appvoid commited on
Commit
110a015
·
verified ·
1 Parent(s): 652466b

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": 77473.98472718996,
26
- "required_pages_per_second_upper_estimate": 0.005359830027838271
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": 9684.248090898745,
40
- "required_pages_per_second_upper_estimate": 0.004054592445562816
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": 19368.49618179749,
54
- "required_pages_per_second_upper_estimate": 0.004033946973075529
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": 29052.744272696233,
68
- "required_pages_per_second_upper_estimate": 0.0017416676581762183
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": 29052.744272696233,
82
- "required_pages_per_second_upper_estimate": 0.002258723077685289
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": 29052.744272696233,
96
- "required_pages_per_second_upper_estimate": 0.37265740912374434
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": 64,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.05252354896876099,
162
- "producer_p95_seconds": 0.13148544900013803,
163
- "gpu_update_seconds": 0.951648482514734,
164
- "producer_batches_per_second": 19.039079034715684,
165
- "gpu_batches_per_second": 1.0508081695853673,
166
- "producer_headroom": 18.118510671865266,
167
  "sustainable": true,
168
- "buffer_seconds": 60.905502880942976,
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,
 
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": 77070.83304288298,
26
+ "required_pages_per_second_upper_estimate": 0.005331939058877128
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": 9633.854130360372,
40
+ "required_pages_per_second_upper_estimate": 0.004033493546630961
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": 19267.708260720745,
54
+ "required_pages_per_second_upper_estimate": 0.0040129555070714616
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": 28901.562391081115,
68
+ "required_pages_per_second_upper_estimate": 0.0017326045352147592
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": 28901.562391081115,
82
+ "required_pages_per_second_upper_estimate": 0.002246969351368534
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": 28901.562391081115,
96
+ "required_pages_per_second_upper_estimate": 0.370718210272843
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": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.05512894966405568,
162
+ "producer_p95_seconds": 0.1570716309997806,
163
+ "gpu_update_seconds": 0.9566264835748824,
164
+ "producer_batches_per_second": 18.139289902923807,
165
+ "gpu_batches_per_second": 1.0453400749088946,
166
+ "producer_headroom": 17.35252511437937,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.2674684652176,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 79,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791463982175382982-d4a3b372/train.jsonl CHANGED
@@ -28,3 +28,4 @@
28
  {"step": 927800, "loss": 0.5746034383773804, "lr": 3e-05, "grad_norm": 0.3320251980846008, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135803567729, "tokens_per_second": 153915.41208555488, "data_wait_s": 1.2613999842869816e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 927900, "loss": 0.555435299873352, "lr": 3e-05, "grad_norm": 0.333205786955412, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135818387398, "tokens_per_second": 155677.48903948313, "data_wait_s": 9.808999948290875e-06, "gpu_peak_gb": 89.37899398803711}
30
  {"step": 928000, "loss": 0.5244894027709961, "lr": 3e-05, "grad_norm": 0.34271158779188887, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135833040179, "tokens_per_second": 153403.62911481096, "data_wait_s": 1.130600003307336e-05, "gpu_peak_gb": 89.37899398803711}
 
 
28
  {"step": 927800, "loss": 0.5746034383773804, "lr": 3e-05, "grad_norm": 0.3320251980846008, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135803567729, "tokens_per_second": 153915.41208555488, "data_wait_s": 1.2613999842869816e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 927900, "loss": 0.555435299873352, "lr": 3e-05, "grad_norm": 0.333205786955412, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135818387398, "tokens_per_second": 155677.48903948313, "data_wait_s": 9.808999948290875e-06, "gpu_peak_gb": 89.37899398803711}
30
  {"step": 928000, "loss": 0.5244894027709961, "lr": 3e-05, "grad_norm": 0.34271158779188887, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135833040179, "tokens_per_second": 153403.62911481096, "data_wait_s": 1.130600003307336e-05, "gpu_peak_gb": 89.37899398803711}
31
+ {"step": 928100, "loss": 0.5575608015060425, "lr": 3e-05, "grad_norm": 0.35268365170265237, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 135847590762, "tokens_per_second": 152781.83261242192, "data_wait_s": 6.800999926781515e-06, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791463982.2e076f0c-3fe2-4316-8f17-4e30b2e706e2-b5972.360.0.1791463982175382982-d4a3b372 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2170dd3481c8dfe141df955db2c9cf1cebe22cd4dab06896d2629faea6a97712
3
- size 63705
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c22a2cf8437612b5ee5c57bf82cc43006ede9d29673f51d79376ad0a9662285
3
+ size 65823