 35,
486
- "shards": 4
487
  },
488
  {
489
- "seconds": 0.9499405899987323,
490
- "bytes": 1172384,
491
  "pages": 36,
492
  "shards": 1
493
  },
494
  {
495
- "seconds": 1.1931303429955733,
496
- "bytes": 6041594,
497
  "pages": 29,
498
  "shards": 2
499
  },
500
  {
501
- "seconds": 1.0443037180011743,
502
- "bytes": 7667563,
503
- "pages": 39,
504
  "shards": 3
505
  },
506
  {
507
- "seconds": 1.1781920779976645,
508
- "bytes": 5828706,
509
- "pages": 34,
510
- "shards": 2
511
- },
512
- {
513
- "seconds": 1.4654938949970528,
514
- "bytes": 13087295,
515
- "pages": 32,
516
  "shards": 4
517
  },
518
  {
519
- "seconds": 1.2222818860027473,
520
- "bytes": 6269271,
521
- "pages": 38,
522
  "shards": 2
523
  },
524
  {
525
- "seconds": 1.2005412010039436,
526
- "bytes": 962816,
527
- "pages": 30,
528
- "shards": 1
529
- },
530
- {
531
- "seconds": 1.2577394740001182,
532
- "bytes": 11361644,
533
- "pages": 34,
534
- "shards": 4
535
- },
536
- {
537
- "seconds": 0.925975751997612,
538
- "bytes": 1221763,
539
- "pages": 37,
540
- "shards": 1
541
  },
542
  {
543
- "seconds": 1.0596703449991765,
544
- "bytes": 5842037,
545
- "pages": 31,
546
  "shards": 2
547
  },
548
  {
549
- "seconds": 1.226734026997292,
550
- "bytes": 13551509,
551
- "pages": 37,
552
- "shards": 4
553
  },
554
  {
555
- "seconds": 1.065682766995451,
556
- "bytes": 5277203,
557
- "pages": 35,
558
  "shards": 2
559
- },
560
- {
561
- "seconds": 1.4707806820006226,
562
- "bytes": 7704301,
563
- "pages": 35,
564
- "shards": 3
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": 76924.88308471063,
26
+ "required_pages_per_second_upper_estimate": 0.005321841902120205
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": 9615.610385588829,
40
+ "required_pages_per_second_upper_estimate": 0.004025855271667829
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": 19231.220771177657,
54
+ "required_pages_per_second_upper_estimate": 0.004005356125239444
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": 28846.831156766486,
68
+ "required_pages_per_second_upper_estimate": 0.0017293234812873531
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": 28846.831156766486,
82
+ "required_pages_per_second_upper_estimate": 0.002242714238637829
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": 28846.831156766486,
96
+ "required_pages_per_second_upper_estimate": 0.37001617676487586
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": 704566,
161
+ "producer_batch_seconds": 0.5073274579607983,
162
+ "producer_p95_seconds": 0.34248151299834717,
163
+ "gpu_update_seconds": 0.9584414956966502,
164
+ "producer_batches_per_second": 1.9711134974233369,
165
+ "gpu_batches_per_second": 1.0433605019085102,
166
+ "producer_headroom": 1.8891969686582784,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.38181422888896,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 78,
172
  "max_pending_pages": 4096,
173
+ "generation_bytes_per_second": 270603.4737486303,
174
+ "upload_bytes_per_second": 5111100.11035178,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 58.06520613998587,
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": 0.8855937030020868,
184
+ "bytes": 2335409,
185
  "pages": 29,
186
+ "shards": 2
187
  },
188
  {
189
+ "seconds": 1.0506481749980594,
190
+ "bytes": 6177700,
191
+ "pages": 37,
192
+ "shards": 2
193
  },
194
  {
195
+ "seconds": 1.1661089699991862,
196
+ "bytes": 11581137,
197
+ "pages": 34,
198
+ "shards": 4
199
  },
200
  {
201
+ "seconds": 1.0408890239996254,
202
+ "bytes": 7081099,
203
+ "pages": 36,
204
+ "shards": 2
205
  },
206
  {
207
+ "seconds": 1.0998482240029261,
208
+ "bytes": 5731633,
209
+ "pages": 29,
210
  "shards": 2
211
  },
212
  {
213
+ "seconds": 1.2273955099954037,
214
+ "bytes": 7571144,
215
+ "pages": 41,
216
+ "shards": 3
217
+ },
218
+ {
219
+ "seconds": 1.115527177003969,
220
+ "bytes": 5389343,
221
+ "pages": 30,
222
  "shards": 2
223
  },
224
  {
225
+ "seconds": 1.2195735799978138,
226
+ "bytes": 6166768,
227
+ "pages": 38,
228
+ "shards": 2
229
  },
230
  {
231
+ "seconds": 0.954921390999516,
232
+ "bytes": 2330284,
233
+ "pages": 28,
234
  "shards": 2
235
  },
236
  {
237
+ "seconds": 1.4671854679982061,
238
+ "bytes": 11708465,
239
+ "pages": 38,
240
  "shards": 3
241
  },
242
  {
243
+ "seconds": 1.1728415899997344,
244
+ "bytes": 11022627,
245
+ "pages": 35,
246
+ "shards": 4
247
  },
248
  {
249
+ "seconds": 0.9499405899987323,
250
+ "bytes": 1172384,
251
+ "pages": 36,
252
+ "shards": 1
253
+ },
254
+ {
255
+ "seconds": 1.1931303429955733,
256
+ "bytes": 6041594,
257
+ "pages": 29,
258
  "shards": 2
259
  },
260
  {
261
+ "seconds": 1.0443037180011743,
262
+ "bytes": 7667563,
263
+ "pages": 39,
264
  "shards": 3
265
  },
266
  {
267
+ "seconds": 1.1781920779976645,
268
+ "bytes": 5828706,
269
+ "pages": 34,
270
  "shards": 2
271
  },
272
  {
273
+ "seconds": 1.4654938949970528,
274
+ "bytes": 13087295,
275
  "pages": 32,
276
+ "shards": 4
277
  },
278
  {
279
+ "seconds": 1.2222818860027473,
280
+ "bytes": 6269271,
281
  "pages": 38,
282
  "shards": 2
283
  },
284
  {
285
+ "seconds": 1.2005412010039436,
286
+ "bytes": 962816,
287
+ "pages": 30,
288
+ "shards": 1
289
  },
290
  {
291
+ "seconds": 1.2577394740001182,
292
+ "bytes": 11361644,
293
+ "pages": 34,
294
+ "shards": 4
295
  },
296
  {
297
+ "seconds": 0.925975751997612,
298
+ "bytes": 1221763,
299
+ "pages": 37,
300
+ "shards": 1
301
  },
302
  {
303
+ "seconds": 1.0596703449991765,
304
+ "bytes": 5842037,
305
+ "pages": 31,
306
  "shards": 2
307
  },
308
  {
309
+ "seconds": 1.226734026997292,
310
+ "bytes": 13551509,
311
+ "pages": 37,
312
  "shards": 4
313
  },
314
  {
315
+ "seconds": 1.065682766995451,
316
+ "bytes": 5277203,
317
+ "pages": 35,
318
  "shards": 2
319
  },
320
  {
321
+ "seconds": 1.4707806820006226,
322
+ "bytes": 7704301,
323
  "pages": 35,
324
+ "shards": 3
325
  },
326
  {
327
+ "seconds": 1.0301591320021544,
328
+ "bytes": 1143669,
329
+ "pages": 32,
330
+ "shards": 1
331
  },
332
  {
333
+ "seconds": 2.9000736189991585,
334
+ "bytes": 10986050,
335
+ "pages": 37,
336
  "shards": 3
337
  },
338
  {
339
+ "seconds": 1.3636105609984952,
340
+ "bytes": 7351661,
341
+ "pages": 31,
342
  "shards": 3
343
  },
344
  {
345
+ "seconds": 1.2681014190020505,
346
+ "bytes": 6981210,
347
+ "pages": 39,
348
+ "shards": 3
349
  },
350
  {
351
+ "seconds": 1.0006479639996542,
352
+ "bytes": 7455931,
353
+ "pages": 31,
354
+ "shards": 3
355
  },
356
  {
357
+ "seconds": 1.9897420429988415,
358
+ "bytes": 1194300,
359
  "pages": 35,
360
+ "shards": 1
361
+ },
362
+ {
363
+ "seconds": 1.1884745839997777,
364
+ "bytes": 9577292,
365
+ "pages": 31,
366
  "shards": 3
367
  },
368
  {
369
+ "seconds": 1.1139246599996113,
370
+ "bytes": 7438556,
371
+ "pages": 40,
372
+ "shards": 3
373
  },
374
  {
375
+ "seconds": 2.768625452001288,
376
+ "bytes": 924425,
377
  "pages": 28,
378
+ "shards": 1
379
  },
380
  {
381
+ "seconds": 1.3904317609994905,
382
+ "bytes": 18015559,
383
+ "pages": 40,
384
+ "shards": 5
385
  },
386
  {
387
+ "seconds": 0.9437920670025051,
388
+ "bytes": 911442,
389
+ "pages": 27,
390
+ "shards": 1
391
  },
392
  {
393
+ "seconds": 1.0681643999996595,
394
+ "bytes": 6436647,
395
+ "pages": 38,
396
  "shards": 2
397
  },
398
  {
399
+ "seconds": 1.4995007219986292,
400
+ "bytes": 7398571,
401
+ "pages": 31,
402
+ "shards": 3
403
  },
404
  {
405
+ "seconds": 1.022340563002217,
406
+ "bytes": 921929,
407
+ "pages": 27,
408
+ "shards": 1
409
  },
410
  {
411
+ "seconds": 1.1844563459962956,
412
+ "bytes": 10179020,
413
+ "pages": 46,
414
  "shards": 3
415
  },
416
  {
417
+ "seconds": 1.3179188790018088,
418
+ "bytes": 13975639,
419
  "pages": 37,
420
+ "shards": 4
421
  },
422
  {
423
+ "seconds": 1.3881760859949281,
424
+ "bytes": 920491,
425
+ "pages": 27,
426
+ "shards": 1
427
+ },
428
+ {
429
+ "seconds": 1.721910915999615,
430
+ "bytes": 12247927,
431
  "pages": 36,
432
+ "shards": 4
433
  },
434
  {
435
+ "seconds": 1.029901437999797,
436
+ "bytes": 1106234,
437
+ "pages": 32,
438
+ "shards": 1
439
  },
440
  {
441
+ "seconds": 0.968032641998434,
442
+ "bytes": 6353702,
443
+ "pages": 39,
444
  "shards": 2
445
  },
446
  {
447
+ "seconds": 1.354905072999827,
448
+ "bytes": 7189345,
449
+ "pages": 31,
450
+ "shards": 3
451
  },
452
  {
453
+ "seconds": 1.0060001309975632,
454
+ "bytes": 6436150,
455
+ "pages": 22,
456
  "shards": 2
457
  },
458
  {
459
+ "seconds": 2.564948939005262,
460
+ "bytes": 12088456,
461
+ "pages": 50,
462
  "shards": 4
463
  },
464
  {
465
+ "seconds": 0.9611534259965993,
466
+ "bytes": 862575,
467
+ "pages": 27,
468
+ "shards": 1
469
  },
470
  {
471
+ "seconds": 1.0102765199990245,
472
+ "bytes": 6370774,
473
+ "pages": 38,
474
  "shards": 2
475
  },
476
  {
477
+ "seconds": 1.36733477999951,
478
+ "bytes": 11825285,
479
+ "pages": 33,
480
+ "shards": 4
481
  },
482
  {
483
+ "seconds": 0.7963763439984177,
484
+ "bytes": 1200164,
485
+ "pages": 36,
486
+ "shards": 1
487
  },
488
  {
489
+ "seconds": 4.367263744999946,
490
+ "bytes": 11457575,
491
+ "pages": 30,
492
+ "shards": 3
493
  },
494
  {
495
+ "seconds": 0.8959607119977591,
496
+ "bytes": 2404108,
497
+ "pages": 30,
498
  "shards": 2
499
  },
500
  {
501
+ "seconds": 1.3112088260022574,
502
+ "bytes": 10449012,
503
+ "pages": 40,
504
  "shards": 3
505
  },
506
  {
507
+ "seconds": 1.3342199069957132,
508
+ "bytes": 7580590,
509
+ "pages": 31,
510
+ "shards": 3
511
  },
512
  {
513
+ "seconds": 1.0893147949973354,
514
+ "bytes": 1277647,
515
  "pages": 36,
516
  "shards": 1
517
  },
518
  {
519
+ "seconds": 1.0343618879996939,
520
+ "bytes": 5958933,
521
  "pages": 29,
522
  "shards": 2
523
  },
524
  {
525
+ "seconds": 1.130363664997276,
526
+ "bytes": 7892634,
527
+ "pages": 23,
528
  "shards": 3
529
  },
530
  {
531
+ "seconds": 1.3904960489962832,
532
+ "bytes": 10986012,
533
+ "pages": 47,
 
 
 
 
 
 
534
  "shards": 4
535
  },
536
  {
537
+ "seconds": 1.050227845000336,
538
+ "bytes": 5898529,
539
+ "pages": 29,
540
  "shards": 2
541
  },
542
  {
543
+ "seconds": 0.9206480690045282,
544
+ "bytes": 2670829,
545
+ "pages": 39,
546
+ "shards": 2
 
 
 
 
 
 
 
 
 
 
 
 
547
  },
548
  {
549
+ "seconds": 1.0737715180002851,
550
+ "bytes": 5694268,
551
+ "pages": 23,
552
  "shards": 2
553
  },
554
  {
555
+ "seconds": 1.1375886099995114,
556
+ "bytes": 7292178,
557
+ "pages": 43,
558
+ "shards": 3
559
  },
560
  {
561
+ "seconds": 1.6739402300008805,
562
+ "bytes": 5561171,
563
+ "pages": 22,
564
  "shards": 2
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
trainer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ef0d375c6273462fad4461943f3ee97eb93fad33803aa4f396c84603803639c6
3
- size 721768511
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bb0a7558188a5dfb0a78d6ae95be1812ae85f92993c239c65c6dc7be7f1be7d
3
+ size 721770431