appvoid commited on
Commit
9286b23
·
verified ·
1 Parent(s): 9af8108

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": 76109.03168328833,
26
- "required_pages_per_second_upper_estimate": 0.006596937969730177
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": 9513.62896041104,
40
- "required_pages_per_second_upper_estimate": 0.0037951288337366527
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": 19027.25792082208,
54
- "required_pages_per_second_upper_estimate": 0.004401621251679954
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": 28540.88688123312,
68
- "required_pages_per_second_upper_estimate": 0.0017379902686063384
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": 28540.88688123312,
82
- "required_pages_per_second_upper_estimate": 0.0021971564783488367
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": 28540.88688123312,
96
- "required_pages_per_second_upper_estimate": 0.35438229486115846
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": 0.0530297397421684,
162
- "producer_p95_seconds": 0.13247945100010838,
163
- "gpu_update_seconds": 0.9687155173226157,
164
- "producer_batches_per_second": 18.857343159932878,
165
- "gpu_batches_per_second": 1.0322948090723785,
166
- "producer_headroom": 18.267400934504465,
167
  "sustainable": true,
168
- "buffer_seconds": 60.06036207400218,
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": 76301.26309665144,
26
+ "required_pages_per_second_upper_estimate": 0.006613600101434429
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": 9537.65788708143,
40
+ "required_pages_per_second_upper_estimate": 0.00380471433184994
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": 19075.31577416286,
54
+ "required_pages_per_second_upper_estimate": 0.004412738590261012
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": 28612.973661244287,
68
+ "required_pages_per_second_upper_estimate": 0.0017423799752989133
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": 28612.973661244287,
82
+ "required_pages_per_second_upper_estimate": 0.002202705918223075
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": 28612.973661244287,
96
+ "required_pages_per_second_upper_estimate": 0.3552773714099642
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": 0.0569888585546785,
162
+ "producer_p95_seconds": 0.1384437749998142,
163
+ "gpu_update_seconds": 0.9662749606989879,
164
+ "producer_batches_per_second": 17.547289511695354,
165
+ "gpu_batches_per_second": 1.034902114483662,
166
+ "producer_headroom": 16.955506483287188,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.875322524036235,
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
@@ -21,3 +21,4 @@
21
  {"step": 277100, "loss": 0.561845388263464, "lr": 0.0003, "grad_norm": 0.13909139895340553, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40559593939, "tokens_per_second": 151182.7518431791, "data_wait_s": 1.3313000181369716e-05, "gpu_peak_gb": 85.46427917480469}
22
  {"step": 277200, "loss": 0.6728517189621925, "lr": 0.0003, "grad_norm": 0.1942734196814653, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40574057787, "tokens_per_second": 149074.59816718844, "data_wait_s": 9.401000170328189e-06, "gpu_peak_gb": 85.46427917480469}
23
  {"step": 277300, "loss": 0.5808306690305471, "lr": 0.0003, "grad_norm": 0.15871520061710287, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40588611599, "tokens_per_second": 149994.4769097146, "data_wait_s": 1.2185999821667792e-05, "gpu_peak_gb": 85.46427917480469}
 
 
21
  {"step": 277100, "loss": 0.561845388263464, "lr": 0.0003, "grad_norm": 0.13909139895340553, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40559593939, "tokens_per_second": 151182.7518431791, "data_wait_s": 1.3313000181369716e-05, "gpu_peak_gb": 85.46427917480469}
22
  {"step": 277200, "loss": 0.6728517189621925, "lr": 0.0003, "grad_norm": 0.1942734196814653, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40574057787, "tokens_per_second": 149074.59816718844, "data_wait_s": 9.401000170328189e-06, "gpu_peak_gb": 85.46427917480469}
23
  {"step": 277300, "loss": 0.5808306690305471, "lr": 0.0003, "grad_norm": 0.15871520061710287, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40588611599, "tokens_per_second": 149994.4769097146, "data_wait_s": 1.2185999821667792e-05, "gpu_peak_gb": 85.46427917480469}
24
+ {"step": 277400, "loss": 0.567566778510809, "lr": 0.0003, "grad_norm": 0.14623475431320984, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 40603211126, "tokens_per_second": 150490.46349487503, "data_wait_s": 1.1923999863938661e-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:bf2e90b7250251c74f76aebdfb9636a44b630be2cb0bfb3a4d77accb0aad13be
3
- size 48879
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec7094213c591c6642ea0ef59f1fe0b057849f176883426c809bea8f9795e7b4
3
+ size 50997