appvoid commited on
Commit
64755c4
·
verified ·
1 Parent(s): 3932527

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": 76008.31219904328,
26
- "required_pages_per_second_upper_estimate": 0.0065882078601070156
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": 9501.03902488041,
40
- "required_pages_per_second_upper_estimate": 0.0037901065202171737
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": 19002.07804976082,
54
- "required_pages_per_second_upper_estimate": 0.004395796331660528
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": 28503.117074641228,
68
- "required_pages_per_second_upper_estimate": 0.0017356902855477523
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": 28503.117074641228,
82
- "required_pages_per_second_upper_estimate": 0.0021942488540838763
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": 28503.117074641228,
96
- "required_pages_per_second_upper_estimate": 0.3539133202706983
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": 690670,
161
- "producer_batch_seconds": 0.05806758685935165,
162
- "producer_p95_seconds": 0.1541919940000298,
163
- "gpu_update_seconds": 0.9699991733394656,
164
- "producer_batches_per_second": 17.22131147660999,
165
- "gpu_batches_per_second": 1.0309287136371972,
166
- "producer_headroom": 16.704657896133142,
167
  "sustainable": true,
168
- "buffer_seconds": 60.139948747046866,
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": 76240.37723507191,
26
+ "required_pages_per_second_upper_estimate": 0.0066083226692665205
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": 9530.047154383989,
40
+ "required_pages_per_second_upper_estimate": 0.0038016782967863366
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": 19060.094308767977,
54
+ "required_pages_per_second_upper_estimate": 0.004409217372130548
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": 28590.141463151966,
68
+ "required_pages_per_second_upper_estimate": 0.0017409896142263764
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": 28590.141463151966,
82
+ "required_pages_per_second_upper_estimate": 0.0022009482324103586
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": 28590.141463151966,
96
+ "required_pages_per_second_upper_estimate": 0.354993871924109
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": 690670,
161
+ "producer_batch_seconds": 0.052247876749923705,
162
+ "producer_p95_seconds": 0.15275691499937238,
163
+ "gpu_update_seconds": 0.9670466316381738,
164
+ "producer_batches_per_second": 19.13953374194216,
165
+ "gpu_batches_per_second": 1.0340762971336794,
166
+ "producer_headroom": 18.508821636270337,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.92393779320495,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1790696019647648547-3fdea0a8/train.jsonl CHANGED
@@ -36,3 +36,4 @@
36
  {"step": 278600, "loss": 0.5726190768182278, "lr": 0.0003, "grad_norm": 0.13401794960571295, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40779598508, "tokens_per_second": 149480.0919833245, "data_wait_s": 1.6554000467294827e-05, "gpu_peak_gb": 85.46427917480469}
37
  {"step": 278700, "loss": 0.5746679967269301, "lr": 0.0003, "grad_norm": 0.15670635691392462, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40794039350, "tokens_per_second": 148513.58695770724, "data_wait_s": 1.127800078393193e-05, "gpu_peak_gb": 85.46427917480469}
38
  {"step": 278800, "loss": 0.6013021990656853, "lr": 0.0003, "grad_norm": 0.16295544019408925, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40808645024, "tokens_per_second": 150309.74284495253, "data_wait_s": 1.3943999874754809e-05, "gpu_peak_gb": 85.46427917480469}
 
 
36
  {"step": 278600, "loss": 0.5726190768182278, "lr": 0.0003, "grad_norm": 0.13401794960571295, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40779598508, "tokens_per_second": 149480.0919833245, "data_wait_s": 1.6554000467294827e-05, "gpu_peak_gb": 85.46427917480469}
37
  {"step": 278700, "loss": 0.5746679967269301, "lr": 0.0003, "grad_norm": 0.15670635691392462, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40794039350, "tokens_per_second": 148513.58695770724, "data_wait_s": 1.127800078393193e-05, "gpu_peak_gb": 85.46427917480469}
38
  {"step": 278800, "loss": 0.6013021990656853, "lr": 0.0003, "grad_norm": 0.16295544019408925, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40808645024, "tokens_per_second": 150309.74284495253, "data_wait_s": 1.3943999874754809e-05, "gpu_peak_gb": 85.46427917480469}
39
+ {"step": 278900, "loss": 0.5618883538991213, "lr": 0.0003, "grad_norm": 0.1863066547620779, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40823246291, "tokens_per_second": 150619.76547717387, "data_wait_s": 1.0720999853219837e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790696019.3e6d470a-8cc7-4d7e-a224-51a0f9a2ce55-85btc.22407.0.1790696019647648547-3fdea0a8 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9c1219223534b9f9453fad7b8c2ec0c18976193f48bbf6c0c0071c5a2bfd6078
3
- size 80649
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d667597cd22edc35726295996819816779b81ea3c0876bf020b655f6e4df5161
3
+ size 82767