appvoid commited on
Commit
cd2c352
·
verified ·
1 Parent(s): 648f6da

Update TensorBoard training metrics

Browse files
benchmark_cache/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": 76213.41939936919,
26
- "required_pages_per_second_upper_estimate": 0.006605986032391819
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": 9526.677424921148,
40
- "required_pages_per_second_upper_estimate": 0.0038003340613216646
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": 19053.354849842297,
54
- "required_pages_per_second_upper_estimate": 0.0044076583169185175
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": 28580.032274763445,
68
- "required_pages_per_second_upper_estimate": 0.001740374017692328
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": 28580.032274763445,
82
- "required_pages_per_second_upper_estimate": 0.0022001699991041858
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": 28580.032274763445,
96
- "required_pages_per_second_upper_estimate": 0.35486834963760067
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,20 +155,20 @@
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.06228470236717243,
162
- "producer_p95_seconds": 0.1804239270004473,
163
- "gpu_update_seconds": 0.9673886906143755,
164
- "producer_batches_per_second": 16.055306712472255,
165
- "gpu_batches_per_second": 1.033710658086062,
166
- "producer_headroom": 15.531722137990728,
167
  "sustainable": true,
168
- "buffer_seconds": 60.945487508705654,
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": 0,
174
  "upload_bytes_per_second": null,
 
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": 75977.04011188867,
26
+ "required_pages_per_second_upper_estimate": 0.006585497274850834
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": 9497.130013986083,
40
+ "required_pages_per_second_upper_estimate": 0.0037885471573265053
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": 18994.260027972166,
54
+ "required_pages_per_second_upper_estimate": 0.004393987769912206
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": 28491.39004195825,
68
+ "required_pages_per_second_upper_estimate": 0.0017349761707843444
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": 28491.39004195825,
82
+ "required_pages_per_second_upper_estimate": 0.0021933460746454366
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": 28491.39004195825,
96
+ "required_pages_per_second_upper_estimate": 0.35376770977262934
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": 0.05420650439063479,
162
+ "producer_p95_seconds": 0.17275564200008375,
163
+ "gpu_update_seconds": 0.9703984242005665,
164
+ "producer_batches_per_second": 18.447970612411766,
165
+ "gpu_batches_per_second": 1.0305045588092538,
166
+ "producer_headroom": 17.901881611982738,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.164702300435124,
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": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl CHANGED
@@ -37,3 +37,4 @@
37
  {"step": 118700, "loss": 0.6312964111566544, "lr": 0.0003, "grad_norm": 0.14389962333335488, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17370383301, "tokens_per_second": 151218.79794761713, "data_wait_s": 1.2882000191893894e-05, "gpu_peak_gb": 85.46427917480469}
38
  {"step": 118800, "loss": 0.6232862267643213, "lr": 0.0003, "grad_norm": 0.14507646341057645, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17384892783, "tokens_per_second": 149216.5196001673, "data_wait_s": 1.2628000149561558e-05, "gpu_peak_gb": 85.46427917480469}
39
  {"step": 118900, "loss": 0.5925244968384504, "lr": 0.0003, "grad_norm": 0.1351402662102939, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17399398353, "tokens_per_second": 148453.03547389834, "data_wait_s": 1.2634000086109154e-05, "gpu_peak_gb": 85.46427917480469}
 
 
37
  {"step": 118700, "loss": 0.6312964111566544, "lr": 0.0003, "grad_norm": 0.14389962333335488, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17370383301, "tokens_per_second": 151218.79794761713, "data_wait_s": 1.2882000191893894e-05, "gpu_peak_gb": 85.46427917480469}
38
  {"step": 118800, "loss": 0.6232862267643213, "lr": 0.0003, "grad_norm": 0.14507646341057645, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17384892783, "tokens_per_second": 149216.5196001673, "data_wait_s": 1.2628000149561558e-05, "gpu_peak_gb": 85.46427917480469}
39
  {"step": 118900, "loss": 0.5925244968384504, "lr": 0.0003, "grad_norm": 0.1351402662102939, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17399398353, "tokens_per_second": 148453.03547389834, "data_wait_s": 1.2634000086109154e-05, "gpu_peak_gb": 85.46427917480469}
40
+ {"step": 119000, "loss": 0.6057021394371986, "lr": 0.0003, "grad_norm": 0.15349554496346607, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17414016155, "tokens_per_second": 150603.53231536408, "data_wait_s": 1.1053999514842872e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d342a041caaeaa1ed0e8330feda3bd98bd041bb2c8fee42c89fbe6fc212eb34d
3
- size 82767
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d720ba93f4f42d878a206330b78456b368a225e438219624a673be271123e68
3
+ size 84885