appvoid commited on
Commit
39ac37b
·
verified ·
1 Parent(s): 75bfb6b

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": 77131.52601695101,
26
- "required_pages_per_second_upper_estimate": 0.005336137939650256
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": 9641.440752118877,
40
- "required_pages_per_second_upper_estimate": 0.00403666990673457
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": 19282.881504237754,
54
- "required_pages_per_second_upper_estimate": 0.004016115693550714
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": 28924.32225635663,
68
- "required_pages_per_second_upper_estimate": 0.001733968954385724
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": 28924.32225635663,
82
- "required_pages_per_second_upper_estimate": 0.0022487388307836378
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": 28924.32225635663,
96
- "required_pages_per_second_upper_estimate": 0.371010149386958
97
  }
98
  ],
99
  "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -155,17 +155,17 @@
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": 0.06653700627341408,
162
- "producer_p95_seconds": 0.15965528699962306,
163
- "gpu_update_seconds": 0.9558737368141397,
164
- "producer_batches_per_second": 15.029230438934942,
165
- "gpu_batches_per_second": 1.0461632760545656,
166
- "producer_headroom": 14.366046661105555,
167
  "sustainable": true,
168
- "buffer_seconds": 60.2200454192908,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 82,
 
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": 75558.19694967395,
26
+ "required_pages_per_second_upper_estimate": 0.005227291384149635
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": 9444.774618709243,
40
+ "required_pages_per_second_upper_estimate": 0.003954329903531828
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": 18889.549237418487,
54
+ "required_pages_per_second_upper_estimate": 0.003934194955241682
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": 28334.32385612773,
68
+ "required_pages_per_second_upper_estimate": 0.0016985994511673963
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": 28334.32385612773,
82
+ "required_pages_per_second_upper_estimate": 0.0022028690502945416
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": 28334.32385612773,
96
+ "required_pages_per_second_upper_estimate": 0.36344228339974766
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": 0.05478669785155432,
162
+ "producer_p95_seconds": 0.14724258800015377,
163
+ "gpu_update_seconds": 0.9757776518821253,
164
+ "producer_batches_per_second": 18.252605818834354,
165
+ "gpu_batches_per_second": 1.0248236348425828,
166
+ "producer_headroom": 17.810484846632203,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.498214416691766,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
  "upload_pages_per_commit": 82,
metrics/void-byte-v1/1791300814123050567-72be911d/train.jsonl CHANGED
@@ -35,3 +35,4 @@
35
  {"step": 803500, "loss": 0.5135341286659241, "lr": 3e-05, "grad_norm": 0.28913286282964196, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117609406982, "tokens_per_second": 154507.92294203243, "data_wait_s": 1.105800038203597e-05, "gpu_peak_gb": 89.37899398803711}
36
  {"step": 803600, "loss": 0.5231519937515259, "lr": 3e-05, "grad_norm": 0.2646122519530628, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117623837178, "tokens_per_second": 150324.62267700967, "data_wait_s": 1.383400012855418e-05, "gpu_peak_gb": 89.37899398803711}
37
  {"step": 803700, "loss": 0.5389294624328613, "lr": 3e-05, "grad_norm": 0.2732866946955584, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117638367567, "tokens_per_second": 150884.93780683746, "data_wait_s": 1.6381000023102388e-05, "gpu_peak_gb": 89.37899398803711}
 
 
35
  {"step": 803500, "loss": 0.5135341286659241, "lr": 3e-05, "grad_norm": 0.28913286282964196, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117609406982, "tokens_per_second": 154507.92294203243, "data_wait_s": 1.105800038203597e-05, "gpu_peak_gb": 89.37899398803711}
36
  {"step": 803600, "loss": 0.5231519937515259, "lr": 3e-05, "grad_norm": 0.2646122519530628, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117623837178, "tokens_per_second": 150324.62267700967, "data_wait_s": 1.383400012855418e-05, "gpu_peak_gb": 89.37899398803711}
37
  {"step": 803700, "loss": 0.5389294624328613, "lr": 3e-05, "grad_norm": 0.2732866946955584, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117638367567, "tokens_per_second": 150884.93780683746, "data_wait_s": 1.6381000023102388e-05, "gpu_peak_gb": 89.37899398803711}
38
+ {"step": 803800, "loss": 0.5099411606788635, "lr": 3e-05, "grad_norm": 0.270241294812033, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 117652931556, "tokens_per_second": 152130.01819733303, "data_wait_s": 1.70599996636156e-05, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791300814.0c9d201b-2dbc-40fb-b739-4ee2350a0f54-7b2tx.840.0.1791300814123050567-72be911d CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3f76d28777c2f2fd5364d395ed3e8ee96eff6c7621c2f3b1af4b88e6602a9307
3
- size 78531
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ed2debfb071d674d42ee42afd31df3a3d8d086f1b53cdfd05d6f61c955482da
3
+ size 80649