appvoid commited on
Commit
3b135ca
·
verified ·
1 Parent(s): d1152a7

Update TensorBoard training metrics

Browse files
benchmark_cache/shard_scheduler.json CHANGED
@@ -22,8 +22,8 @@
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
- "required_token_positions_per_second": 76156.70045956816,
26
- "required_pages_per_second_upper_estimate": 0.006601069778442705
27
  },
28
  {
29
  "source": "rewrite6",
@@ -36,8 +36,8 @@
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
- "required_token_positions_per_second": 9519.58755744602,
40
- "required_pages_per_second_upper_estimate": 0.003797505807182871
41
  },
42
  {
43
  "source": "openmath",
@@ -50,8 +50,8 @@
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
- "required_token_positions_per_second": 19039.17511489204,
54
- "required_pages_per_second_upper_estimate": 0.0044043780847925875
55
  },
56
  {
57
  "source": "ultra_qa",
@@ -64,8 +64,8 @@
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
- "required_token_positions_per_second": 28558.762672338064,
68
- "required_pages_per_second_upper_estimate": 0.0017390788104976018
69
  },
70
  {
71
  "source": "ultra_style",
@@ -78,8 +78,8 @@
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
- "required_token_positions_per_second": 28558.762672338064,
82
- "required_pages_per_second_upper_estimate": 0.002198532606231452
83
  },
84
  {
85
  "source": "cortex",
@@ -92,8 +92,8 @@
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
- "required_token_positions_per_second": 28558.762672338064,
96
- "required_pages_per_second_upper_estimate": 0.35460425236025755
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": 0.1434574083060457,
157
  "selected": {
158
- "prefetch_batches": 62,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
- "producer_batch_seconds": 0.052907974976548644,
162
- "producer_p95_seconds": 0.14818189600009646,
163
- "gpu_update_seconds": 0.9681091690565353,
164
- "producer_batches_per_second": 18.900742287778886,
165
- "gpu_batches_per_second": 1.03294135822982,
166
- "producer_headroom": 18.297981910773334,
167
  "sustainable": true,
168
- "buffer_seconds": 60.02276848150519,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 83,
172
  "max_pending_pages": 4096,
173
  "generation_bytes_per_second": 0,
174
  "upload_bytes_per_second": null,
 
22
  "tokens_per_second": 1062967.6168812835,
23
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
24
  "estimated_pages_for_buffer": 1,
25
+ "required_token_positions_per_second": 76221.4173211979,
26
+ "required_pages_per_second_upper_estimate": 0.006606679271985385
27
  },
28
  {
29
  "source": "rewrite6",
 
36
  "tokens_per_second": 1206672.263377472,
37
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
38
  "estimated_pages_for_buffer": 1,
39
+ "required_token_positions_per_second": 9527.677165149738,
40
+ "required_pages_per_second_upper_estimate": 0.0038007328726462975
41
  },
42
  {
43
  "source": "openmath",
 
50
  "tokens_per_second": 410042.7581134458,
51
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
52
  "estimated_pages_for_buffer": 1,
53
+ "required_token_positions_per_second": 19055.354330299477,
54
+ "required_pages_per_second_upper_estimate": 0.004408120861532632
55
  },
56
  {
57
  "source": "ultra_qa",
 
64
  "tokens_per_second": 2163486.5624068812,
65
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
66
  "estimated_pages_for_buffer": 1,
67
+ "required_token_positions_per_second": 28583.031495449213,
68
+ "required_pages_per_second_upper_estimate": 0.0017405566544963964
69
  },
70
  {
71
  "source": "ultra_style",
 
78
  "tokens_per_second": 1546694.2737633225,
79
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
80
  "estimated_pages_for_buffer": 1,
81
+ "required_token_positions_per_second": 28583.031495449213,
82
+ "required_pages_per_second_upper_estimate": 0.0022004008874149505
83
  },
84
  {
85
  "source": "cortex",
 
92
  "tokens_per_second": 1015177.8395898147,
93
  "note": "Bounded fresh sample; full-page latency is an extrapolation.",
94
  "estimated_pages_for_buffer": 22,
95
+ "required_token_positions_per_second": 28583.031495449213,
96
+ "required_pages_per_second_upper_estimate": 0.354905589920772
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": 0.1434574083060457,
157
  "selected": {
158
+ "prefetch_batches": 63,
159
  "prefetch_capacity_limit": 512,
160
  "batch_memory_estimate_bytes": 688946,
161
+ "producer_batch_seconds": 0.056569761343766345,
162
+ "producer_p95_seconds": 0.1495996820003711,
163
+ "gpu_update_seconds": 0.967287182411072,
164
+ "producer_batches_per_second": 17.677288647607032,
165
+ "gpu_batches_per_second": 1.0338191368435046,
166
+ "producer_headroom": 17.09901472861104,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.93909249189754,
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,
metrics/void-byte-v1/1790519810984249790-cb9c9513/train.jsonl CHANGED
@@ -24,3 +24,4 @@
24
  {"step": 117400, "loss": 0.5866196602582932, "lr": 0.0003, "grad_norm": 0.1448304762316454, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17179656845, "tokens_per_second": 150238.43506845544, "data_wait_s": 1.3821000266034389e-05, "gpu_peak_gb": 85.46427917480469}
25
  {"step": 117500, "loss": 0.600348450243473, "lr": 0.0003, "grad_norm": 0.13086762944878322, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17194356680, "tokens_per_second": 151583.4417689883, "data_wait_s": 1.4385000213223975e-05, "gpu_peak_gb": 85.46427917480469}
26
  {"step": 117600, "loss": 0.6016160920262337, "lr": 0.0003, "grad_norm": 0.1248867513326305, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17209104158, "tokens_per_second": 151913.17321309296, "data_wait_s": 1.3078000392852118e-05, "gpu_peak_gb": 85.46427917480469}
 
 
24
  {"step": 117400, "loss": 0.5866196602582932, "lr": 0.0003, "grad_norm": 0.1448304762316454, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17179656845, "tokens_per_second": 150238.43506845544, "data_wait_s": 1.3821000266034389e-05, "gpu_peak_gb": 85.46427917480469}
25
  {"step": 117500, "loss": 0.600348450243473, "lr": 0.0003, "grad_norm": 0.13086762944878322, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17194356680, "tokens_per_second": 151583.4417689883, "data_wait_s": 1.4385000213223975e-05, "gpu_peak_gb": 85.46427917480469}
26
  {"step": 117600, "loss": 0.6016160920262337, "lr": 0.0003, "grad_norm": 0.1248867513326305, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17209104158, "tokens_per_second": 151913.17321309296, "data_wait_s": 1.3078000392852118e-05, "gpu_peak_gb": 85.46427917480469}
27
+ {"step": 117700, "loss": 0.6191273359581828, "lr": 0.0003, "grad_norm": 0.16089004303991808, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17223795313, "tokens_per_second": 151289.57461195142, "data_wait_s": 1.3388000297709368e-05, "gpu_peak_gb": 85.46427917480469}
runs/void-byte-v1/events.out.tfevents.1790519810.cad0b19a-59f2-43f0-8fb4-2a94097843e2-tjh7c.1336.0.1790519810984249790-cb9c9513 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b9638a026efc361c062cd815b1db36c545baf56d3c1aafe049b9a5f43c2a1841
3
- size 55233
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bd3502e59ad1d7c34cf949f338a3f8d54554bdb3e4d6b2e2d95941bb2c0456e
3
+ size 57351