appvoid commited on
Commit
e9988a2
·
verified ·
1 Parent(s): 0a8a17d

Qwen90M step 237163; complete resumable checkpoint

Browse files
Files changed (7) hide show
  1. complete.json +6 -6
  2. coverage.json +44 -44
  3. latest.json +1 -1
  4. model.safetensors +1 -1
  5. shard_scheduler.json +133 -133
  6. trainer.pt +2 -2
  7. training_config.json +2 -2
complete.json CHANGED
@@ -1,18 +1,18 @@
1
  {
2
- "step": 235000,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
- "model.safetensors": "2e35e5395cb287d1981044a5822eddab4ba9c62bc411de4f902b706cfb265591",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
- "trainer.pt": "bdfb4bcc1f514f4993f7557ffa73c6c78e0fbe0f7eae146e44f89fbac2919cdd",
12
- "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf",
13
  "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
14
- "shard_scheduler.json": "b38e01cf77934ef444db8086750e69ecf4d69bd7e1c45ee0fec8d3051a6abc4d",
15
- "coverage.json": "d2f27535dad5eb059b3164dfa9e026883a87ba1676e6c7dc163b3f5eb36c7ad5",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
 
1
  {
2
+ "step": 237163,
3
  "files": {
4
  "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
5
  "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
6
+ "model.safetensors": "2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8",
7
  "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
8
  "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
9
  "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
10
  "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
11
+ "trainer.pt": "3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39",
12
+ "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783",
13
  "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522",
14
+ "shard_scheduler.json": "7642e374dd789aa75440578f87a3db0e21df171d6322760efedf97e6eb520c03",
15
+ "coverage.json": "70e3c6c517ccfd837f7cb477fd294e1034ff341fbf2983c3135f90eb9c011c05",
16
  "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
17
  },
18
  "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063"
coverage.json CHANGED
@@ -1,107 +1,107 @@
1
  {
2
- "step": 235000,
3
  "stats": {
4
- "climbmix/blocks": 8460000,
5
- "climbmix/targets": 14142088161,
6
- "climbmix/input_tokens": 14142984336,
7
- "ultra_qa/blocks": 3172500,
8
- "ultra_qa/targets": 5175699264,
9
- "ultra_qa/input_tokens": 5175709288,
10
- "cortex/blocks": 3172500,
11
- "cortex/targets": 120690717,
12
- "cortex/input_tokens": 5042698249,
13
- "ultra_style/blocks": 3172500,
14
- "ultra_style/targets": 5018546024,
15
- "ultra_style/input_tokens": 5018809138,
16
- "openmath/blocks": 2115000,
17
- "openmath/targets": 3267999434,
18
- "openmath/input_tokens": 3269088966,
19
- "rewrite6/blocks": 1057500,
20
- "rewrite6/targets": 1733635352,
21
- "rewrite6/input_tokens": 1735535955
22
  },
23
  "sources": {
24
  "climbmix": {
25
- "page": 1158,
26
- "row": 2810,
27
  "start": {
28
  "file": 56,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
- "row_group": 4,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
- "documents": 4745978
38
  },
39
  "rewrite6": {
40
- "page": 9,
41
- "row": 555,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
- "row": 36864
47
  }
48
  },
49
  "epoch": 13,
50
  "offset": 0,
51
- "documents": 2707671
52
  },
53
  "openmath": {
54
- "page": 724,
55
- "row": 1423,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
- "row_group": 345,
61
- "row": 604
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
- "documents": 2966927
67
  },
68
  "ultra_qa": {
69
- "page": 313,
70
- "row": 492,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
- "row": 49063
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
- "documents": 1282540
82
  },
83
  "ultra_style": {
84
- "page": 477,
85
- "row": 3572,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
- "row_group": 6,
91
- "row": 73569
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
- "documents": 1957364
97
  },
98
  "cortex": {
99
- "page": 61524,
100
- "row": 82,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
- "documents": 7875154
105
  }
106
  }
107
  }
 
1
  {
2
+ "step": 237163,
3
  "stats": {
4
+ "climbmix/blocks": 8537867,
5
+ "climbmix/targets": 14271950160,
6
+ "climbmix/input_tokens": 14272854649,
7
+ "ultra_qa/blocks": 3201701,
8
+ "ultra_qa/targets": 5223297877,
9
+ "ultra_qa/input_tokens": 5223307978,
10
+ "cortex/blocks": 3201700,
11
+ "cortex/targets": 121802303,
12
+ "cortex/input_tokens": 5089120155,
13
+ "ultra_style/blocks": 3201701,
14
+ "ultra_style/targets": 5064464924,
15
+ "ultra_style/input_tokens": 5064730515,
16
+ "openmath/blocks": 2134467,
17
+ "openmath/targets": 3298047515,
18
+ "openmath/input_tokens": 3299146987,
19
+ "rewrite6/blocks": 1067234,
20
+ "rewrite6/targets": 1749558797,
21
+ "rewrite6/input_tokens": 1751477016
22
  },
23
  "sources": {
24
  "climbmix": {
25
+ "page": 1169,
26
+ "row": 1585,
27
  "start": {
28
  "file": 56,
29
  "reader": {
30
  "format": "parquet-row-v1",
31
+ "row_group": 48,
32
  "row": 1024
33
  }
34
  },
35
  "epoch": 0,
36
  "offset": 2048,
37
+ "documents": 4789809
38
  },
39
  "rewrite6": {
40
+ "page": 15,
41
+ "row": 1043,
42
  "start": {
43
  "file": 0,
44
  "reader": {
45
  "format": "jsonl-row-v1",
46
+ "row": 61440
47
  }
48
  },
49
  "epoch": 13,
50
  "offset": 0,
51
+ "documents": 2732735
52
  },
53
  "openmath": {
54
+ "page": 730,
55
+ "row": 4073,
56
  "start": {
57
  "file": 6,
58
  "reader": {
59
  "format": "parquet-row-v1",
60
+ "row_group": 370,
61
+ "row": 180
62
  }
63
  },
64
  "epoch": 0,
65
  "offset": 0,
66
+ "documents": 2994153
67
  },
68
  "ultra_qa": {
69
+ "page": 315,
70
+ "row": 4076,
71
  "start": {
72
  "file": 2,
73
  "reader": {
74
  "format": "parquet-row-v1",
75
  "row_group": 3,
76
+ "row": 57255
77
  }
78
  },
79
  "epoch": 0,
80
  "offset": 2048,
81
+ "documents": 1294316
82
  },
83
  "ultra_style": {
84
+ "page": 482,
85
+ "row": 1439,
86
  "start": {
87
  "file": 2,
88
  "reader": {
89
  "format": "parquet-row-v1",
90
+ "row_group": 7,
91
+ "row": 8799
92
  }
93
  },
94
  "epoch": 0,
95
  "offset": 2048,
96
+ "documents": 1975711
97
  },
98
  "cortex": {
99
+ "page": 62090,
100
+ "row": 107,
101
  "start": null,
102
  "epoch": 0,
103
  "offset": 0,
104
+ "documents": 7947627
105
  }
106
  }
107
  }
latest.json CHANGED
@@ -1 +1 @@
1
- {"step": 235000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2e35e5395cb287d1981044a5822eddab4ba9c62bc411de4f902b706cfb265591", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "bdfb4bcc1f514f4993f7557ffa73c6c78e0fbe0f7eae146e44f89fbac2919cdd", "training_config.json": "2122ad6f8105987f9b87dd55d13b632084853908566b081c829ffe801259e2bf", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "b38e01cf77934ef444db8086750e69ecf4d69bd7e1c45ee0fec8d3051a6abc4d", "coverage.json": "d2f27535dad5eb059b3164dfa9e026883a87ba1676e6c7dc163b3f5eb36c7ad5", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000235000"}
 
1
+ {"step": 237163, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39", "training_config.json": "95ba9e002790d11959d1191b36ed994bfa10a6da51c22c772e143bec517a9783", "performance.json": "7f72f0a0e868a4baec06f238f922d0c5d7e397a51386c11ac32acc776d4d3522", "shard_scheduler.json": "7642e374dd789aa75440578f87a3db0e21df171d6322760efedf97e6eb520c03", "coverage.json": "70e3c6c517ccfd837f7cb477fd294e1034ff341fbf2983c3135f90eb9c011c05", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "9fee47918905b0750bee87876aa424f021ccc4507ee8db22a11d01fbc6109063", "folder": "step-000000237163"}
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2e35e5395cb287d1981044a5822eddab4ba9c62bc411de4f902b706cfb265591
3
  size 360622744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b20b64846b5ec5f1a1266ad70e66bdcf4161c507eb849f41d4457b347e1fef8
3
  size 360622744
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": 73759.62675692099,
26
- "required_pages_per_second_upper_estimate": 0.006393297505225003
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": 9219.953344615124,
40
- "required_pages_per_second_upper_estimate": 0.003677977239753919
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": 18439.906689230247,
54
- "required_pages_per_second_upper_estimate": 0.004265747881279806
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": 27659.86003384537,
68
- "required_pages_per_second_upper_estimate": 0.0016843403559910553
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": 27659.86003384537,
82
- "required_pages_per_second_upper_estimate": 0.0021293325927985215
83
  },
84
  {
85
  "source": "cortex",
@@ -91,9 +91,9 @@
91
  "read_seconds": 0.042780299000014566,
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
- "estimated_pages_for_buffer": 21,
95
- "required_token_positions_per_second": 27659.86003384537,
96
- "required_pages_per_second_upper_estimate": 0.3434428900237825
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,138 +155,30 @@
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
157
  "selected": {
158
- "prefetch_batches": 61,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.48818316624266345,
162
- "producer_p95_seconds": 0.43546763699851,
163
- "gpu_update_seconds": 0.9995712185878433,
164
- "producer_batches_per_second": 2.0484114757511436,
165
- "gpu_batches_per_second": 1.0004289653445229,
166
- "producer_headroom": 2.0475331549858926,
167
  "sustainable": true,
168
- "buffer_seconds": 60.97384433385844,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 81,
172
  "max_pending_pages": 4096,
173
- "generation_bytes_per_second": 293726.8938308259,
174
- "upload_bytes_per_second": 5635627.239740325,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
- "estimated_full_shards_per_hour": 63.02695380395492,
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": 0.9491155960022297,
184
- "bytes": 6968517,
185
- "pages": 33,
186
- "shards": 3
187
- },
188
- {
189
- "seconds": 1.2095077359990682,
190
- "bytes": 6033276,
191
- "pages": 36,
192
- "shards": 2
193
- },
194
- {
195
- "seconds": 2.290356701996643,
196
- "bytes": 13205046,
197
- "pages": 34,
198
- "shards": 4
199
- },
200
- {
201
- "seconds": 1.0317470599984517,
202
- "bytes": 1158559,
203
- "pages": 34,
204
- "shards": 1
205
- },
206
- {
207
- "seconds": 1.1566925789957168,
208
- "bytes": 9706808,
209
- "pages": 33,
210
- "shards": 3
211
- },
212
- {
213
- "seconds": 0.9952422389978892,
214
- "bytes": 7289939,
215
- "pages": 30,
216
- "shards": 3
217
- },
218
- {
219
- "seconds": 0.823028221995628,
220
- "bytes": 1028230,
221
- "pages": 34,
222
- "shards": 1
223
- },
224
- {
225
- "seconds": 1.2517391190049239,
226
- "bytes": 5921573,
227
- "pages": 37,
228
- "shards": 2
229
- },
230
- {
231
- "seconds": 1.3183760180036188,
232
- "bytes": 12970748,
233
- "pages": 31,
234
- "shards": 4
235
- },
236
- {
237
- "seconds": 1.064266913002939,
238
- "bytes": 6180437,
239
- "pages": 37,
240
- "shards": 2
241
- },
242
- {
243
- "seconds": 1.110571836004965,
244
- "bytes": 7330551,
245
- "pages": 29,
246
- "shards": 3
247
- },
248
- {
249
- "seconds": 1.0676300510021974,
250
- "bytes": 1203686,
251
- "pages": 35,
252
- "shards": 1
253
- },
254
- {
255
- "seconds": 1.0235948650006321,
256
- "bytes": 9738901,
257
- "pages": 35,
258
- "shards": 3
259
- },
260
- {
261
- "seconds": 0.8542503440039582,
262
- "bytes": 2305951,
263
- "pages": 28,
264
- "shards": 2
265
- },
266
- {
267
- "seconds": 0.9167553060033242,
268
- "bytes": 5839408,
269
- "pages": 28,
270
- "shards": 2
271
- },
272
- {
273
- "seconds": 1.0627733270011959,
274
- "bytes": 12238667,
275
- "pages": 44,
276
- "shards": 3
277
- },
278
- {
279
- "seconds": 1.0165252450024127,
280
- "bytes": 7299775,
281
- "pages": 36,
282
- "shards": 4
283
- },
284
- {
285
- "seconds": 1.2853216939984122,
286
- "bytes": 5894461,
287
- "pages": 29,
288
- "shards": 2
289
- },
290
  {
291
  "seconds": 1.3212683820020175,
292
  "bytes": 7578218,
@@ -562,6 +454,114 @@
562
  "bytes": 7202195,
563
  "pages": 30,
564
  "shards": 3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
565
  }
566
  ],
567
  "cache_key": {
 
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": 75472.3444404108,
26
+ "required_pages_per_second_upper_estimate": 0.006541751533186617
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": 9434.04305505135,
40
+ "required_pages_per_second_upper_estimate": 0.0037633808261733487
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": 18868.0861101027,
54
+ "required_pages_per_second_upper_estimate": 0.004364799654598218
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": 28302.129165154052,
68
+ "required_pages_per_second_upper_estimate": 0.0017234511763620484
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": 28302.129165154052,
82
+ "required_pages_per_second_upper_estimate": 0.002178776248441411
83
  },
84
  {
85
  "source": "cortex",
 
91
  "read_seconds": 0.042780299000014566,
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": 28302.129165154052,
96
+ "required_pages_per_second_upper_estimate": 0.35141772309812946
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
 
155
  "shard_benchmark_scope": "Local packing/read measurements; upload capacity measured only on real data commits. Targets larger than available pages are sample-limited.",
156
  "estimated_fresh_generation_batch_seconds": 0.1434574083060457,
157
  "selected": {
158
+ "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.12409712750809376,
162
+ "producer_p95_seconds": 0.4301016780009377,
163
+ "gpu_update_seconds": 0.9768876340950552,
164
+ "producer_batches_per_second": 8.05820424759452,
165
+ "gpu_batches_per_second": 1.0236591856609538,
166
+ "producer_headroom": 7.871960082487336,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.56703331389342,
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": 379406.6242124819,
174
+ "upload_bytes_per_second": 5575493.562052467,
175
  "upload_sustainable": true,
176
  "logical_page_rows": 4096,
177
  "logical_cortex_page_rows": 128,
178
+ "estimated_full_shards_per_hour": 81.41182942181437,
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": 1.3212683820020175,
184
  "bytes": 7578218,
 
454
  "bytes": 7202195,
455
  "pages": 30,
456
  "shards": 3
457
+ },
458
+ {
459
+ "seconds": 1.457639926004049,
460
+ "bytes": 7115441,
461
+ "pages": 39,
462
+ "shards": 2
463
+ },
464
+ {
465
+ "seconds": 1.29948361699644,
466
+ "bytes": 9783282,
467
+ "pages": 32,
468
+ "shards": 3
469
+ },
470
+ {
471
+ "seconds": 1.0601626389980083,
472
+ "bytes": 7547567,
473
+ "pages": 29,
474
+ "shards": 3
475
+ },
476
+ {
477
+ "seconds": 1.118395095996675,
478
+ "bytes": 1248062,
479
+ "pages": 35,
480
+ "shards": 1
481
+ },
482
+ {
483
+ "seconds": 0.9650491309948848,
484
+ "bytes": 5725695,
485
+ "pages": 29,
486
+ "shards": 2
487
+ },
488
+ {
489
+ "seconds": 1.1484884869933012,
490
+ "bytes": 6640918,
491
+ "pages": 32,
492
+ "shards": 3
493
+ },
494
+ {
495
+ "seconds": 1.1328601960049127,
496
+ "bytes": 11867525,
497
+ "pages": 34,
498
+ "shards": 3
499
+ },
500
+ {
501
+ "seconds": 1.0216524100032984,
502
+ "bytes": 7553152,
503
+ "pages": 40,
504
+ "shards": 3
505
+ },
506
+ {
507
+ "seconds": 0.8235062559979269,
508
+ "bytes": 854296,
509
+ "pages": 26,
510
+ "shards": 1
511
+ },
512
+ {
513
+ "seconds": 1.2072945199979586,
514
+ "bytes": 10797475,
515
+ "pages": 40,
516
+ "shards": 3
517
+ },
518
+ {
519
+ "seconds": 1.3132626469960087,
520
+ "bytes": 2377167,
521
+ "pages": 28,
522
+ "shards": 2
523
+ },
524
+ {
525
+ "seconds": 0.8975952609980595,
526
+ "bytes": 5947609,
527
+ "pages": 37,
528
+ "shards": 2
529
+ },
530
+ {
531
+ "seconds": 1.3013317000004463,
532
+ "bytes": 11518633,
533
+ "pages": 35,
534
+ "shards": 3
535
+ },
536
+ {
537
+ "seconds": 1.1227065580023918,
538
+ "bytes": 6254929,
539
+ "pages": 31,
540
+ "shards": 3
541
+ },
542
+ {
543
+ "seconds": 1.5046920809982112,
544
+ "bytes": 5774998,
545
+ "pages": 29,
546
+ "shards": 2
547
+ },
548
+ {
549
+ "seconds": 0.8562071050037048,
550
+ "bytes": 2686484,
551
+ "pages": 37,
552
+ "shards": 3
553
+ },
554
+ {
555
+ "seconds": 1.2383149749948643,
556
+ "bytes": 6204112,
557
+ "pages": 30,
558
+ "shards": 2
559
+ },
560
+ {
561
+ "seconds": 1.4738188229966909,
562
+ "bytes": 10769021,
563
+ "pages": 39,
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:bdfb4bcc1f514f4993f7557ffa73c6c78e0fbe0f7eae146e44f89fbac2919cdd
3
- size 721759615
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3817012a1ce6a6ac292529de34664fef83f2285d26044dba167d3aaa20957a39
3
+ size 721779903
training_config.json CHANGED
@@ -46,8 +46,8 @@
46
  "cpu_threads": 4,
47
  "page_rows": 4096,
48
  "cortex_page_rows": 128,
49
- "prefetch_batches": 61,
50
- "prefetch_low_water": 30,
51
  "memory_pages": 12,
52
  "cache_gb": 20,
53
  "force_rebenchmark": false,
 
46
  "cpu_threads": 4,
47
  "page_rows": 4096,
48
  "cortex_page_rows": 128,
49
+ "prefetch_batches": 62,
50
+ "prefetch_low_water": 31,
51
  "memory_pages": 12,
52
  "cache_gb": 20,
53
  "force_rebenchmark": false,