appvoid commited on
Commit
7942805
·
verified ·
1 Parent(s): 7b1c6e4

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": 75963.27878278683,
26
- "required_pages_per_second_upper_estimate": 0.005255315885283733
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": 9495.409847848354,
40
- "required_pages_per_second_upper_estimate": 0.00397552981658855
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": 18990.819695696708,
54
- "required_pages_per_second_upper_estimate": 0.003955286921019455
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": 28486.22954354506,
68
- "required_pages_per_second_upper_estimate": 0.0017077059651815108
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": 28486.22954354506,
82
- "required_pages_per_second_upper_estimate": 0.0022146790493287336
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": 22,
95
- "required_token_positions_per_second": 28486.22954354506,
96
- "required_pages_per_second_upper_estimate": 0.3653907664543177
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": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 720750,
161
- "producer_batch_seconds": 0.0535936187968753,
162
- "producer_p95_seconds": 0.12324112800001785,
163
- "gpu_update_seconds": 0.9705742192990576,
164
- "producer_batches_per_second": 18.658937807317905,
165
- "gpu_batches_per_second": 1.0303179088377121,
166
- "producer_headroom": 18.109883995287245,
167
  "sustainable": true,
168
- "buffer_seconds": 60.175601596541576,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 47,
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": 76666.80380649082,
26
+ "required_pages_per_second_upper_estimate": 0.0053039873788265305
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": 9583.350475811352,
40
+ "required_pages_per_second_upper_estimate": 0.00401234872110752
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": 19166.700951622704,
54
+ "required_pages_per_second_upper_estimate": 0.00399191834832821
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": 28750.051427434057,
68
+ "required_pages_per_second_upper_estimate": 0.0017235216842879642
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": 28750.051427434057,
82
+ "required_pages_per_second_upper_estimate": 0.0022351900403712738
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": 23,
95
+ "required_token_positions_per_second": 28750.051427434057,
96
+ "required_pages_per_second_upper_estimate": 0.3687747903109767
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": 0.04395623965625095,
162
+ "producer_p95_seconds": 0.11782706700000745,
163
+ "gpu_update_seconds": 0.9616678450048806,
164
+ "producer_batches_per_second": 22.749898713362565,
165
+ "gpu_batches_per_second": 1.0398600776705027,
166
+ "producer_headroom": 21.877846069758682,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.585074235307474,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
+ "upload_pages_per_commit": 61,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl CHANGED
@@ -1,3 +1,4 @@
1
  {"step": 720100, "loss": 0.5276756882667542, "lr": 5e-05, "grad_norm": 0.19444800064694195, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105402386187, "tokens_per_second": 151641.9104605178, "data_wait_s": 1.1152999945807096e-05, "gpu_peak_gb": 89.37899398803711}
2
  {"step": 720200, "loss": 0.5311113595962524, "lr": 5e-05, "grad_norm": 0.2152443446279208, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105416991235, "tokens_per_second": 150859.00996360285, "data_wait_s": 8.567000008952164e-06, "gpu_peak_gb": 89.37899398803711}
3
  {"step": 720300, "loss": 0.5242825150489807, "lr": 5e-05, "grad_norm": 0.20403188480716858, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105431760929, "tokens_per_second": 153686.81241369364, "data_wait_s": 9.222000016961829e-06, "gpu_peak_gb": 89.37899398803711}
 
 
1
  {"step": 720100, "loss": 0.5276756882667542, "lr": 5e-05, "grad_norm": 0.19444800064694195, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105402386187, "tokens_per_second": 151641.9104605178, "data_wait_s": 1.1152999945807096e-05, "gpu_peak_gb": 89.37899398803711}
2
  {"step": 720200, "loss": 0.5311113595962524, "lr": 5e-05, "grad_norm": 0.2152443446279208, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105416991235, "tokens_per_second": 150859.00996360285, "data_wait_s": 8.567000008952164e-06, "gpu_peak_gb": 89.37899398803711}
3
  {"step": 720300, "loss": 0.5242825150489807, "lr": 5e-05, "grad_norm": 0.20403188480716858, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105431760929, "tokens_per_second": 153686.81241369364, "data_wait_s": 9.222000016961829e-06, "gpu_peak_gb": 89.37899398803711}
4
+ {"step": 720400, "loss": 0.5264872312545776, "lr": 5e-05, "grad_norm": 0.20526975189012117, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105446489981, "tokens_per_second": 153182.13289004244, "data_wait_s": 1.7705999994177546e-05, "gpu_peak_gb": 89.37899398803711}
runs/void-byte-v1/events.out.tfevents.1791209746.ebd20fd8-0561-42d8-bf2b-56d08e7106ef-6qqk2.329.0.1791209746770022608-f3fe5fb7 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3871b4e7628337d317a6bfbec038cf44bb0da73fc7ec013addbbc1b7c8eb39be
3
- size 6519
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42cd5ca4ce1b92612e542e223c344cd4dea26d8ec93a456e6b52901595399acf
3
+ size 8637