appvoid commited on
Commit
5ed9603
·
verified ·
1 Parent(s): 5aa54c2

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": 76007.14455885073,
26
- "required_pages_per_second_upper_estimate": 0.006588106652014483
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": 9500.893069856342,
40
- "required_pages_per_second_upper_estimate": 0.0037900482965758505
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": 19001.786139712684,
54
- "required_pages_per_second_upper_estimate": 0.004395728803408332
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": 28502.679209569025,
68
- "required_pages_per_second_upper_estimate": 0.0017356636218621559
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": 28502.679209569025,
82
- "required_pages_per_second_upper_estimate": 0.00219421514601853
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": 28502.679209569025,
96
- "required_pages_per_second_upper_estimate": 0.3539078834519417
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": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.04570335661718694,
162
- "producer_p95_seconds": 0.11890651299995625,
163
- "gpu_update_seconds": 0.9700140747020691,
164
- "producer_batches_per_second": 21.880230994323636,
165
- "gpu_batches_per_second": 1.0309128765035092,
166
- "producer_headroom": 21.224132022226378,
167
  "sustainable": true,
168
- "buffer_seconds": 60.14087263152829,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 52,
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": 76262.40147915448,
26
+ "required_pages_per_second_upper_estimate": 0.0066102316749236615
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": 9532.80018489431,
40
+ "required_pages_per_second_upper_estimate": 0.003802776521818378
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": 19065.60036978862,
54
+ "required_pages_per_second_upper_estimate": 0.004410491102444309
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": 28598.400554682932,
68
+ "required_pages_per_second_upper_estimate": 0.0017414925495685078
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": 28598.400554682932,
82
+ "required_pages_per_second_upper_estimate": 0.002201584040139041
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": 28598.400554682932,
96
+ "required_pages_per_second_upper_estimate": 0.35509642219952237
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.04658439075000054,
162
+ "producer_p95_seconds": 0.12595866000003753,
163
+ "gpu_update_seconds": 0.9667673528501822,
164
+ "producer_batches_per_second": 21.466417911669016,
165
+ "gpu_batches_per_second": 1.0343750200623167,
166
+ "producer_headroom": 20.753032019639992,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.90634322956148,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 55,
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
@@ -2,3 +2,4 @@
2
  {"step": 115200, "loss": 0.5842613279819489, "lr": 0.0003, "grad_norm": 0.13957367498710943, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16857707507, "tokens_per_second": 151021.34513643512, "data_wait_s": 2.3415999976350577e-05, "gpu_peak_gb": 85.46427917480469}
3
  {"step": 115300, "loss": 0.617220900952816, "lr": 0.0003, "grad_norm": 0.12405028442484603, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16872235692, "tokens_per_second": 149967.25268749727, "data_wait_s": 1.4869999972688674e-05, "gpu_peak_gb": 85.46427917480469}
4
  {"step": 115400, "loss": 0.610551755875349, "lr": 0.0003, "grad_norm": 0.15634487661387214, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16886971989, "tokens_per_second": 152076.35164390146, "data_wait_s": 1.957099993887823e-05, "gpu_peak_gb": 85.46427917480469}
 
 
2
  {"step": 115200, "loss": 0.5842613279819489, "lr": 0.0003, "grad_norm": 0.13957367498710943, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16857707507, "tokens_per_second": 151021.34513643512, "data_wait_s": 2.3415999976350577e-05, "gpu_peak_gb": 85.46427917480469}
3
  {"step": 115300, "loss": 0.617220900952816, "lr": 0.0003, "grad_norm": 0.12405028442484603, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16872235692, "tokens_per_second": 149967.25268749727, "data_wait_s": 1.4869999972688674e-05, "gpu_peak_gb": 85.46427917480469}
4
  {"step": 115400, "loss": 0.610551755875349, "lr": 0.0003, "grad_norm": 0.15634487661387214, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16886971989, "tokens_per_second": 152076.35164390146, "data_wait_s": 1.957099993887823e-05, "gpu_peak_gb": 85.46427917480469}
5
+ {"step": 115500, "loss": 0.5834813602268696, "lr": 0.0003, "grad_norm": 0.15759322127084688, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 16901693141, "tokens_per_second": 151467.96586301902, "data_wait_s": 1.4530000044032931e-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:fa4f5c029465cc904ef3ce81e1b31d427598cbaa6d4cdcfa993f228d3f6a4c8d
3
- size 8637
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4bc66c6b379270cdf878f3298835135441158c38e4acfacdaaf69c017fb13500
3
+ size 10755