appvoid commited on
Commit
17bb428
·
verified ·
1 Parent(s): 73a7f59

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": 77526.26304064406,
26
- "required_pages_per_second_upper_estimate": 0.005363446762865425
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": 9690.782880080507,
40
- "required_pages_per_second_upper_estimate": 0.004057328425331304
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": 19381.565760161015,
54
- "required_pages_per_second_upper_estimate": 0.004036669021580692
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": 29072.348640241522,
68
- "required_pages_per_second_upper_estimate": 0.0017428429100764225
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": 29072.348640241522,
82
- "required_pages_per_second_upper_estimate": 0.0022602472310314357
83
  },
84
  {
85
  "source": "cortex",
@@ -91,9 +91,9 @@
91
  "read_seconds": 0.04061180899998362,
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": 29072.348640241522,
96
- "required_pages_per_second_upper_estimate": 0.37290887290108543
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": 0.05213821181249756,
162
- "producer_p95_seconds": 0.13747843300006934,
163
- "gpu_update_seconds": 0.9510067570437031,
164
- "producer_batches_per_second": 19.179790891107995,
165
- "gpu_batches_per_second": 1.0515172395920689,
166
- "producer_headroom": 18.240110736128972,
167
  "sustainable": true,
168
- "buffer_seconds": 60.864432450797,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 76,
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": 76010.09132657066,
26
+ "required_pages_per_second_upper_estimate": 0.005258554485682764
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": 9501.261415821333,
40
+ "required_pages_per_second_upper_estimate": 0.003977979745904202
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": 19002.522831642666,
54
+ "required_pages_per_second_upper_estimate": 0.0039577243756045485
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": 28503.784247464002,
68
+ "required_pages_per_second_upper_estimate": 0.0017087583428769704
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": 28503.784247464002,
82
+ "required_pages_per_second_upper_estimate": 0.0022160438503434493
83
  },
84
  {
85
  "source": "cortex",
 
91
  "read_seconds": 0.04061180899998362,
92
  "tokens_per_second": 910315.184299418,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
+ "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28503.784247464002,
96
+ "required_pages_per_second_upper_estimate": 0.3656159393474173
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": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
+ "producer_batch_seconds": 0.054260372234395504,
162
+ "producer_p95_seconds": 0.13925519799977337,
163
+ "gpu_update_seconds": 0.9699764690879549,
164
+ "producer_batches_per_second": 18.429656097458594,
165
+ "gpu_batches_per_second": 1.030952844598669,
166
+ "producer_headroom": 17.876332747918184,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.138541083453205,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 84,
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
@@ -27,3 +27,4 @@
27
  {"step": 762700, "loss": 0.5494645833969116, "lr": 3e-05, "grad_norm": 0.2529401843485463, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111636742854, "tokens_per_second": 154850.33984594542, "data_wait_s": 1.6755000160628697e-05, "gpu_peak_gb": 89.37899398803711}
28
  {"step": 762800, "loss": 0.5381356477737427, "lr": 3e-05, "grad_norm": 0.2355803750415523, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111651462115, "tokens_per_second": 154221.52309780213, "data_wait_s": 2.231799999208306e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 762900, "loss": 0.5140604376792908, "lr": 3e-05, "grad_norm": 0.2409726487696894, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111666008715, "tokens_per_second": 151746.17577388752, "data_wait_s": 1.589900011822465e-05, "gpu_peak_gb": 89.37899398803711}
 
 
27
  {"step": 762700, "loss": 0.5494645833969116, "lr": 3e-05, "grad_norm": 0.2529401843485463, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111636742854, "tokens_per_second": 154850.33984594542, "data_wait_s": 1.6755000160628697e-05, "gpu_peak_gb": 89.37899398803711}
28
  {"step": 762800, "loss": 0.5381356477737427, "lr": 3e-05, "grad_norm": 0.2355803750415523, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111651462115, "tokens_per_second": 154221.52309780213, "data_wait_s": 2.231799999208306e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 762900, "loss": 0.5140604376792908, "lr": 3e-05, "grad_norm": 0.2409726487696894, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111666008715, "tokens_per_second": 151746.17577388752, "data_wait_s": 1.589900011822465e-05, "gpu_peak_gb": 89.37899398803711}
30
+ {"step": 763000, "loss": 0.577847421169281, "lr": 3e-05, "grad_norm": 0.2346762175215377, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 111680576336, "tokens_per_second": 152632.09971097927, "data_wait_s": 1.1066000297432765e-05, "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 sha256:6ca60e196224a50dd5be82c2760529d590f519e23776b4dc97c06ca37cde5408
3
- size 61587
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:134da2ea91ccc6e6254ed91daa36dbbb6ab0c9e5d5a3ac3444948cc16bf066f9
3
+ size 63705