appvoid commited on
Commit
7f3de42
·
verified ·
1 Parent(s): 35a0b63

Update TensorBoard training metrics

Browse files
benchmark_cache/shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 77399.48758423836,
26
- "required_pages_per_second_upper_estimate": 0.005354676142631693
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9674.935948029795,
40
- "required_pages_per_second_upper_estimate": 0.004050693645803243
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19349.87189605959,
54
- "required_pages_per_second_upper_estimate": 0.004030068025511768
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 29024.807844089388,
68
- "required_pages_per_second_upper_estimate": 0.0017399929119376945
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 29024.807844089388,
82
- "required_pages_per_second_upper_estimate": 0.00225655114324047
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
- "required_token_positions_per_second": 29024.807844089388,
96
- "required_pages_per_second_upper_estimate": 0.37229907061337575
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": 2.558590814491007,
157
  "selected": {
158
- "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.052796195554690506,
162
- "producer_p95_seconds": 0.11202996099996199,
163
- "gpu_update_seconds": 0.9525644458531787,
164
- "producer_batches_per_second": 18.940758694707846,
165
- "gpu_batches_per_second": 1.0497977374164276,
166
- "producer_headroom": 18.042293310063158,
167
  "sustainable": true,
168
- "buffer_seconds": 60.01156008875026,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 41,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
 
22
  "tokens_per_second": 598315.063954628,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 77442.14330750014,
26
+ "required_pages_per_second_upper_estimate": 0.005357627164541857
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 8533.957213384612,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9680.267913437518,
40
+ "required_pages_per_second_upper_estimate": 0.004052926028375357
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 380144.35761751566,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19360.535826875035,
54
+ "required_pages_per_second_upper_estimate": 0.0040322890410738575
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 43742.16556775435,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 29040.803740312553,
68
+ "required_pages_per_second_upper_estimate": 0.001740951841491954
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 22018.86276523721,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 29040.803740312553,
82
+ "required_pages_per_second_upper_estimate": 0.002257794753813309
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 23,
95
+ "required_token_positions_per_second": 29040.803740312553,
96
+ "required_pages_per_second_upper_estimate": 0.37250424879507127
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": 2.558590814491007,
157
  "selected": {
158
+ "prefetch_batches": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.04780414914843778,
162
+ "producer_p95_seconds": 0.12575584399996842,
163
+ "gpu_update_seconds": 0.9520397660902493,
164
+ "producer_batches_per_second": 20.918686302623577,
165
+ "gpu_batches_per_second": 1.0503762926907028,
166
+ "producer_headroom": 19.915421214465056,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.930545029775956,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 53,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791143623306220418-eb05a537/train.jsonl CHANGED
@@ -1,3 +1,4 @@
1
  {"step": 685100, "loss": 0.5776553750038147, "lr": 8e-05, "grad_norm": 0.20740935451668885, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100281063608, "tokens_per_second": 152913.44909213492, "data_wait_s": 1.2569999967126932e-05, "gpu_peak_gb": 89.37899398803711}
2
  {"step": 685200, "loss": 0.5670952200889587, "lr": 8e-05, "grad_norm": 0.21379803612541587, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100295657942, "tokens_per_second": 153444.8572623894, "data_wait_s": 8.530000059181475e-06, "gpu_peak_gb": 89.37899398803711}
3
  {"step": 685300, "loss": 0.6020703315734863, "lr": 8e-05, "grad_norm": 0.2038634766871492, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100310140405, "tokens_per_second": 152199.12537603173, "data_wait_s": 1.5465999922525953e-05, "gpu_peak_gb": 89.37899398803711}
 
 
1
  {"step": 685100, "loss": 0.5776553750038147, "lr": 8e-05, "grad_norm": 0.20740935451668885, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100281063608, "tokens_per_second": 152913.44909213492, "data_wait_s": 1.2569999967126932e-05, "gpu_peak_gb": 89.37899398803711}
2
  {"step": 685200, "loss": 0.5670952200889587, "lr": 8e-05, "grad_norm": 0.21379803612541587, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100295657942, "tokens_per_second": 153444.8572623894, "data_wait_s": 8.530000059181475e-06, "gpu_peak_gb": 89.37899398803711}
3
  {"step": 685300, "loss": 0.6020703315734863, "lr": 8e-05, "grad_norm": 0.2038634766871492, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100310140405, "tokens_per_second": 152199.12537603173, "data_wait_s": 1.5465999922525953e-05, "gpu_peak_gb": 89.37899398803711}
4
+ {"step": 685400, "loss": 0.5669589638710022, "lr": 8e-05, "grad_norm": 0.1897535348704872, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 100324882735, "tokens_per_second": 154863.5560354908, "data_wait_s": 1.8494000073587813e-05, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791143623.3eb6ddb3-ffce-42e5-bd48-c0d653189e98-8pb7d.369.0.1791143623306220418-eb05a537 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:64b102ee9c34547274516a33c366ed20ff573ddeecf7c04a054cec3b4015654f
3
- size 6519
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79c5eacfa85523e8c225d21cf7645706e6c9514be1ec08e4aedb0877fe438696
3
+ size 8637