appvoid commited on
Commit
5b55c01
·
verified ·
1 Parent(s): 26f5cff

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": 77445.08511839515,
26
- "required_pages_per_second_upper_estimate": 0.005357830685845515
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": 9680.635639799393,
40
- "required_pages_per_second_upper_estimate": 0.004053079987724075
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": 19361.271279598786,
54
- "required_pages_per_second_upper_estimate": 0.004032442216481023
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": 29041.90691939818,
68
- "required_pages_per_second_upper_estimate": 0.0017410179753936804
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": 29041.90691939818,
82
- "required_pages_per_second_upper_estimate": 0.0022578805211348448
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": 29041.90691939818,
96
- "required_pages_per_second_upper_estimate": 0.3725183991918803
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": 0.04851462940623996,
162
- "producer_p95_seconds": 0.12414443499983463,
163
- "gpu_update_seconds": 0.9520036021303017,
164
- "producer_batches_per_second": 20.61233925186657,
165
- "gpu_batches_per_second": 1.0504161935546217,
166
- "producer_headroom": 19.62302121610878,
167
  "sustainable": true,
168
- "buffer_seconds": 60.928230536339306,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 69,
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": 77282.53438266298,
26
+ "required_pages_per_second_upper_estimate": 0.005346585048777893
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": 9660.316797832873,
40
+ "required_pages_per_second_upper_estimate": 0.004044572912898362
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": 19320.633595665746,
54
+ "required_pages_per_second_upper_estimate": 0.004023978458605563
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": 28980.950393498617,
68
+ "required_pages_per_second_upper_estimate": 0.0017373637247412273
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": 28980.950393498617,
82
+ "required_pages_per_second_upper_estimate": 0.0022531414193655754
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": 28980.950393498617,
96
+ "required_pages_per_second_upper_estimate": 0.3717365143276589
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": 0.057878406593772525,
162
+ "producer_p95_seconds": 0.1468161810003039,
163
+ "gpu_update_seconds": 0.9540059806390048,
164
+ "producer_batches_per_second": 17.277600729727,
165
+ "gpu_batches_per_second": 1.048211458098185,
166
+ "producer_headroom": 16.48293442725239,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.1023767802573,
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,
metrics/void-byte-v1/1791254470898546162-7724d9f6/train.jsonl CHANGED
@@ -16,3 +16,4 @@
16
  {"step": 761600, "loss": 0.5470552444458008, "lr": 3e-05, "grad_norm": 0.25005745380968136, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111475755654, "tokens_per_second": 151434.22608220068, "data_wait_s": 1.94229999124218e-05, "gpu_peak_gb": 89.37899398803711}
17
  {"step": 761700, "loss": 0.5007930397987366, "lr": 3e-05, "grad_norm": 0.23310555419277346, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111490271322, "tokens_per_second": 152385.58246989583, "data_wait_s": 1.2020999747619499e-05, "gpu_peak_gb": 89.37899398803711}
18
  {"step": 761800, "loss": 0.5855912566184998, "lr": 3e-05, "grad_norm": 0.27354794981320624, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111505032369, "tokens_per_second": 154600.75998703312, "data_wait_s": 1.6495000181748765e-05, "gpu_peak_gb": 89.37899398803711}
 
 
16
  {"step": 761600, "loss": 0.5470552444458008, "lr": 3e-05, "grad_norm": 0.25005745380968136, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111475755654, "tokens_per_second": 151434.22608220068, "data_wait_s": 1.94229999124218e-05, "gpu_peak_gb": 89.37899398803711}
17
  {"step": 761700, "loss": 0.5007930397987366, "lr": 3e-05, "grad_norm": 0.23310555419277346, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111490271322, "tokens_per_second": 152385.58246989583, "data_wait_s": 1.2020999747619499e-05, "gpu_peak_gb": 89.37899398803711}
18
  {"step": 761800, "loss": 0.5855912566184998, "lr": 3e-05, "grad_norm": 0.27354794981320624, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111505032369, "tokens_per_second": 154600.75998703312, "data_wait_s": 1.6495000181748765e-05, "gpu_peak_gb": 89.37899398803711}
19
+ {"step": 761900, "loss": 0.5362855195999146, "lr": 3e-05, "grad_norm": 0.24191452364117072, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111519760426, "tokens_per_second": 154197.6498428081, "data_wait_s": 9.1489996520977e-06, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791254470.1e0c9aa3-e42e-441e-9e17-2e25895f438d-9bhv5.366.0.1791254470898546162-7724d9f6 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e3b9d56a83bf16ea78baaed36a604672357da1d60f7f243ed33d1bf265b100b4
3
- size 38289
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f074399ef8d78431a1c04b1b123afdea05d5678d95bffb89f427eaf092e5a2d
3
+ size 40407