appvoid commited on
Commit
4e388c5
·
verified ·
1 Parent(s): 41fbf81

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": 75677.68217031317,
26
- "required_pages_per_second_upper_estimate": 0.005235557648957365
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": 9459.710271289146,
40
- "required_pages_per_second_upper_estimate": 0.003960583149375141
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": 18919.420542578293,
54
- "required_pages_per_second_upper_estimate": 0.003940416360347178
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": 28379.130813867443,
68
- "required_pages_per_second_upper_estimate": 0.0017012855598676291
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": 28379.130813867443,
82
- "required_pages_per_second_upper_estimate": 0.0022063525941738256
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": 28379.130813867443,
96
- "required_pages_per_second_upper_estimate": 0.36401701894366983
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": 0.05531432020313787,
162
  "producer_p95_seconds": 0.1398530239998763,
163
- "gpu_update_seconds": 0.9742370258390657,
164
- "producer_batches_per_second": 18.07850112461966,
165
- "gpu_batches_per_second": 1.026444256867312,
166
- "producer_headroom": 17.612745167277662,
167
  "sustainable": true,
168
- "buffer_seconds": 60.40269560202207,
169
  "buffer_drain_seconds": null,
170
  "shard_target_mib": 16,
171
- "upload_pages_per_commit": 81,
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": 76765.74061191983,
26
+ "required_pages_per_second_upper_estimate": 0.0053108320565911265
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": 9595.717576489978,
40
+ "required_pages_per_second_upper_estimate": 0.004017526567907232
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": 19191.435152979957,
54
+ "required_pages_per_second_upper_estimate": 0.003997069830186171
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": 28787.152729469937,
68
+ "required_pages_per_second_upper_estimate": 0.0017257458506946073
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": 28787.152729469937,
82
+ "required_pages_per_second_upper_estimate": 0.0022380745034131827
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": 28787.152729469937,
96
+ "required_pages_per_second_upper_estimate": 0.36925068597721855
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": 0.05145686314061848,
162
  "producer_p95_seconds": 0.1398530239998763,
163
+ "gpu_update_seconds": 0.9604284334690814,
164
+ "producer_batches_per_second": 19.43375361353169,
165
+ "gpu_batches_per_second": 1.0412019939767772,
166
+ "producer_headroom": 18.66472953946834,
167
  "sustainable": true,
168
+ "buffer_seconds": 60.506991308552124,
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,
metrics/void-byte-v1/1791209746770022608-f3fe5fb7/train.jsonl CHANGED
@@ -28,3 +28,4 @@
28
  {"step": 722800, "loss": 0.5469934344291687, "lr": 5e-05, "grad_norm": 0.21146521720862674, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105798063516, "tokens_per_second": 153817.08469523638, "data_wait_s": 1.5890000213403255e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 722900, "loss": 0.5500357151031494, "lr": 5e-05, "grad_norm": 0.2232098414074729, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105812839171, "tokens_per_second": 152915.06034777255, "data_wait_s": 1.3776999821857316e-05, "gpu_peak_gb": 89.37899398803711}
30
  {"step": 723000, "loss": 0.580702006816864, "lr": 5e-05, "grad_norm": 0.2446644316669375, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105827552658, "tokens_per_second": 152596.44957717348, "data_wait_s": 1.8846999864763347e-05, "gpu_peak_gb": 89.37899398803711}
 
 
28
  {"step": 722800, "loss": 0.5469934344291687, "lr": 5e-05, "grad_norm": 0.21146521720862674, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105798063516, "tokens_per_second": 153817.08469523638, "data_wait_s": 1.5890000213403255e-05, "gpu_peak_gb": 89.37899398803711}
29
  {"step": 722900, "loss": 0.5500357151031494, "lr": 5e-05, "grad_norm": 0.2232098414074729, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105812839171, "tokens_per_second": 152915.06034777255, "data_wait_s": 1.3776999821857316e-05, "gpu_peak_gb": 89.37899398803711}
30
  {"step": 723000, "loss": 0.580702006816864, "lr": 5e-05, "grad_norm": 0.2446644316669375, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105827552658, "tokens_per_second": 152596.44957717348, "data_wait_s": 1.8846999864763347e-05, "gpu_peak_gb": 89.37899398803711}
31
+ {"step": 723100, "loss": 0.5263711214065552, "lr": 5e-05, "grad_norm": 0.22687436973325104, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 105842314080, "tokens_per_second": 153348.85338377196, "data_wait_s": 1.1264000022492837e-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 sha256:7ac40cd8d5b6604badb892ef8def3e3d2b34204c3d6f868088b11d682e63784d
3
- size 63705
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cb31993abe0af93d36ab3299846286ebfc15b5435fe26346a71567dfaba9820
3
+ size 65823