MarxistLeninist commited on
Commit
5db3ba4
·
1 Parent(s): 56731b1

Upload checkpoints/step2865113_20260911/serving_candidate.json with huggingface_hub

Browse files
checkpoints/step2865113_20260911/serving_candidate.json ADDED
@@ -0,0 +1,710 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "agillm43_provenance": {
3
+ "batch_size": 56,
4
+ "block_size": 2048,
5
+ "checkpoint_path": "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02865113_from02852036_20260911T1839Z.pt",
6
+ "checkpoint_type": "full",
7
+ "created_at_iso": "2026-09-11T18:39:12Z",
8
+ "created_at_unix": 1789151952.4425156,
9
+ "created_from_scratch": false,
10
+ "dataset_provenance": {
11
+ "chat": false,
12
+ "dataset_field_text": "text",
13
+ "dataset_hotload_config_path": "/workspace/agillm43_dataset_hotload.json",
14
+ "dataset_hotload_config_sha256": "3f48fe33b475c719d43e14cc4d0867c19e17dfa1d2cb304cd291eea6a879a7c8",
15
+ "dataset_hotload_config_stamp": [
16
+ 50,
17
+ 1124990893,
18
+ 1789091829464680808,
19
+ 7210
20
+ ],
21
+ "dataset_hotload_config_used": true,
22
+ "dataset_hotload_enabled": true,
23
+ "dataset_hotload_generation": 13,
24
+ "dataset_hotload_schema": "agillm.dataset_hotload.config.v2",
25
+ "dataset_revision_map_sha256": "8361995a504a2365136b20e1a879ba78944a06309f13f3f013bbd0cdfc72225b",
26
+ "dataset_revisions": {
27
+ "EleutherAI/proof-pile-2": "901a9273a770e9d4138c5ddd91802f9c5c6cdc4b",
28
+ "HuggingFaceFW/fineweb": "9bb295ddab0e05d785b879661af7260fed5140fc",
29
+ "HuggingFaceFW/fineweb-edu": "87f09149ef4734204d70ed1d046ddc9ca3f2b8f9",
30
+ "HuggingFaceTB/cosmopedia": "0ae6ec63f91742bd2d1eaef4f02232c55d719385",
31
+ "HuggingFaceTB/finemath": "e92b25a616738fe95dc186b64dfb19f9c8525594",
32
+ "HuggingFaceTB/smollm-corpus": "3ba9d605774198c5868892d7a8deda78031a781f",
33
+ "Skylion007/openwebtext": "79d93d786212f7344586290adb811d4ae6a1762c",
34
+ "allenai/c4": "1588ec454efa1a09f29cd18ddd04fe05fc8653a2",
35
+ "allenai/dolma": "7f48140530a023e9ea4c5cfb141160922727d4d3",
36
+ "allenai/dolmino-mix-1124": "a319f19eef1e257417b11ea8c30da266ae175557",
37
+ "codeparrot/codeparrot-clean": "35a59fb025bc0a102f7d96eac09d145b896d487b",
38
+ "mlfoundations/dclm-baseline-1.0": "a3b142c183aebe5af344955ae20836eb34dcf69b",
39
+ "wikimedia/wikipedia": "b04c8d1ceb2f5cd4588862100d08de323dccfbaa"
40
+ },
41
+ "dolma_payload_integrity": "external_url_manifest_pinned_payload_hashes_unavailable",
42
+ "dolma_url_list_sha256": "bc6ef505be920c94bb9e7147445638f606fdc3f5af622cd1d887dc48e21a2d5b",
43
+ "dolma_url_manifest_raw_sha256": "6fe8c40d51e5bebc9303453d7f533b232d69f7ae122123621a82621b1cd779b7",
44
+ "has_language_mix": true,
45
+ "has_numeracy": false,
46
+ "hot_config_expected_sha256": "",
47
+ "hot_config_has_append": true,
48
+ "hot_config_has_datasets": false,
49
+ "hot_config_mtime": 1789092253.2119184,
50
+ "hot_config_path": "/workspace/agillm43_production_v14/contracts/hot_config_owner_v25.json",
51
+ "hot_config_used": true,
52
+ "hot_reload_capable": true,
53
+ "phase": "pretrain",
54
+ "proof_pile_lfs_manifest_sha256": "d020dfb796b02a75564e4e11d9e03185a55b3f9c6639f87830cfbe10737ed329",
55
+ "proof_pile_paths_sha256": "8a30018017e3b3d9a9bcfa06a3c404ccd113297b0b1d0c48e2e2aadf113e52b0",
56
+ "schema": "agillm.dataset_provenance.v1",
57
+ "sequence_prefetch_depth": 256,
58
+ "source_count": 23,
59
+ "source_effective": "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96,wikimedia/wikipedia:20231101.en|1.63,allenai/c4:en|0.36,Skylion007/openwebtext|0.18,HuggingFaceTB/cosmopedia:web_samples_v2|0.64,HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65,EleutherAI/proof-pile-2:all|0.67,allenai/dolma:v1_6-sample|0.5,codeparrot/codeparrot-clean|0.24,HuggingFaceTB/finemath:finemath-4plus|0.32,HuggingFaceTB/finemath:infiwebmath-4plus|0.23,HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05,allenai/dolmino-mix-1124|0.78,mlfoundations/dclm-baseline-1.0|1.05,HuggingFaceFW/fineweb-edu:sample-100BT|2.4,json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35",
60
+ "source_requested": "HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25",
61
+ "source_specs": [
62
+ "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96",
63
+ "wikimedia/wikipedia:20231101.en|1.63",
64
+ "allenai/c4:en|0.36",
65
+ "Skylion007/openwebtext|0.18",
66
+ "HuggingFaceTB/cosmopedia:web_samples_v2|0.64",
67
+ "HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65",
68
+ "EleutherAI/proof-pile-2:all|0.67",
69
+ "allenai/dolma:v1_6-sample|0.5",
70
+ "codeparrot/codeparrot-clean|0.24",
71
+ "HuggingFaceTB/finemath:finemath-4plus|0.32",
72
+ "HuggingFaceTB/finemath:infiwebmath-4plus|0.23",
73
+ "HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05",
74
+ "allenai/dolmino-mix-1124|0.78",
75
+ "mlfoundations/dclm-baseline-1.0|1.05",
76
+ "HuggingFaceFW/fineweb-edu:sample-100BT|2.4",
77
+ "json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03",
78
+ "json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45",
79
+ "json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35",
80
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35",
81
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35",
82
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2",
83
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2",
84
+ "json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35"
85
+ ],
86
+ "val_source_effective": "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96,wikimedia/wikipedia:20231101.en|1.63,allenai/c4:en|0.36,Skylion007/openwebtext|0.18,HuggingFaceTB/cosmopedia:web_samples_v2|0.64,HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65,EleutherAI/proof-pile-2:all|0.67,allenai/dolma:v1_6-sample|0.5,codeparrot/codeparrot-clean|0.24,HuggingFaceTB/finemath:finemath-4plus|0.32,HuggingFaceTB/finemath:infiwebmath-4plus|0.23,HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05,allenai/dolmino-mix-1124|0.78,mlfoundations/dclm-baseline-1.0|1.05,HuggingFaceFW/fineweb-edu:sample-100BT|2.4,json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35",
87
+ "val_source_requested": ""
88
+ },
89
+ "effective_global_step": 5717149,
90
+ "effective_seen_tok": 285479475200,
91
+ "global_origin_seen_tok": 0,
92
+ "global_origin_step": 2852036,
93
+ "gpu": {
94
+ "allocated_gb": 7.54,
95
+ "peak_allocated_gb": 8.97,
96
+ "reserved_gb": 13.47
97
+ },
98
+ "hostname": "ead6505f92c4",
99
+ "lane": "pretrain",
100
+ "launch": {
101
+ "argv": [
102
+ "/workspace/agillm44_throughput_80k_20260910/runtime_retain_activations_v2.py",
103
+ "train",
104
+ "--preset",
105
+ "agillm4_floor",
106
+ "--block",
107
+ "2048",
108
+ "--source",
109
+ "HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25",
110
+ "--nat_mask_token_id",
111
+ "2",
112
+ "--tie_weights",
113
+ "--tie_kv",
114
+ "--moe_ffn",
115
+ "--moe_experts",
116
+ "2",
117
+ "--moe_top_k",
118
+ "1",
119
+ "--moe_mlp_mult",
120
+ "4",
121
+ "--moe_shared_experts",
122
+ "1",
123
+ "--moe_shared_mlp_mult",
124
+ "2",
125
+ "--moe_aux_coef",
126
+ "0.01",
127
+ "--moe_z_coef",
128
+ "0.001",
129
+ "--amp",
130
+ "--grad_checkpoint",
131
+ "--attn_backend",
132
+ "sdpa",
133
+ "--sublinear_window",
134
+ "128",
135
+ "--sublinear_stride",
136
+ "128",
137
+ "--sublinear_max_anchors",
138
+ "128",
139
+ "--sublinear_chunk",
140
+ "128",
141
+ "--sublinear_sinks",
142
+ "4",
143
+ "--sublinear_recent_anchors",
144
+ "64",
145
+ "--no-sublinear_pooled_landmarks",
146
+ "--anchor_stride",
147
+ "256",
148
+ "--anchor_max",
149
+ "2048",
150
+ "--anchor_position",
151
+ "-1",
152
+ "--alibi_mode",
153
+ "corrected",
154
+ "--alibi_scale",
155
+ "0.0",
156
+ "--optimizer",
157
+ "adamw8bit",
158
+ "--weight_decay",
159
+ "0.0",
160
+ "--lr_decay",
161
+ "cosine",
162
+ "--lr_decay_tokens",
163
+ "49000954880",
164
+ "--lr_min_mult",
165
+ "0.2",
166
+ "--lr_schedule_reset_on_resume",
167
+ "--lr_warmup_tokens",
168
+ "10000000",
169
+ "--lr_warmup_min_mult",
170
+ "0.1",
171
+ "--dblock",
172
+ "--dblock_schedule",
173
+ "roundrobin",
174
+ "--dblock_router_ramp_steps",
175
+ "64",
176
+ "--dblock_warmup_steps",
177
+ "14",
178
+ "--dblock_checkpoint_skip_tail",
179
+ "0",
180
+ "--dblock_loop_layers",
181
+ "0",
182
+ "--dblock_loop_start",
183
+ "0",
184
+ "--dblock_loop_cond_scale",
185
+ "1.0",
186
+ "--dblock_sigma_curriculum_steps",
187
+ "2000",
188
+ "--dblock_sigma_sampling",
189
+ "lognormal",
190
+ "--dblock_sigma_stratified",
191
+ "--dblock_sigma_min",
192
+ "0.002",
193
+ "--dblock_sigma_max",
194
+ "2.0",
195
+ "--dblock_sigma_pmean",
196
+ "-1.2",
197
+ "--dblock_sigma_pstd",
198
+ "1.2",
199
+ "--dblock_edm_wmax",
200
+ "5.0",
201
+ "--dblock_ar_prob",
202
+ "0.8",
203
+ "--dblock_sat_prob",
204
+ "0.1",
205
+ "--dblock_nat_prob",
206
+ "0.1",
207
+ "--dblock_ar_loss_tokens",
208
+ "8192",
209
+ "--dblock_sat_loss_tokens",
210
+ "4096",
211
+ "--dblock_nat_loss_tokens",
212
+ "4096",
213
+ "--dblock_nat_embed_noise_mode",
214
+ "off",
215
+ "--dblock_nat_embed_noise_scale",
216
+ "1.0",
217
+ "--nat_loss_weight",
218
+ "1.0",
219
+ "--nat_mask_ratio",
220
+ "0.5",
221
+ "--nat_max_tokens",
222
+ "0",
223
+ "--nat_span_mask_prob",
224
+ "0.35",
225
+ "--nat_suffix_mask_prob",
226
+ "0.20",
227
+ "--nat_span_max_tokens",
228
+ "0",
229
+ "--loss_spike_skip",
230
+ "0.0",
231
+ "--sat_every",
232
+ "1",
233
+ "--nat_every",
234
+ "1",
235
+ "--dblock_fullstack_anchor_spike_skip",
236
+ "0.0",
237
+ "--dblock_fullstack_anchor_max_ce",
238
+ "0.0",
239
+ "--dblock_anchor_softcap_mult",
240
+ "4.0",
241
+ "--dblock_anchor_softcap_min_ce",
242
+ "25.0",
243
+ "--dblock_anchor_softcap_max_ce",
244
+ "50.0",
245
+ "--dblock_fullstack_ar_every",
246
+ "3",
247
+ "--dblock_fullstack_ar_offset",
248
+ "0",
249
+ "--dblock_fullstack_ar_tokens",
250
+ "256",
251
+ "--dblock_fullstack_sat_every",
252
+ "3",
253
+ "--dblock_fullstack_sat_offset",
254
+ "1",
255
+ "--dblock_fullstack_sat_tokens",
256
+ "128",
257
+ "--dblock_fullstack_nat_every",
258
+ "3",
259
+ "--dblock_fullstack_nat_offset",
260
+ "2",
261
+ "--dblock_fullstack_nat_tokens",
262
+ "128",
263
+ "--dblock_fullstack_nat_mask_id",
264
+ "-1",
265
+ "--ckpt_codec",
266
+ "block-sharded-zstd",
267
+ "--async_update_every_steps",
268
+ "0",
269
+ "--dblock_router_hidden",
270
+ "64",
271
+ "--dblock_router_heads",
272
+ "4",
273
+ "--dblock_router_layers",
274
+ "2",
275
+ "--dblock_router_lr",
276
+ "0.0005",
277
+ "--dblock_explore",
278
+ "0.08",
279
+ "--dblock_max_stale_steps",
280
+ "64",
281
+ "--dblock_max_count_skew",
282
+ "1.35",
283
+ "--dblock_stale_bonus",
284
+ "0.35",
285
+ "--dblock_undertrain_bonus",
286
+ "0.25",
287
+ "--dblock_activation_offload_min_mb",
288
+ "1.0",
289
+ "--nat_document_boundary_aware",
290
+ "--dblock_fullstack_anchor_deterministic_eval",
291
+ "--dblock_fullstack_anchor_spike_retry_limit",
292
+ "3",
293
+ "--dblock_local_spike_retry_limit",
294
+ "3",
295
+ "--dblock_fullstack_nat_no_valid_crop_limit",
296
+ "8",
297
+ "--target_tokens",
298
+ "400000000000",
299
+ "--lr_schedule_reanchor_on_resume",
300
+ "--save_every_sec",
301
+ "1800",
302
+ "--delta_max_keep",
303
+ "0",
304
+ "--max_ckpts",
305
+ "2",
306
+ "--disk_free_floor_gb",
307
+ "12.0",
308
+ "--heartbeat_every_sec",
309
+ "30",
310
+ "--dblock_fullstack_ar_weight",
311
+ "1.0",
312
+ "--dblock_fullstack_sat_weight",
313
+ "0.1",
314
+ "--dblock_fullstack_nat_weight",
315
+ "0.1",
316
+ "--dblock_local_aux_warn_ce",
317
+ "0.0",
318
+ "--dblock_local_aux_max_ce",
319
+ "0.0",
320
+ "--lr_schedule_origin_tokens",
321
+ "85002387456",
322
+ "--dblock_stop_after_commits",
323
+ "0",
324
+ "--no-oom_auto_backoff",
325
+ "--require_exact_resume_state",
326
+ "--dblock_objective_mode",
327
+ "committed_9_1_1",
328
+ "--dblock_ar_weight",
329
+ "0.25",
330
+ "--dblock_sat_weight",
331
+ "0.25",
332
+ "--dblock_nat_weight",
333
+ "0.25",
334
+ "--dblock_aux_grad_ratio",
335
+ "0.5",
336
+ "--dblock_min_ar_share",
337
+ "0.8",
338
+ "--lr_core",
339
+ "2e-7",
340
+ "--lr_head",
341
+ "2e-7",
342
+ "--dblock_checkpoint_stride",
343
+ "2",
344
+ "--empty_cache_every_steps",
345
+ "0",
346
+ "--dblock_log_every",
347
+ "25",
348
+ "--cuda_max_reserved_mib",
349
+ "0",
350
+ "--cuda_min_free_mib",
351
+ "0",
352
+ "--repo",
353
+ "--repo_retrofit",
354
+ "--repo_start_layer",
355
+ "9",
356
+ "--repo_dim",
357
+ "160",
358
+ "--repo_theta",
359
+ "500000.0",
360
+ "--repo_lr",
361
+ "1e-5",
362
+ "--dblock_clean_context",
363
+ "e2e",
364
+ "--dblock_local_rows",
365
+ "2",
366
+ "--dblock_clean_context_noise_scale",
367
+ "0.0",
368
+ "--dblock_depth_anchor_every",
369
+ "50",
370
+ "--dblock_fullstack_anchor_cap",
371
+ "256",
372
+ "--dblock_fullstack_ar_rows",
373
+ "1",
374
+ "--detachable_50m_upgrade_multimode",
375
+ "--detachable_50m_upgrade_logit_distill",
376
+ "--detachable_50m_upgrade_sat_gate_v2713",
377
+ "--detachable_50m_create_from_legacy",
378
+ "--detachable_50m_seed_from_mature",
379
+ "--detachable_50m_lr",
380
+ "0.0001",
381
+ "--detachable_50m_weight_decay",
382
+ "0.0",
383
+ "--detachable_50m_update_mode",
384
+ "sublayer",
385
+ "--detachable_50m_batch",
386
+ "1",
387
+ "--detachable_50m_tokens",
388
+ "64",
389
+ "--detachable_50m_loss_tokens",
390
+ "64",
391
+ "--detachable_50m_every",
392
+ "1",
393
+ "--detachable_50m_teacher_every",
394
+ "1",
395
+ "--detachable_50m_distill_weight",
396
+ "1.0",
397
+ "--detachable_50m_logit_distill_alpha",
398
+ "0.30",
399
+ "--detachable_50m_logit_distill_temperature",
400
+ "1.0",
401
+ "--detachable_50m_logit_distill_tokens",
402
+ "64",
403
+ "--detachable_50m_hidden_distill_weight",
404
+ "0.0",
405
+ "--detachable_50m_anchor_every",
406
+ "128",
407
+ "--detachable_50m_anchor_tokens",
408
+ "32",
409
+ "--detachable_50m_seed",
410
+ "20260830",
411
+ "--detachable_50m_log_every",
412
+ "10",
413
+ "--detachable_50m_telemetry_keep",
414
+ "128",
415
+ "--weight_accel_mode",
416
+ "off",
417
+ "--weight_accel_dir",
418
+ "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/weight_trajectory_accel",
419
+ "--weight_accel_every",
420
+ "1",
421
+ "--weight_accel_train_every",
422
+ "4",
423
+ "--weight_accel_record_every",
424
+ "14",
425
+ "--weight_accel_state_every",
426
+ "256",
427
+ "--detachable_50m",
428
+ "--detachable_50m_optimizer",
429
+ "paged_adamw8bit",
430
+ "--delta_every_steps",
431
+ "0",
432
+ "--delta_every_sec",
433
+ "0",
434
+ "--resume",
435
+ "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt",
436
+ "--save_dir",
437
+ "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B",
438
+ "--data_seed",
439
+ "2852078",
440
+ "--batch_size",
441
+ "56",
442
+ "--dblock_blocks",
443
+ "14",
444
+ "--dblock_direct56_enabled",
445
+ "1",
446
+ "--dblock_layers_per_update",
447
+ "0",
448
+ "--dblock_router",
449
+ "shadow",
450
+ "--dblock_router_blend",
451
+ "0.0",
452
+ "--dblock_router_full_anchor_every",
453
+ "128",
454
+ "--dblock_router_train_layer_policy",
455
+ "cyclic",
456
+ "--dblock_router_train_layers_per_update",
457
+ "1",
458
+ "--dblock_sublayer_mode",
459
+ "off",
460
+ "--dblock_target_route_activate_from_checkpoint",
461
+ "1",
462
+ "--dblock_target_route_activation_checkpoint_step",
463
+ "2717509",
464
+ "--dblock_target_route_activation_parent_pointer_sha256",
465
+ "04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827",
466
+ "--dblock_target_route_offset",
467
+ "1",
468
+ "--dblock_target_route_policy",
469
+ "xor_fold56_v1",
470
+ "--dblock_train_layer_policy",
471
+ "cyclic",
472
+ "--dblock_train_layers_per_update",
473
+ "1",
474
+ "--dblock_train_sublayer_policy",
475
+ "cyclic",
476
+ "--dblock_train_sublayers_per_update",
477
+ "1",
478
+ "--granularity_label",
479
+ "one-sublayer-local-grad-direct56-teacher-v27.16",
480
+ "--oom_known_safe_batch",
481
+ "56",
482
+ "--val_every_sec",
483
+ "600",
484
+ "--val_file",
485
+ "/workspace/agillm43_heldout_v3/token_ids.json",
486
+ "--val_sha256",
487
+ "513b30203ffd18167655deb2d9cd610541df87d1ea0b8a0f9241c58e68703d21"
488
+ ],
489
+ "argv_string": "/workspace/agillm44_throughput_80k_20260910/runtime_retain_activations_v2.py train --preset agillm4_floor --block 2048 --source HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25 --nat_mask_token_id 2 --tie_weights --tie_kv --moe_ffn --moe_experts 2 --moe_top_k 1 --moe_mlp_mult 4 --moe_shared_experts 1 --moe_shared_mlp_mult 2 --moe_aux_coef 0.01 --moe_z_coef 0.001 --amp --grad_checkpoint --attn_backend sdpa --sublinear_window 128 --sublinear_stride 128 --sublinear_max_anchors 128 --sublinear_chunk 128 --sublinear_sinks 4 --sublinear_recent_anchors 64 --no-sublinear_pooled_landmarks --anchor_stride 256 --anchor_max 2048 --anchor_position -1 --alibi_mode corrected --alibi_scale 0.0 --optimizer adamw8bit --weight_decay 0.0 --lr_decay cosine --lr_decay_tokens 49000954880 --lr_min_mult 0.2 --lr_schedule_reset_on_resume --lr_warmup_tokens 10000000 --lr_warmup_min_mult 0.1 --dblock --dblock_schedule roundrobin --dblock_router_ramp_steps 64 --dblock_warmup_steps 14 --dblock_checkpoint_skip_tail 0 --dblock_loop_layers 0 --dblock_loop_start 0 --dblock_loop_cond_scale 1.0 --dblock_sigma_curriculum_steps 2000 --dblock_sigma_sampling lognormal --dblock_sigma_stratified --dblock_sigma_min 0.002 --dblock_sigma_max 2.0 --dblock_sigma_pmean -1.2 --dblock_sigma_pstd 1.2 --dblock_edm_wmax 5.0 --dblock_ar_prob 0.8 --dblock_sat_prob 0.1 --dblock_nat_prob 0.1 --dblock_ar_loss_tokens 8192 --dblock_sat_loss_tokens 4096 --dblock_nat_loss_tokens 4096 --dblock_nat_embed_noise_mode off --dblock_nat_embed_noise_scale 1.0 --nat_loss_weight 1.0 --nat_mask_ratio 0.5 --nat_max_tokens 0 --nat_span_mask_prob 0.35 --nat_suffix_mask_prob 0.20 --nat_span_max_tokens 0 --loss_spike_skip 0.0 --sat_every 1 --nat_every 1 --dblock_fullstack_anchor_spike_skip 0.0 --dblock_fullstack_anchor_max_ce 0.0 --dblock_anchor_softcap_mult 4.0 --dblock_anchor_softcap_min_ce 25.0 --dblock_anchor_softcap_max_ce 50.0 --dblock_fullstack_ar_every 3 --dblock_fullstack_ar_offset 0 --dblock_fullstack_ar_tokens 256 --dblock_fullstack_sat_every 3 --dblock_fullstack_sat_offset 1 --dblock_fullstack_sat_tokens 128 --dblock_fullstack_nat_every 3 --dblock_fullstack_nat_offset 2 --dblock_fullstack_nat_tokens 128 --dblock_fullstack_nat_mask_id -1 --ckpt_codec block-sharded-zstd --async_update_every_steps 0 --dblock_router_hidden 64 --dblock_router_heads 4 --dblock_router_layers 2 --dblock_router_lr 0.0005 --dblock_explore 0.08 --dblock_max_stale_steps 64 --dblock_max_count_skew 1.35 --dblock_stale_bonus 0.35 --dblock_undertrain_bonus 0.25 --dblock_activation_offload_min_mb 1.0 --nat_document_boundary_aware --dblock_fullstack_anchor_deterministic_eval --dblock_fullstack_anchor_spike_retry_limit 3 --dblock_local_spike_retry_limit 3 --dblock_fullstack_nat_no_valid_crop_limit 8 --target_tokens 400000000000 --lr_schedule_reanchor_on_resume --save_every_sec 1800 --delta_max_keep 0 --max_ckpts 2 --disk_free_floor_gb 12.0 --heartbeat_every_sec 30 --dblock_fullstack_ar_weight 1.0 --dblock_fullstack_sat_weight 0.1 --dblock_fullstack_nat_weight 0.1 --dblock_local_aux_warn_ce 0.0 --dblock_local_aux_max_ce 0.0 --lr_schedule_origin_tokens 85002387456 --dblock_stop_after_commits 0 --no-oom_auto_backoff --require_exact_resume_state --dblock_objective_mode committed_9_1_1 --dblock_ar_weight 0.25 --dblock_sat_weight 0.25 --dblock_nat_weight 0.25 --dblock_aux_grad_ratio 0.5 --dblock_min_ar_share 0.8 --lr_core 2e-7 --lr_head 2e-7 --dblock_checkpoint_stride 2 --empty_cache_every_steps 0 --dblock_log_every 25 --cuda_max_reserved_mib 0 --cuda_min_free_mib 0 --repo --repo_retrofit --repo_start_layer 9 --repo_dim 160 --repo_theta 500000.0 --repo_lr 1e-5 --dblock_clean_context e2e --dblock_local_rows 2 --dblock_clean_context_noise_scale 0.0 --dblock_depth_anchor_every 50 --dblock_fullstack_anchor_cap 256 --dblock_fullstack_ar_rows 1 --detachable_50m_upgrade_multimode --detachable_50m_upgrade_logit_distill --detachable_50m_upgrade_sat_gate_v2713 --detachable_50m_create_from_legacy --detachable_50m_seed_from_mature --detachable_50m_lr 0.0001 --detachable_50m_weight_decay 0.0 --detachable_50m_update_mode sublayer --detachable_50m_batch 1 --detachable_50m_tokens 64 --detachable_50m_loss_tokens 64 --detachable_50m_every 1 --detachable_50m_teacher_every 1 --detachable_50m_distill_weight 1.0 --detachable_50m_logit_distill_alpha 0.30 --detachable_50m_logit_distill_temperature 1.0 --detachable_50m_logit_distill_tokens 64 --detachable_50m_hidden_distill_weight 0.0 --detachable_50m_anchor_every 128 --detachable_50m_anchor_tokens 32 --detachable_50m_seed 20260830 --detachable_50m_log_every 10 --detachable_50m_telemetry_keep 128 --weight_accel_mode off --weight_accel_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/weight_trajectory_accel --weight_accel_every 1 --weight_accel_train_every 4 --weight_accel_record_every 14 --weight_accel_state_every 256 --detachable_50m --detachable_50m_optimizer paged_adamw8bit --delta_every_steps 0 --delta_every_sec 0 --resume /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt --save_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B --data_seed 2852078 --batch_size 56 --dblock_blocks 14 --dblock_direct56_enabled 1 --dblock_layers_per_update 0 --dblock_router shadow --dblock_router_blend 0.0 --dblock_router_full_anchor_every 128 --dblock_router_train_layer_policy cyclic --dblock_router_train_layers_per_update 1 --dblock_sublayer_mode off --dblock_target_route_activate_from_checkpoint 1 --dblock_target_route_activation_checkpoint_step 2717509 --dblock_target_route_activation_parent_pointer_sha256 04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827 --dblock_target_route_offset 1 --dblock_target_route_policy xor_fold56_v1 --dblock_train_layer_policy cyclic --dblock_train_layers_per_update 1 --dblock_train_sublayer_policy cyclic --dblock_train_sublayers_per_update 1 --granularity_label one-sublayer-local-grad-direct56-teacher-v27.16 --oom_known_safe_batch 56 --val_every_sec 600 --val_file /workspace/agillm43_heldout_v3/token_ids.json --val_sha256 513b30203ffd18167655deb2d9cd610541df87d1ea0b8a0f9241c58e68703d21",
490
+ "cwd": "/workspace/agillm44_throughput_80k_20260910",
491
+ "env": {
492
+ "AGILLM43_DBLOCK_CC_NOISE_SCALE": "0.0",
493
+ "AGILLM43_DBLOCK_CLEAN_CONTEXT": "e2e",
494
+ "AGILLM43_DBLOCK_CLEAN_CONTEXT_NOISE_SCALE": "0.0",
495
+ "AGILLM43_DBLOCK_DEPTH_ANCHOR_EVERY": "50",
496
+ "AGILLM43_DBLOCK_FULLSTACK_ANCHOR_CAP": "256",
497
+ "AGILLM43_DBLOCK_FULLSTACK_AR_ROWS": "1",
498
+ "AGILLM43_DBLOCK_LOCAL_ROWS": "2",
499
+ "AGILLM43_DBLOCK_TARGET_ROUTE_ACTIVATE_FROM_CHECKPOINT": "1",
500
+ "AGILLM43_DBLOCK_TARGET_ROUTE_ACTIVATION_CHECKPOINT_STEP": "2717509",
501
+ "AGILLM43_DBLOCK_TARGET_ROUTE_ACTIVATION_PARENT_POINTER_SHA256": "04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827",
502
+ "AGILLM43_DBLOCK_TARGET_ROUTE_OFFSET": "1",
503
+ "AGILLM43_DBLOCK_TARGET_ROUTE_POLICY": "xor_fold56_v1",
504
+ "AGILLM43_DECOMPRESS_CACHE": "0",
505
+ "AGILLM43_DETACHABLE50M_ANCHOR_EVERY": "128",
506
+ "AGILLM43_DETACHABLE50M_ANCHOR_TOKENS": "<redacted>",
507
+ "AGILLM43_DETACHABLE50M_BATCH": "1",
508
+ "AGILLM43_DETACHABLE50M_CREATE_FROM_LEGACY": "1",
509
+ "AGILLM43_DETACHABLE50M_DISTILL_WEIGHT": "1.0",
510
+ "AGILLM43_DETACHABLE50M_ENABLED": "1",
511
+ "AGILLM43_DETACHABLE50M_EVERY": "1",
512
+ "AGILLM43_DETACHABLE50M_GATE_V2713_START_CHECKPOINT_SHA256": "39f442fc0e6f9fad6f875e067f63db67f33a41a53e8cbc49ce1591010bb4aec3",
513
+ "AGILLM43_DETACHABLE50M_HIDDEN_DISTILL_WEIGHT": "0.0",
514
+ "AGILLM43_DETACHABLE50M_LOGIT_DISTILL_ALPHA": "0.30",
515
+ "AGILLM43_DETACHABLE50M_LOGIT_DISTILL_TEMPERATURE": "1.0",
516
+ "AGILLM43_DETACHABLE50M_LOGIT_DISTILL_TOKENS": "<redacted>",
517
+ "AGILLM43_DETACHABLE50M_LOG_EVERY": "10",
518
+ "AGILLM43_DETACHABLE50M_LOSS_TOKENS": "<redacted>",
519
+ "AGILLM43_DETACHABLE50M_LR": "0.0001",
520
+ "AGILLM43_DETACHABLE50M_OPTIMIZER": "paged_adamw8bit",
521
+ "AGILLM43_DETACHABLE50M_SEED": "20260830",
522
+ "AGILLM43_DETACHABLE50M_SEED_FROM_MATURE": "1",
523
+ "AGILLM43_DETACHABLE50M_TEACHER_EVERY": "1",
524
+ "AGILLM43_DETACHABLE50M_TELEMETRY_KEEP": "128",
525
+ "AGILLM43_DETACHABLE50M_TOKENS": "<redacted>",
526
+ "AGILLM43_DETACHABLE50M_UPDATE_MODE": "sublayer",
527
+ "AGILLM43_DETACHABLE50M_UPGRADE_LOGIT_DISTILL": "1",
528
+ "AGILLM43_DETACHABLE50M_UPGRADE_MULTIMODE": "1",
529
+ "AGILLM43_DETACHABLE50M_UPGRADE_SAT_GATE_V2713": "1",
530
+ "AGILLM43_DETACHABLE50M_WEIGHT_DECAY": "0.0",
531
+ "AGILLM43_REPO_DIM": "160",
532
+ "AGILLM43_REPO_ENABLED": "1",
533
+ "AGILLM43_REPO_LR": "1e-5",
534
+ "AGILLM43_REPO_RETROFIT": "1",
535
+ "AGILLM43_REPO_START_LAYER": "9",
536
+ "AGILLM43_REPO_THETA": "500000.0",
537
+ "AGILLM43_SEQUENCE_PREFETCH_DEPTH": "256",
538
+ "AGILLM43_SINGLEFILE_PROFILE": "intelligence-v25-satvar-pretrain",
539
+ "AGILLM43_SUPERVISE_SET_ARGS_JSON": "{\"--batch_size\":\"56\",\"--dblock_blocks\":\"14\",\"--dblock_direct56_enabled\":\"1\",\"--dblock_layers_per_update\":\"0\",\"--dblock_router\":\"shadow\",\"--dblock_router_blend\":\"0.0\",\"--dblock_router_full_anchor_every\":\"128\",\"--dblock_router_train_layer_policy\":\"cyclic\",\"--dblock_router_train_layers_per_update\":\"1\",\"--dblock_sublayer_mode\":\"off\",\"--dblock_target_route_activate_from_checkpoint\":\"1\",\"--dblock_target_route_activation_checkpoint_step\":\"2717509\",\"--dblock_target_route_activation_parent_pointer_sha256\":\"04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827\",\"--dblock_target_route_offset\":\"1\",\"--dblock_target_route_policy\":\"xor_fold56_v1\",\"--dblock_train_layer_policy\":\"cyclic\",\"--dblock_train_layers_per_update\":\"1\",\"--dblock_train_sublayer_policy\":\"cyclic\",\"--dblock_train_sublayers_per_update\":\"1\",\"--granularity_label\":\"one-sublayer-local-grad-direct56-teacher-v27.16\",\"--oom_known_safe_batch\":\"56\",\"--val_every_sec\":\"600\",\"--val_file\":\"/workspace/agillm43_heldout_v3/token_ids.json\",\"--val_sha256\":\"513b30203ffd18167655deb2d9cd610541df87d1ea0b8a0f9241c58e68703d21\"}",
540
+ "AGILLM43_WEIGHT_ACCEL_MODE": "off",
541
+ "AGILLM_DATASETS_STREAMING_READ_MAX_RETRIES": "2",
542
+ "AGILLM_DATASETS_STREAMING_READ_RETRY_INTERVAL": "1",
543
+ "AGILLM_DATASET_AGENT_ROUTER": "0",
544
+ "AGILLM_DATASET_HOTLOAD_CONFIG": "/workspace/agillm43_dataset_hotload.json",
545
+ "AGILLM_DATASET_HOT_RELOAD_CHECK_TOKENS": "<redacted>",
546
+ "AGILLM_DATASET_HOT_RELOAD_SEC": "5",
547
+ "AGILLM_DATASET_NN_ROUTER": "1",
548
+ "AGILLM_DATASET_ROUTER_BLEND": "0.25",
549
+ "AGILLM_DATASET_ROUTER_EXPLORE": "0.03",
550
+ "AGILLM_DATASET_ROUTER_LR": "0.005",
551
+ "AGILLM_DATASET_ROUTER_MIN_SCORE": "0.15",
552
+ "AGILLM_DATASET_ROUTER_SHARPNESS": "2.0",
553
+ "AGILLM_ENABLE_CUPY": "0",
554
+ "AGILLM_HF_HUB_DOWNLOAD_TIMEOUT": "30",
555
+ "AGILLM_HF_TRANSPORT": "geth_hf_relay",
556
+ "AGILLM_HOT_CONFIG": "/workspace/agillm43_production_v14/contracts/hot_config_owner_v25.json",
557
+ "AGILLM_HOT_CONFIG_SHA256": "",
558
+ "AGILLM_MAX_EXAMPLE_CHARS": "32768",
559
+ "AGILLM_MAX_EXAMPLE_TOKENS": "<redacted>",
560
+ "AGILLM_PARENT_POINTER_SHA256": "0a42102b9ce5e3b54f4c8a94ff8e127b948246a14fb5faf36e2c364320837d9f",
561
+ "AGILLM_REMOTE_SHARDS_PER_ITER": "4",
562
+ "AGILLM_RUN_ID": "singlefile_intelligence_v25_20260911T121655Z_from_step000002852036",
563
+ "AGILLM_STREAM_NEXT_TIMEOUT_SEC": "60",
564
+ "AGILLM_STREAM_OPEN_TIMEOUT_SEC": "120",
565
+ "AGILLM_STREAM_SOCKET_TIMEOUT_SEC": "30",
566
+ "AGILLM_STREAM_SOURCE_FATAL_COOLDOWN_SEC": "600",
567
+ "AGILLM_STREAM_SOURCE_MAX_COOLDOWN_SEC": "120",
568
+ "AGILLM_SYNTHETIC_TOKENIZER": "<redacted>",
569
+ "CUDA_MODULE_LOADING": "LAZY",
570
+ "CUDA_VERSION": "12.1.1",
571
+ "CUDA_VISIBLE_DEVICES": "0",
572
+ "HF_HUB_DOWNLOAD_TIMEOUT": "30",
573
+ "HF_HUB_ETAG_TIMEOUT": "10",
574
+ "PYTORCH_CUDA_ALLOC_CONF": "backend:cudaMallocAsync",
575
+ "PYTORCH_VERSION": "2.2.1"
576
+ },
577
+ "parent_proc_cmdline": "/opt/conda/bin/python -u /workspace/agillm44_throughput_80k_20260910/runtime_retain_activations_v2.py supervise --save_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B --log /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/logs/singlefile_intelligence_v25_satvar_pretrain.log --sleep_sec 15",
578
+ "pid": 1937223,
579
+ "ppid": 1937141,
580
+ "proc_cmdline": "/opt/conda/bin/python -u /workspace/agillm44_throughput_80k_20260910/runtime_retain_activations_v2.py train --preset agillm4_floor --block 2048 --source HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25 --nat_mask_token_id 2 --tie_weights --tie_kv --moe_ffn --moe_experts 2 --moe_top_k 1 --moe_mlp_mult 4 --moe_shared_experts 1 --moe_shared_mlp_mult 2 --moe_aux_coef 0.01 --moe_z_coef 0.001 --amp --grad_checkpoint --attn_backend sdpa --sublinear_window 128 --sublinear_stride 128 --sublinear_max_anchors 128 --sublinear_chunk 128 --sublinear_sinks 4 --sublinear_recent_anchors 64 --no-sublinear_pooled_landmarks --anchor_stride 256 --anchor_max 2048 --anchor_position -1 --alibi_mode corrected --alibi_scale 0.0 --optimizer adamw8bit --weight_decay 0.0 --lr_decay cosine --lr_decay_tokens 49000954880 --lr_min_mult 0.2 --lr_schedule_reset_on_resume --lr_warmup_tokens 10000000 --lr_warmup_min_mult 0.1 --dblock --dblock_schedule roundrobin --dblock_router_ramp_steps 64 --dblock_warmup_steps 14 --dblock_checkpoint_skip_tail 0 --dblock_loop_layers 0 --dblock_loop_start 0 --dblock_loop_cond_scale 1.0 --dblock_sigma_curriculum_steps 2000 --dblock_sigma_sampling lognormal --dblock_sigma_stratified --dblock_sigma_min 0.002 --dblock_sigma_max 2.0 --dblock_sigma_pmean -1.2 --dblock_sigma_pstd 1.2 --dblock_edm_wmax 5.0 --dblock_ar_prob 0.8 --dblock_sat_prob 0.1 --dblock_nat_prob 0.1 --dblock_ar_loss_tokens 8192 --dblock_sat_loss_tokens 4096 --dblock_nat_loss_tokens 4096 --dblock_nat_embed_noise_mode off --dblock_nat_embed_noise_scale 1.0 --nat_loss_weight 1.0 --nat_mask_ratio 0.5 --nat_max_tokens 0 --nat_span_mask_prob 0.35 --nat_suffix_mask_prob 0.20 --nat_span_max_tokens 0 --loss_spike_skip 0.0 --sat_every 1 --nat_every 1 --dblock_fullstack_anchor_spike_skip 0.0 --dblock_fullstack_anchor_max_ce 0.0 --dblock_anchor_softcap_mult 4.0 --dblock_anchor_softcap_min_ce 25.0 --dblock_anchor_softcap_max_ce 50.0 --dblock_fullstack_ar_every 3 --dblock_fullstack_ar_offset 0 --dblock_fullstack_ar_tokens 256 --dblock_fullstack_sat_every 3 --dblock_fullstack_sat_offset 1 --dblock_fullstack_sat_tokens 128 --dblock_fullstack_nat_every 3 --dblock_fullstack_nat_offset 2 --dblock_fullstack_nat_tokens 128 --dblock_fullstack_nat_mask_id -1 --ckpt_codec block-sharded-zstd --async_update_every_steps 0 --dblock_router_hidden 64 --dblock_router_heads 4 --dblock_router_layers 2 --dblock_router_lr 0.0005 --dblock_explore 0.08 --dblock_max_stale_steps 64 --dblock_max_count_skew 1.35 --dblock_stale_bonus 0.35 --dblock_undertrain_bonus 0.25 --dblock_activation_offload_min_mb 1.0 --nat_document_boundary_aware --dblock_fullstack_anchor_deterministic_eval --dblock_fullstack_anchor_spike_retry_limit 3 --dblock_local_spike_retry_limit 3 --dblock_fullstack_nat_no_valid_crop_limit 8 --target_tokens 400000000000 --lr_schedule_reanchor_on_resume --save_every_sec 1800 --delta_max_keep 0 --max_ckpts 2 --disk_free_floor_gb 12.0 --heartbeat_every_sec 30 --dblock_fullstack_ar_weight 1.0 --dblock_fullstack_sat_weight 0.1 --dblock_fullstack_nat_weight 0.1 --dblock_local_aux_warn_ce 0.0 --dblock_local_aux_max_ce 0.0 --lr_schedule_origin_tokens 85002387456 --dblock_stop_after_commits 0 --no-oom_auto_backoff --require_exact_resume_state --dblock_objective_mode committed_9_1_1 --dblock_ar_weight 0.25 --dblock_sat_weight 0.25 --dblock_nat_weight 0.25 --dblock_aux_grad_ratio 0.5 --dblock_min_ar_share 0.8 --lr_core 2e-7 --lr_head 2e-7 --dblock_checkpoint_stride 2 --empty_cache_every_steps 0 --dblock_log_every 25 --cuda_max_reserved_mib 0 --cuda_min_free_mib 0 --repo --repo_retrofit --repo_start_layer 9 --repo_dim 160 --repo_theta 500000.0 --repo_lr 1e-5 --dblock_clean_context e2e --dblock_local_rows 2 --dblock_clean_context_noise_scale 0.0 --dblock_depth_anchor_every 50 --dblock_fullstack_anchor_cap 256 --dblock_fullstack_ar_rows 1 --detachable_50m_upgrade_multimode --detachable_50m_upgrade_logit_distill --detachable_50m_upgrade_sat_gate_v2713 --detachable_50m_create_from_legacy --detachable_50m_seed_from_mature --detachable_50m_lr 0.0001 --detachable_50m_weight_decay 0.0 --detachable_50m_update_mode sublayer --detachable_50m_batch 1 --detachable_50m_tokens 64 --detachable_50m_loss_tokens 64 --detachable_50m_every 1 --detachable_50m_teacher_every 1 --detachable_50m_distill_weight 1.0 --detachable_50m_logit_distill_alpha 0.30 --detachable_50m_logit_distill_temperature 1.0 --detachable_50m_logit_distill_tokens 64 --detachable_50m_hidden_distill_weight 0.0 --detachable_50m_anchor_every 128 --detachable_50m_anchor_tokens 32 --detachable_50m_seed 20260830 --detachable_50m_log_every 10 --detachable_50m_telemetry_keep 128 --weight_accel_mode off --weight_accel_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/weight_trajectory_accel --weight_accel_every 1 --weight_accel_train_every 4 --weight_accel_record_every 14 --weight_accel_state_every 256 --detachable_50m --detachable_50m_optimizer paged_adamw8bit --delta_every_steps 0 --delta_every_sec 0 --resume /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt --save_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B --data_seed 2852078 --batch_size 56 --dblock_blocks 14 --dblock_direct56_enabled 1 --dblock_layers_per_update 0 --dblock_router shadow --dblock_router_blend 0.0 --dblock_router_full_anchor_every 128 --dblock_router_train_layer_policy cyclic --dblock_router_train_layers_per_update 1 --dblock_sublayer_mode off --dblock_target_route_activate_from_checkpoint 1 --dblock_target_route_activation_checkpoint_step 2717509 --dblock_target_route_activation_parent_pointer_sha256 04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827 --dblock_target_route_offset 1 --dblock_target_route_policy xor_fold56_v1 --dblock_train_layer_policy cyclic --dblock_train_layers_per_update 1 --dblock_train_sublayer_policy cyclic --dblock_train_sublayers_per_update 1 --granularity_label one-sublayer-local-grad-direct56-teacher-v27.16 --oom_known_safe_batch 56 --val_every_sec 600 --val_file /workspace/agillm43_heldout_v3/token_ids.json --val_sha256 513b30203ffd18167655deb2d9cd610541df87d1ea0b8a0f9241c58e68703d21",
581
+ "schema": "agillm.launch.v1"
582
+ },
583
+ "local_step": 2865113,
584
+ "loss": 5.242419242858887,
585
+ "nat_mask_contract": "agillm.nat-mask.v1",
586
+ "nat_mask_contract_source": "explicit-legacy-recovery",
587
+ "nat_mask_eos_id": 1,
588
+ "nat_mask_legacy_blank_id": 1,
589
+ "nat_mask_migrated_from_id": 1,
590
+ "nat_mask_schema_version": 1,
591
+ "nat_mask_token_id": 2,
592
+ "pid": 1937223,
593
+ "schema_version": 1,
594
+ "seen_tok": 285479475200,
595
+ "source_is_lane_checkpoint": true,
596
+ "source_is_master_checkpoint": false,
597
+ "source_is_non_master_checkpoint": false,
598
+ "step": 2865113,
599
+ "train_argv": "/workspace/agillm44_throughput_80k_20260910/runtime_retain_activations_v2.py train --preset agillm4_floor --block 2048 --source HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25 --nat_mask_token_id 2 --tie_weights --tie_kv --moe_ffn --moe_experts 2 --moe_top_k 1 --moe_mlp_mult 4 --moe_shared_experts 1 --moe_shared_mlp_mult 2 --moe_aux_coef 0.01 --moe_z_coef 0.001 --amp --grad_checkpoint --attn_backend sdpa --sublinear_window 128 --sublinear_stride 128 --sublinear_max_anchors 128 --sublinear_chunk 128 --sublinear_sinks 4 --sublinear_recent_anchors 64 --no-sublinear_pooled_landmarks --anchor_stride 256 --anchor_max 2048 --anchor_position -1 --alibi_mode corrected --alibi_scale 0.0 --optimizer adamw8bit --weight_decay 0.0 --lr_decay cosine --lr_decay_tokens 49000954880 --lr_min_mult 0.2 --lr_schedule_reset_on_resume --lr_warmup_tokens 10000000 --lr_warmup_min_mult 0.1 --dblock --dblock_schedule roundrobin --dblock_router_ramp_steps 64 --dblock_warmup_steps 14 --dblock_checkpoint_skip_tail 0 --dblock_loop_layers 0 --dblock_loop_start 0 --dblock_loop_cond_scale 1.0 --dblock_sigma_curriculum_steps 2000 --dblock_sigma_sampling lognormal --dblock_sigma_stratified --dblock_sigma_min 0.002 --dblock_sigma_max 2.0 --dblock_sigma_pmean -1.2 --dblock_sigma_pstd 1.2 --dblock_edm_wmax 5.0 --dblock_ar_prob 0.8 --dblock_sat_prob 0.1 --dblock_nat_prob 0.1 --dblock_ar_loss_tokens 8192 --dblock_sat_loss_tokens 4096 --dblock_nat_loss_tokens 4096 --dblock_nat_embed_noise_mode off --dblock_nat_embed_noise_scale 1.0 --nat_loss_weight 1.0 --nat_mask_ratio 0.5 --nat_max_tokens 0 --nat_span_mask_prob 0.35 --nat_suffix_mask_prob 0.20 --nat_span_max_tokens 0 --loss_spike_skip 0.0 --sat_every 1 --nat_every 1 --dblock_fullstack_anchor_spike_skip 0.0 --dblock_fullstack_anchor_max_ce 0.0 --dblock_anchor_softcap_mult 4.0 --dblock_anchor_softcap_min_ce 25.0 --dblock_anchor_softcap_max_ce 50.0 --dblock_fullstack_ar_every 3 --dblock_fullstack_ar_offset 0 --dblock_fullstack_ar_tokens 256 --dblock_fullstack_sat_every 3 --dblock_fullstack_sat_offset 1 --dblock_fullstack_sat_tokens 128 --dblock_fullstack_nat_every 3 --dblock_fullstack_nat_offset 2 --dblock_fullstack_nat_tokens 128 --dblock_fullstack_nat_mask_id -1 --ckpt_codec block-sharded-zstd --async_update_every_steps 0 --dblock_router_hidden 64 --dblock_router_heads 4 --dblock_router_layers 2 --dblock_router_lr 0.0005 --dblock_explore 0.08 --dblock_max_stale_steps 64 --dblock_max_count_skew 1.35 --dblock_stale_bonus 0.35 --dblock_undertrain_bonus 0.25 --dblock_activation_offload_min_mb 1.0 --nat_document_boundary_aware --dblock_fullstack_anchor_deterministic_eval --dblock_fullstack_anchor_spike_retry_limit 3 --dblock_local_spike_retry_limit 3 --dblock_fullstack_nat_no_valid_crop_limit 8 --target_tokens 400000000000 --lr_schedule_reanchor_on_resume --save_every_sec 1800 --delta_max_keep 0 --max_ckpts 2 --disk_free_floor_gb 12.0 --heartbeat_every_sec 30 --dblock_fullstack_ar_weight 1.0 --dblock_fullstack_sat_weight 0.1 --dblock_fullstack_nat_weight 0.1 --dblock_local_aux_warn_ce 0.0 --dblock_local_aux_max_ce 0.0 --lr_schedule_origin_tokens 85002387456 --dblock_stop_after_commits 0 --no-oom_auto_backoff --require_exact_resume_state --dblock_objective_mode committed_9_1_1 --dblock_ar_weight 0.25 --dblock_sat_weight 0.25 --dblock_nat_weight 0.25 --dblock_aux_grad_ratio 0.5 --dblock_min_ar_share 0.8 --lr_core 2e-7 --lr_head 2e-7 --dblock_checkpoint_stride 2 --empty_cache_every_steps 0 --dblock_log_every 25 --cuda_max_reserved_mib 0 --cuda_min_free_mib 0 --repo --repo_retrofit --repo_start_layer 9 --repo_dim 160 --repo_theta 500000.0 --repo_lr 1e-5 --dblock_clean_context e2e --dblock_local_rows 2 --dblock_clean_context_noise_scale 0.0 --dblock_depth_anchor_every 50 --dblock_fullstack_anchor_cap 256 --dblock_fullstack_ar_rows 1 --detachable_50m_upgrade_multimode --detachable_50m_upgrade_logit_distill --detachable_50m_upgrade_sat_gate_v2713 --detachable_50m_create_from_legacy --detachable_50m_seed_from_mature --detachable_50m_lr 0.0001 --detachable_50m_weight_decay 0.0 --detachable_50m_update_mode sublayer --detachable_50m_batch 1 --detachable_50m_tokens 64 --detachable_50m_loss_tokens 64 --detachable_50m_every 1 --detachable_50m_teacher_every 1 --detachable_50m_distill_weight 1.0 --detachable_50m_logit_distill_alpha 0.30 --detachable_50m_logit_distill_temperature 1.0 --detachable_50m_logit_distill_tokens 64 --detachable_50m_hidden_distill_weight 0.0 --detachable_50m_anchor_every 128 --detachable_50m_anchor_tokens 32 --detachable_50m_seed 20260830 --detachable_50m_log_every 10 --detachable_50m_telemetry_keep 128 --weight_accel_mode off --weight_accel_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/weight_trajectory_accel --weight_accel_every 1 --weight_accel_train_every 4 --weight_accel_record_every 14 --weight_accel_state_every 256 --detachable_50m --detachable_50m_optimizer paged_adamw8bit --delta_every_steps 0 --delta_every_sec 0 --resume /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt --save_dir /workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B --data_seed 2852078 --batch_size 56 --dblock_blocks 14 --dblock_direct56_enabled 1 --dblock_layers_per_update 0 --dblock_router shadow --dblock_router_blend 0.0 --dblock_router_full_anchor_every 128 --dblock_router_train_layer_policy cyclic --dblock_router_train_layers_per_update 1 --dblock_sublayer_mode off --dblock_target_route_activate_from_checkpoint 1 --dblock_target_route_activation_checkpoint_step 2717509 --dblock_target_route_activation_parent_pointer_sha256 04638cfbc19c0f9db6e6dfa03b2509d50a0bf3c56048a2cbd6e67154bc85c827 --dblock_target_route_offset 1 --dblock_target_route_policy xor_fold56_v1 --dblock_train_layer_policy cyclic --dblock_train_layers_per_update 1 --dblock_train_sublayer_policy cyclic --dblock_train_sublayers_per_update 1 --granularity_label one-sublayer-local-grad-direct56-teacher-v27.16 --oom_known_safe_batch 56 --val_every_sec 600 --val_file /workspace/agillm43_heldout_v3/token_ids.json --val_sha256 513b30203ffd18167655deb2d9cd610541df87d1ea0b8a0f9241c58e68703d21",
600
+ "train_script": "runtime_retain_activations_v2.py",
601
+ "train_script_sha256": "989a7ee5e4344b7b53b67b70d0df27dc42461d864084283d1d7acb262e0a7081",
602
+ "warmstart_base_seen_tok": 0,
603
+ "warmstart_base_step": 2852036,
604
+ "warmstart_kind": "warmstarted_from_lane_checkpoint",
605
+ "warmstart_source_path": "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt",
606
+ "warmstart_source_step": 2852036
607
+ },
608
+ "alibi_mode": "corrected",
609
+ "alibi_scale": 0.0,
610
+ "checkpoint_summary": "warmstarted_from_lane_checkpoint; source=/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt; path=/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02865113_from02852036_20260911T1839Z.pt",
611
+ "dataset_provenance": {
612
+ "chat": false,
613
+ "dataset_field_text": "text",
614
+ "dataset_hotload_config_path": "/workspace/agillm43_dataset_hotload.json",
615
+ "dataset_hotload_config_sha256": "3f48fe33b475c719d43e14cc4d0867c19e17dfa1d2cb304cd291eea6a879a7c8",
616
+ "dataset_hotload_config_stamp": [
617
+ 50,
618
+ 1124990893,
619
+ 1789091829464680808,
620
+ 7210
621
+ ],
622
+ "dataset_hotload_config_used": true,
623
+ "dataset_hotload_enabled": true,
624
+ "dataset_hotload_generation": 13,
625
+ "dataset_hotload_schema": "agillm.dataset_hotload.config.v2",
626
+ "dataset_revision_map_sha256": "8361995a504a2365136b20e1a879ba78944a06309f13f3f013bbd0cdfc72225b",
627
+ "dataset_revisions": {
628
+ "EleutherAI/proof-pile-2": "901a9273a770e9d4138c5ddd91802f9c5c6cdc4b",
629
+ "HuggingFaceFW/fineweb": "9bb295ddab0e05d785b879661af7260fed5140fc",
630
+ "HuggingFaceFW/fineweb-edu": "87f09149ef4734204d70ed1d046ddc9ca3f2b8f9",
631
+ "HuggingFaceTB/cosmopedia": "0ae6ec63f91742bd2d1eaef4f02232c55d719385",
632
+ "HuggingFaceTB/finemath": "e92b25a616738fe95dc186b64dfb19f9c8525594",
633
+ "HuggingFaceTB/smollm-corpus": "3ba9d605774198c5868892d7a8deda78031a781f",
634
+ "Skylion007/openwebtext": "79d93d786212f7344586290adb811d4ae6a1762c",
635
+ "allenai/c4": "1588ec454efa1a09f29cd18ddd04fe05fc8653a2",
636
+ "allenai/dolma": "7f48140530a023e9ea4c5cfb141160922727d4d3",
637
+ "allenai/dolmino-mix-1124": "a319f19eef1e257417b11ea8c30da266ae175557",
638
+ "codeparrot/codeparrot-clean": "35a59fb025bc0a102f7d96eac09d145b896d487b",
639
+ "mlfoundations/dclm-baseline-1.0": "a3b142c183aebe5af344955ae20836eb34dcf69b",
640
+ "wikimedia/wikipedia": "b04c8d1ceb2f5cd4588862100d08de323dccfbaa"
641
+ },
642
+ "dolma_payload_integrity": "external_url_manifest_pinned_payload_hashes_unavailable",
643
+ "dolma_url_list_sha256": "bc6ef505be920c94bb9e7147445638f606fdc3f5af622cd1d887dc48e21a2d5b",
644
+ "dolma_url_manifest_raw_sha256": "6fe8c40d51e5bebc9303453d7f533b232d69f7ae122123621a82621b1cd779b7",
645
+ "has_language_mix": true,
646
+ "has_numeracy": false,
647
+ "hot_config_expected_sha256": "",
648
+ "hot_config_has_append": true,
649
+ "hot_config_has_datasets": false,
650
+ "hot_config_mtime": 1789092253.2119184,
651
+ "hot_config_path": "/workspace/agillm43_production_v14/contracts/hot_config_owner_v25.json",
652
+ "hot_config_used": true,
653
+ "hot_reload_capable": true,
654
+ "phase": "pretrain",
655
+ "proof_pile_lfs_manifest_sha256": "d020dfb796b02a75564e4e11d9e03185a55b3f9c6639f87830cfbe10737ed329",
656
+ "proof_pile_paths_sha256": "8a30018017e3b3d9a9bcfa06a3c404ccd113297b0b1d0c48e2e2aadf113e52b0",
657
+ "schema": "agillm.dataset_provenance.v1",
658
+ "sequence_prefetch_depth": 256,
659
+ "source_count": 23,
660
+ "source_effective": "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96,wikimedia/wikipedia:20231101.en|1.63,allenai/c4:en|0.36,Skylion007/openwebtext|0.18,HuggingFaceTB/cosmopedia:web_samples_v2|0.64,HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65,EleutherAI/proof-pile-2:all|0.67,allenai/dolma:v1_6-sample|0.5,codeparrot/codeparrot-clean|0.24,HuggingFaceTB/finemath:finemath-4plus|0.32,HuggingFaceTB/finemath:infiwebmath-4plus|0.23,HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05,allenai/dolmino-mix-1124|0.78,mlfoundations/dclm-baseline-1.0|1.05,HuggingFaceFW/fineweb-edu:sample-100BT|2.4,json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35",
661
+ "source_requested": "HuggingFaceFW/fineweb-edu:sample-10BT|1.60,HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.50,wikimedia/wikipedia:20231101.en|1.10,allenai/c4:en|0.40,Skylion007/openwebtext|0.35,HuggingFaceTB/cosmopedia:web_samples_v2|1.30,HuggingFaceTB/smollm-corpus:cosmopedia-v2|1.30,EleutherAI/proof-pile-2:all|1.50,allenai/dolma:v1_6-sample|0.50,codeparrot/codeparrot-clean|1.25",
662
+ "source_specs": [
663
+ "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96",
664
+ "wikimedia/wikipedia:20231101.en|1.63",
665
+ "allenai/c4:en|0.36",
666
+ "Skylion007/openwebtext|0.18",
667
+ "HuggingFaceTB/cosmopedia:web_samples_v2|0.64",
668
+ "HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65",
669
+ "EleutherAI/proof-pile-2:all|0.67",
670
+ "allenai/dolma:v1_6-sample|0.5",
671
+ "codeparrot/codeparrot-clean|0.24",
672
+ "HuggingFaceTB/finemath:finemath-4plus|0.32",
673
+ "HuggingFaceTB/finemath:infiwebmath-4plus|0.23",
674
+ "HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05",
675
+ "allenai/dolmino-mix-1124|0.78",
676
+ "mlfoundations/dclm-baseline-1.0|1.05",
677
+ "HuggingFaceFW/fineweb-edu:sample-100BT|2.4",
678
+ "json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03",
679
+ "json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45",
680
+ "json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35",
681
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35",
682
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35",
683
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2",
684
+ "json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2",
685
+ "json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35"
686
+ ],
687
+ "val_source_effective": "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96,wikimedia/wikipedia:20231101.en|1.63,allenai/c4:en|0.36,Skylion007/openwebtext|0.18,HuggingFaceTB/cosmopedia:web_samples_v2|0.64,HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65,EleutherAI/proof-pile-2:all|0.67,allenai/dolma:v1_6-sample|0.5,codeparrot/codeparrot-clean|0.24,HuggingFaceTB/finemath:finemath-4plus|0.32,HuggingFaceTB/finemath:infiwebmath-4plus|0.23,HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05,allenai/dolmino-mix-1124|0.78,mlfoundations/dclm-baseline-1.0|1.05,HuggingFaceFW/fineweb-edu:sample-100BT|2.4,json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35",
688
+ "val_source_requested": ""
689
+ },
690
+ "lr_schedule_origin_tok": 85002387456,
691
+ "nat_mask_contract": "agillm.nat-mask.v1",
692
+ "nat_mask_contract_source": "explicit-legacy-recovery",
693
+ "nat_mask_eos_id": 1,
694
+ "nat_mask_legacy_blank_id": 1,
695
+ "nat_mask_migrated_from_id": 1,
696
+ "nat_mask_schema_version": 1,
697
+ "nat_mask_token_id": 2,
698
+ "path": "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02865113_from02852036_20260911T1839Z.pt",
699
+ "promotion_required": true,
700
+ "quality_gate_used_for_training": false,
701
+ "repair_schema": "",
702
+ "saved_at": "2026-09-11T18:40:10Z",
703
+ "schema": "agillm43.serving.candidate.v1",
704
+ "source_effective": "HuggingFaceFW/fineweb:CC-MAIN-2024-10|0.96,wikimedia/wikipedia:20231101.en|1.63,allenai/c4:en|0.36,Skylion007/openwebtext|0.18,HuggingFaceTB/cosmopedia:web_samples_v2|0.64,HuggingFaceTB/smollm-corpus:cosmopedia-v2|0.65,EleutherAI/proof-pile-2:all|0.67,allenai/dolma:v1_6-sample|0.5,codeparrot/codeparrot-clean|0.24,HuggingFaceTB/finemath:finemath-4plus|0.32,HuggingFaceTB/finemath:infiwebmath-4plus|0.23,HuggingFaceTB/smollm-corpus:fineweb-edu-dedup|1.05,allenai/dolmino-mix-1124|0.78,mlfoundations/dclm-baseline-1.0|1.05,HuggingFaceFW/fineweb-edu:sample-100BT|2.4,json:/workspace/agillm_wikitext103_raw_docs/train.jsonl|0.03,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb.jsonl|0.45,json:/workspace/agillm44_ultradata_20260911/ultra_fineweb_l3.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_code.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_math.jsonl|0.35,json:/workspace/agillm44_ultradata_20260911/ultradata_rl.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultradata_sft_agent.jsonl|0.2,json:/workspace/agillm44_ultradata_20260911/ultrax.jsonl|0.35",
705
+ "step": 2865113,
706
+ "training_pointer_schema": "agillm43.training.resume.v2",
707
+ "training_pointer_sha256": "2281c815934ab200c270d4ab4533d3887c6cf2d1ed8b4c3383bdb53154bd0ba7",
708
+ "warmstart_kind": "warmstarted_from_lane_checkpoint",
709
+ "warmstart_source_path": "/workspace/agillm43_production_v14/production_quality_v18_hard_spike_gates/checkpoints_continue400B_from200B/pretrain_step02852036_from02849585_20260911T120900_1789128540505672216_1648411_000002.pt"
710
+ }