appvoid commited on
Commit
4b2403c
·
verified ·
1 Parent(s): 3f8e00f

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": 75989.12012803665,
26
- "required_pages_per_second_upper_estimate": 0.00658654434003402
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": 9498.640016004581,
40
- "required_pages_per_second_upper_estimate": 0.003789149519708226
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": 18997.280032009163,
54
- "required_pages_per_second_upper_estimate": 0.00439468639469591
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": 28495.920048013744,
68
- "required_pages_per_second_upper_estimate": 0.001735252024386019
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": 28495.920048013744,
82
- "required_pages_per_second_upper_estimate": 0.0021936948070514477
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": 28495.920048013744,
96
- "required_pages_per_second_upper_estimate": 0.3538239572868836
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": 0.05719081141406335,
162
- "producer_p95_seconds": 0.17010058800042316,
163
- "gpu_update_seconds": 0.9702441596346054,
164
- "producer_batches_per_second": 17.48532631859281,
165
- "gpu_batches_per_second": 1.030668404514386,
166
- "producer_headroom": 16.96503573991993,
167
  "sustainable": true,
168
- "buffer_seconds": 60.15513789734553,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 79,
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": 76376.25931618252,
26
+ "required_pages_per_second_upper_estimate": 0.0066201005836672025
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": 9547.032414522815,
40
+ "required_pages_per_second_upper_estimate": 0.0038084539710079844
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": 19094.06482904563,
54
+ "required_pages_per_second_upper_estimate": 0.004417075854136045
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": 28641.09724356845,
68
+ "required_pages_per_second_upper_estimate": 0.0017440925539094218
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": 28641.09724356845,
82
+ "required_pages_per_second_upper_estimate": 0.0022048709494414476
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": 28641.09724356845,
96
+ "required_pages_per_second_upper_estimate": 0.35562657217885507
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": 0.05389935951563274,
162
+ "producer_p95_seconds": 0.15546805500025584,
163
+ "gpu_update_seconds": 0.9653261453245666,
164
+ "producer_batches_per_second": 18.553096158962042,
165
+ "gpu_batches_per_second": 1.0359193158119375,
166
+ "producer_headroom": 17.90978879896685,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.815547155447696,
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-v1/1790519810984249790-cb9c9513/train.jsonl CHANGED
@@ -20,3 +20,4 @@
20
  {"step": 117000, "loss": 0.6301776804029942, "lr": 0.0003, "grad_norm": 0.13138056593190925, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17121284753, "tokens_per_second": 149447.57248424378, "data_wait_s": 1.3673000012204284e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 117100, "loss": 0.6071493308991194, "lr": 0.0003, "grad_norm": 0.14773526847702778, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17135712327, "tokens_per_second": 148386.77664299984, "data_wait_s": 1.4575000022887252e-05, "gpu_peak_gb": 85.46427917480469}
22
  {"step": 117200, "loss": 0.6492827702313662, "lr": 0.0003, "grad_norm": 0.15350586535446745, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17150398058, "tokens_per_second": 151387.28465944977, "data_wait_s": 1.5145999896049034e-05, "gpu_peak_gb": 85.46427917480469}
 
 
20
  {"step": 117000, "loss": 0.6301776804029942, "lr": 0.0003, "grad_norm": 0.13138056593190925, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17121284753, "tokens_per_second": 149447.57248424378, "data_wait_s": 1.3673000012204284e-05, "gpu_peak_gb": 85.46427917480469}
21
  {"step": 117100, "loss": 0.6071493308991194, "lr": 0.0003, "grad_norm": 0.14773526847702778, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17135712327, "tokens_per_second": 148386.77664299984, "data_wait_s": 1.4575000022887252e-05, "gpu_peak_gb": 85.46427917480469}
22
  {"step": 117200, "loss": 0.6492827702313662, "lr": 0.0003, "grad_norm": 0.15350586535446745, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17150398058, "tokens_per_second": 151387.28465944977, "data_wait_s": 1.5145999896049034e-05, "gpu_peak_gb": 85.46427917480469}
23
+ {"step": 117300, "loss": 0.6031660884618759, "lr": 0.0003, "grad_norm": 0.12929184356664308, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 86, "input_bytes_and_specials": 17165055303, "tokens_per_second": 150776.66775610272, "data_wait_s": 1.257900021300884e-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:c325d467f9185042c9c9ddd1f63fdb7b133e1f1ce9e23d0145b2647c99a01484
3
- size 46761
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0a81bb7f493bdb1e3684bedaae4e796e633ec0dc40431a9c3df82d77c553b62
3
+ size 48879