Hugging Face's logo

c4tdr0ut
/
Helios-7B

Text Generation
Transformers
Safetensors
Grok
English
mistral
text-generation-inference
axolotl
synthetic-data
full-parameter-finetuning
illegal
millitary
conversational
Model card Files
xet
 Community
Helios-7B
File size: 64,453 Bytes
8186f25
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
[2026-08-11 19:02:12,745] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:31] baseline 0.000GB ()
[2026-08-11 19:02:12,747] [INFO] [axolotl.cli.config.load_cfg:336] [PID:31] config:
{
  "activation_offloading": false,
  "attn_decontaminates_packing": true,
  "attn_implementation": "flash_attention_2",
  "attn_needs_dtype_cast": true,
  "attn_supports_packing": true,
  "attn_uses_flash_lib": true,
  "auto_resume_from_checkpoints": false,
  "axolotl_config_path": "/workspace/task_08e727bf-a87d-47c2-97d0-de595c57c5dc_output/sft_run/sft_training_config.yaml",
  "base_model": "c4tdr0ut/heli-pretrain",
  "base_model_config": "c4tdr0ut/heli-pretrain",
  "batch_size": 150,
  "bf16": true,
  "capabilities": {
    "bf16": true,
    "compute_capability": "sm_90",
    "fp8": true,
    "n_gpu": 1,
    "n_node": 1,
    "tf32": true
  },
  "chat_template": "chatml",
  "context_parallel_size": 1,
  "dataloader_num_workers": 1,
  "dataloader_pin_memory": true,
  "dataloader_prefetch_factor": 256,
  "dataset_num_proc": 17,
  "dataset_prepared_path": "last_finetune_prepared",
  "datasets": [
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "axolotl_rag_conversations_inputs.jsonl",
      "trust_remote_code": false,
      "type": "input_output"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "axolotl_correction_conversations_inputs.json",
      "trust_remote_code": false,
      "type": "input_output"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "pretraining_subset_1515812.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "factual_sft_completion/combined_all_0.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "factual_sft_completion/combined_all_2.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "factual_sft_completion/combined_all_1.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Augmentoolkit-Pippa-Thoughts_748413.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Augmentoolkit-Bluemoon-1mil-thoughts_748413.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Openthoughts-100mil-DifferentFormat_2993654.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Augmentoolkit-LMsys-800k-Thoughts_748413.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Augmentoolkit-Generic-Grabbag-Thoughts_1496827.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    },
    {
      "message_property_mappings": {
        "content": "content",
        "role": "role"
      },
      "path": "generic_sft_completion/Augmentoolkit-Augmentoolkit-Capybara-2point5mil-Thoughts_748413.jsonl",
      "trust_remote_code": false,
      "type": "completion"
    }
  ],
  "ddp": false,
  "device": "cuda:0",
  "device_map": "auto",
  "dion_rank_fraction": 1.0,
  "dion_rank_multiple_of": 1,
  "eaft_alpha": 1.0,
  "eaft_k": 20,
  "env_capabilities": {
    "torch_version": "2.12.0"
  },
  "eval_batch_size": 4,
  "eval_causal_lm_metrics": [
    "sacrebleu",
    "comet",
    "ter",
    "chrf"
  ],
  "eval_max_new_tokens": 128,
  "eval_sample_packing": false,
  "eval_steps": 0.25,
  "eval_table_size": 0,
  "evals_per_epoch": 1,
  "experimental_skip_move_to_device": true,
  "fp16": false,
  "generate_samples": false,
  "generation_do_sample": true,
  "generation_max_new_tokens": 50,
  "generation_prompt_ratio": 0.5,
  "generation_temperature": 0.7,
  "gradient_accumulation_steps": 10,
  "gradient_checkpointing": true,
  "gradient_checkpointing_kwargs": {
    "use_reentrant": true
  },
  "group_by_length": false,
  "hub_model_id": "c4tdr0ut/Helios-7B",
  "hub_strategy": "all_checkpoints",
  "include_tkps": true,
  "is_falcon_derived_model": false,
  "is_llama_derived_model": false,
  "is_mistral_derived_model": true,
  "layer_offloading": false,
  "learning_rate": 2e-05,
  "liger_fused_linear_cross_entropy": true,
  "liger_glu_activation": true,
  "liger_layer_norm": true,
  "liger_rms_norm": true,
  "liger_rope": true,
  "lisa_layers_attribute": "model.layers",
  "load_best_model_at_end": false,
  "load_in_4bit": false,
  "load_in_8bit": false,
  "local_rank": 0,
  "logging_steps": 1,
  "lora_dropout": 0.0,
  "loraplus_lr_embedding": 1e-06,
  "lr_scheduler": "constant",
  "mean_resizing_embeddings": false,
  "merge_method": "memory_efficient",
  "micro_batch_size": 15,
  "model_config_type": "mistral",
  "neftune_noise_alpha": 5.0,
  "num_epochs": 4.0,
  "num_generation_samples": 3,
  "optimizer": "paged_adamw_8bit",
  "otel_metrics_host": "localhost",
  "otel_metrics_port": 8000,
  "output_dir": "./finetune-model-output",
  "pad_to_sequence_len": false,
  "plugins": [
    "axolotl.integrations.liger.LigerPlugin"
  ],
  "pretrain_multipack_attn": true,
  "profiler_steps_start": 0,
  "qgalore_cos_threshold": 0.4,
  "qgalore_gamma_proj": 2,
  "qgalore_proj_bits": 4,
  "qgalore_proj_group_size": 256,
  "qgalore_proj_quant": true,
  "qgalore_proj_type": "std",
  "qgalore_queue_size": 5,
  "qgalore_rank": 256,
  "qgalore_scale": 0.25,
  "qgalore_update_proj_gap": 200,
  "qlora_sharded_model_loading": false,
  "quantize_moe_experts": false,
  "ray_num_workers": 1,
  "relora_prune_method": "magnitude",
  "remove_unused_columns": false,
  "resources_per_worker": {
    "GPU": 1
  },
  "sample_packing": true,
  "sample_packing_bin_size": 200,
  "sample_packing_group_size": 100000,
  "save_only_model": false,
  "save_safetensors": true,
  "save_steps": 0.25,
  "save_total_limit": 2,
  "saves_per_epoch": 1,
  "seed": 1337,
  "sequence_len": 5000,
  "shuffle_before_merging_datasets": false,
  "shuffle_merged_datasets": true,
  "skip_prepare_dataset": false,
  "special_tokens": {
    "pad_token": "<unk>"
  },
  "streaming_multipack_buffer_size": 10000,
  "strict": false,
  "tensor_parallel_size": 1,
  "tf32": false,
  "tiled_mlp_use_original_mlp": true,
  "tokenizer_config": "c4tdr0ut/heli-pretrain",
  "tokenizer_save_jinja_files": true,
  "tokenizer_type": "AutoTokenizer",
  "torch_dtype": "torch.bfloat16",
  "train_on_inputs": false,
  "trl": {
    "async_prefetch": false,
    "log_completions": false,
    "mask_truncated_completions": false,
    "ref_model_mixup_alpha": 0.9,
    "ref_model_sync_steps": 64,
    "replay_buffer_size": 0,
    "replay_recompute_logps": true,
    "reroll_max_groups": 1,
    "reroll_start_fraction": 1.0,
    "reward_num_workers": 1,
    "scale_rewards": true,
    "skip_zero_advantage_batches": true,
    "sync_ref_model": false,
    "use_data_producer": false,
    "use_vllm": false,
    "vllm_lora_sync": false,
    "vllm_server_host": "0.0.0.0",
    "vllm_server_port": 8000
  },
  "type_of_model": "AutoModelForCausalLM",
  "use_otel_metrics": false,
  "use_ray": false,
  "val_set_size": 0.04,
  "vllm": {
    "device": "auto",
    "dtype": "auto",
    "gpu_memory_utilization": 0.9,
    "host": "0.0.0.0",
    "port": 8000
  },
  "warmup_ratio": 0.1,
  "weight_decay": 0.0,
  "world_size": 1
}
[2026-08-11 19:02:17,597] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:31] EOS: 2 / </s>
[2026-08-11 19:02:17,597] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:31] BOS: 1 / <s>
[2026-08-11 19:02:17,598] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:31] PAD: 0 / <unk>
[2026-08-11 19:02:17,598] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:31] UNK: 0 / <unk>
[2026-08-11 19:02:17,601] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:543] [PID:31] Unable to find prepared dataset in last_finetune_prepared/5eab9bc891e347cf10de8154a14a218d
[2026-08-11 19:02:17,601] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:31] Loading raw datasets...
[2026-08-11 19:02:17,601] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:31] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`.

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 500 examples [00:00, 9589.65 examples/s]
[2026-08-11 19:02:18,066] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: axolotl_rag_conversations_inputs.jsonl with base_type: input_output and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/500 [00:00<?, ? examples/s][2026-08-11 19:02:18,496] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/multiprocess/popen_fork.py:66: DeprecationWarning: This process (pid=31) is multi-threaded, use of fork() may lead to deadlocks in the child.
  self.pid = os.fork()


Tokenizing Prompts (num_proc=17):   0%|          | 2/500 [00:01<07:47,  1.06 examples/s]
Tokenizing Prompts (num_proc=17):   4%|▎         | 18/500 [00:02<00:40, 11.79 examples/s]
Tokenizing Prompts (num_proc=17):   7%|▋         | 37/500 [00:02<00:16, 27.42 examples/s]
Tokenizing Prompts (num_proc=17):  14%|█▍        | 72/500 [00:02<00:06, 61.15 examples/s]
Tokenizing Prompts (num_proc=17):  23%|██▎       | 113/500 [00:02<00:03, 107.20 examples/s]
Tokenizing Prompts (num_proc=17):  28%|██▊       | 142/500 [00:02<00:02, 133.58 examples/s]
Tokenizing Prompts (num_proc=17):  40%|████      | 200/500 [00:02<00:01, 213.81 examples/s]
Tokenizing Prompts (num_proc=17):  48%|████▊     | 238/500 [00:02<00:01, 229.44 examples/s]
Tokenizing Prompts (num_proc=17):  58%|█████▊    | 288/500 [00:02<00:00, 284.63 examples/s]
Tokenizing Prompts (num_proc=17):  65%|██████▌   | 327/500 [00:02<00:00, 294.54 examples/s]
Tokenizing Prompts (num_proc=17):  73%|███████▎  | 365/500 [00:03<00:00, 291.66 examples/s]
Tokenizing Prompts (num_proc=17):  82%|████████▏ | 409/500 [00:03<00:00, 325.36 examples/s]
Tokenizing Prompts (num_proc=17):  90%|█████████ | 451/500 [00:03<00:00, 338.40 examples/s]
Tokenizing Prompts (num_proc=17):  98%|█████████▊| 490/500 [00:03<00:00, 289.79 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 500/500 [00:03<00:00, 129.54 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 388 examples [00:00, 2954.03 examples/s]
Generating train split: 388 examples [00:00, 2891.60 examples/s]
[2026-08-11 19:02:22,531] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: axolotl_correction_conversations_inputs.json with base_type: input_output and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/388 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   4%|▍         | 16/388 [00:01<00:43,  8.52 examples/s]
Tokenizing Prompts (num_proc=17):  14%|█▍        | 54/388 [00:02<00:10, 33.33 examples/s]
Tokenizing Prompts (num_proc=17):  27%|██▋       | 104/388 [00:02<00:03, 71.73 examples/s]
Tokenizing Prompts (num_proc=17):  36%|███▌      | 138/388 [00:02<00:02, 88.56 examples/s]
Tokenizing Prompts (num_proc=17):  51%|█████▏    | 199/388 [00:02<00:01, 147.71 examples/s]
Tokenizing Prompts (num_proc=17):  65%|██████▌   | 253/388 [00:02<00:00, 196.74 examples/s]
Tokenizing Prompts (num_proc=17):  75%|███████▌  | 292/388 [00:02<00:00, 208.67 examples/s]
Tokenizing Prompts (num_proc=17):  87%|████████▋ | 338/388 [00:02<00:00, 251.52 examples/s]
Tokenizing Prompts (num_proc=17):  99%|█████████▊| 383/388 [00:03<00:00, 287.97 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 388/388 [00:03<00:00, 120.92 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 11 examples [00:00, 602.06 examples/s]
[2026-08-11 19:02:26,503] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: pretraining_subset_1515812.jsonl with base_type: completion and prompt_style: None
[2026-08-11 19:02:26,507] [WARNING] [datasets.arrow_dataset.map:3413] [PID:31] num_proc must be <= 11. Reducing num_proc to 11 for dataset of size 11.

Tokenizing Prompts (num_proc=11):   0%|          | 0/11 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=11):   9%|▉         | 1/11 [00:01<00:15,  1.60s/ examples]
Tokenizing Prompts (num_proc=11):  27%|██▋       | 3/11 [00:01<00:03,  2.22 examples/s]
Tokenizing Prompts (num_proc=11):  45%|████▌     | 5/11 [00:01<00:01,  3.83 examples/s]
Tokenizing Prompts (num_proc=11):  64%|██████▎   | 7/11 [00:02<00:00,  5.18 examples/s]
Tokenizing Prompts (num_proc=11):  91%|█████████ | 10/11 [00:02<00:00,  7.41 examples/s]
Tokenizing Prompts (num_proc=11): 100%|██████████| 11/11 [00:02<00:00,  3.69 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 1057 examples [00:00, 34419.58 examples/s]
[2026-08-11 19:02:30,018] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: factual_sft_completion/combined_all_0.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/1057 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 63/1057 [00:01<00:25, 38.67 examples/s]
Tokenizing Prompts (num_proc=17):  12%|█▏        | 126/1057 [00:01<00:10, 85.83 examples/s]
Tokenizing Prompts (num_proc=17):  30%|██▉       | 313/1057 [00:01<00:02, 266.76 examples/s]
Tokenizing Prompts (num_proc=17):  41%|████▏     | 437/1057 [00:01<00:01, 386.78 examples/s]
Tokenizing Prompts (num_proc=17):  59%|█████▉    | 623/1057 [00:02<00:00, 551.39 examples/s]
Tokenizing Prompts (num_proc=17):  71%|███████   | 747/1057 [00:02<00:00, 590.90 examples/s]
Tokenizing Prompts (num_proc=17):  94%|█████████▍| 995/1057 [00:02<00:00, 793.97 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 1057/1057 [00:02<00:00, 402.54 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 1041 examples [00:00, 32791.87 examples/s]
[2026-08-11 19:02:33,121] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: factual_sft_completion/combined_all_2.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/1041 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 62/1041 [00:01<00:27, 36.23 examples/s]
Tokenizing Prompts (num_proc=17):  18%|█▊        | 186/1041 [00:01<00:06, 125.98 examples/s]
Tokenizing Prompts (num_proc=17):  36%|███▌      | 370/1041 [00:02<00:02, 275.52 examples/s]
Tokenizing Prompts (num_proc=17):  53%|█████▎    | 553/1041 [00:02<00:01, 432.87 examples/s]
Tokenizing Prompts (num_proc=17):  71%|███████   | 736/1041 [00:02<00:00, 562.13 examples/s]
Tokenizing Prompts (num_proc=17):  88%|████████▊ | 919/1041 [00:02<00:00, 706.13 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 1041/1041 [00:02<00:00, 393.82 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 1041 examples [00:00, 34785.73 examples/s]
[2026-08-11 19:02:36,287] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: factual_sft_completion/combined_all_1.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/1041 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 62/1041 [00:01<00:25, 38.17 examples/s]
Tokenizing Prompts (num_proc=17):  24%|██▍       | 248/1041 [00:01<00:04, 173.56 examples/s]
Tokenizing Prompts (num_proc=17):  36%|███▌      | 370/1041 [00:01<00:02, 269.79 examples/s]
Tokenizing Prompts (num_proc=17):  53%|█████▎    | 553/1041 [00:02<00:01, 427.79 examples/s]
Tokenizing Prompts (num_proc=17):  71%|███████   | 736/1041 [00:02<00:00, 556.70 examples/s]
Tokenizing Prompts (num_proc=17):  94%|█████████▍| 980/1041 [00:02<00:00, 796.34 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 1041/1041 [00:02<00:00, 402.39 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 40 examples [00:00, 4102.51 examples/s]
[2026-08-11 19:02:39,388] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Augmentoolkit-Pippa-Thoughts_748413.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/40 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   8%|▊         | 3/40 [00:01<00:16,  2.22 examples/s]
Tokenizing Prompts (num_proc=17):  22%|██▎       | 9/40 [00:01<00:04,  6.99 examples/s]
Tokenizing Prompts (num_proc=17):  35%|███▌      | 14/40 [00:01<00:02,  9.88 examples/s]
Tokenizing Prompts (num_proc=17):  52%|█████▎    | 21/40 [00:01<00:01, 16.52 examples/s]
Tokenizing Prompts (num_proc=17):  62%|██████▎   | 25/40 [00:02<00:00, 19.74 examples/s]
Tokenizing Prompts (num_proc=17):  85%|████████▌ | 34/40 [00:02<00:00, 30.94 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 40/40 [00:02<00:00, 16.80 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 45 examples [00:00, 3857.74 examples/s]
[2026-08-11 19:02:42,226] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Augmentoolkit-Bluemoon-1mil-thoughts_748413.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/45 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   7%|▋         | 3/45 [00:01<00:21,  1.91 examples/s]
Tokenizing Prompts (num_proc=17):  13%|█▎        | 6/45 [00:01<00:09,  4.22 examples/s]
Tokenizing Prompts (num_proc=17):  40%|████      | 18/45 [00:01<00:01, 15.84 examples/s]
Tokenizing Prompts (num_proc=17):  60%|██████    | 27/45 [00:01<00:00, 23.37 examples/s]
Tokenizing Prompts (num_proc=17):  78%|███████▊  | 35/45 [00:02<00:00, 27.52 examples/s]
Tokenizing Prompts (num_proc=17):  96%|█████████▌| 43/45 [00:02<00:00, 28.48 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 45/45 [00:02<00:00, 17.38 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 389 examples [00:00, 15786.22 examples/s]
[2026-08-11 19:02:45,300] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Openthoughts-100mil-DifferentFormat_2993654.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/389 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 23/389 [00:01<00:27, 13.46 examples/s]
Tokenizing Prompts (num_proc=17):  18%|█▊        | 69/389 [00:01<00:06, 46.97 examples/s]
Tokenizing Prompts (num_proc=17):  24%|██▎       | 92/389 [00:01<00:04, 64.34 examples/s]
Tokenizing Prompts (num_proc=17):  41%|████▏     | 161/389 [00:02<00:01, 133.17 examples/s]
Tokenizing Prompts (num_proc=17):  53%|█████▎    | 207/389 [00:02<00:01, 179.30 examples/s]
Tokenizing Prompts (num_proc=17):  65%|██████▌   | 253/389 [00:02<00:00, 218.99 examples/s]
Tokenizing Prompts (num_proc=17):  83%|████████▎ | 322/389 [00:02<00:00, 294.81 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 389/389 [00:02<00:00, 146.34 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 813 examples [00:00, 70109.16 examples/s]
[2026-08-11 19:02:48,430] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Augmentoolkit-LMsys-800k-Thoughts_748413.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/813 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 48/813 [00:01<00:24, 31.02 examples/s]
Tokenizing Prompts (num_proc=17):  18%|█▊        | 144/813 [00:01<00:06, 108.15 examples/s]
Tokenizing Prompts (num_proc=17):  30%|██▉       | 240/813 [00:01<00:02, 196.60 examples/s]
Tokenizing Prompts (num_proc=17):  41%|████▏     | 336/813 [00:01<00:01, 289.09 examples/s]
Tokenizing Prompts (num_proc=17):  53%|█████▎    | 432/813 [00:02<00:01, 360.93 examples/s]
Tokenizing Prompts (num_proc=17):  71%|███████   | 576/813 [00:02<00:00, 521.34 examples/s]
Tokenizing Prompts (num_proc=17):  83%|████████▎ | 672/813 [00:02<00:00, 603.26 examples/s]
Tokenizing Prompts (num_proc=17):  94%|█████████▍| 766/813 [00:02<00:00, 660.00 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 813/813 [00:02<00:00, 323.82 examples/s]

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 2531 examples [00:00, 97320.19 examples/s]
[2026-08-11 19:02:51,457] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Augmentoolkit-Generic-Grabbag-Thoughts_1496827.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/2531 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 149/2531 [00:02<00:33, 70.55 examples/s]
Tokenizing Prompts (num_proc=17):  41%|████      | 1043/2531 [00:02<00:02, 509.55 examples/s]
Tokenizing Prompts (num_proc=17):  53%|█████▎    | 1341/2531 [00:02<00:01, 670.58 examples/s]
Tokenizing Prompts (num_proc=17):  65%|██████▍   | 1639/2531 [00:02<00:01, 847.87 examples/s]
Tokenizing Prompts (num_proc=17):  82%|████████▏ | 2086/2531 [00:02<00:00, 1192.76 examples/s]
Tokenizing Prompts (num_proc=17):  94%|█████████▍| 2383/2531 [00:03<00:00, 1414.16 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 2531/2531 [00:03<00:00, 787.24 examples/s] 

Generating train split: 0 examples [00:00, ? examples/s]
Generating train split: 444 examples [00:00, 41194.31 examples/s]
[2026-08-11 19:02:55,122] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:31] Loading dataset: generic_sft_completion/Augmentoolkit-Augmentoolkit-Capybara-2point5mil-Thoughts_748413.jsonl with base_type: completion and prompt_style: None

Tokenizing Prompts (num_proc=17):   0%|          | 0/444 [00:00<?, ? examples/s]
Tokenizing Prompts (num_proc=17):   6%|▌         | 27/444 [00:01<00:24, 17.18 examples/s]
Tokenizing Prompts (num_proc=17):  18%|█▊        | 80/444 [00:01<00:06, 58.21 examples/s]
Tokenizing Prompts (num_proc=17):  36%|███▌      | 158/444 [00:01<00:02, 124.21 examples/s]
Tokenizing Prompts (num_proc=17):  47%|████▋     | 210/444 [00:01<00:01, 167.76 examples/s]
Tokenizing Prompts (num_proc=17):  59%|█████▉    | 262/444 [00:02<00:00, 216.33 examples/s]
Tokenizing Prompts (num_proc=17):  71%|███████   | 314/444 [00:02<00:00, 264.54 examples/s]
Tokenizing Prompts (num_proc=17):  82%|████████▏ | 366/444 [00:02<00:00, 313.18 examples/s]
Tokenizing Prompts (num_proc=17):  94%|█████████▍| 418/444 [00:02<00:00, 346.24 examples/s]
Tokenizing Prompts (num_proc=17): 100%|██████████| 444/444 [00:02<00:00, 170.98 examples/s]
[2026-08-11 19:02:57,993] [INFO] [axolotl.utils.data.shared.merge_datasets:712] [PID:31] Merging datasets...
[2026-08-11 19:02:58,037] [DEBUG] [axolotl.utils.data.shared.merge_datasets:716] [PID:31] Shuffling merged datasets...
[2026-08-11 19:02:58,273] [INFO] [axolotl.utils.data.utils._log_dataset_stats:212] [PID:31] min_input_len: 8
[2026-08-11 19:02:58,274] [INFO] [axolotl.utils.data.utils._log_dataset_stats:213] [PID:31] max_input_len: 15017

Dropping Invalid Sequences (<None or >5000) (num_proc=17):   0%|          | 0/9474 [00:00<?, ? examples/s]
Dropping Invalid Sequences (<None or >5000) (num_proc=17):   6%|▌         | 558/9474 [00:01<00:30, 294.29 examples/s]
Dropping Invalid Sequences (<None or >5000) (num_proc=17):  47%|████▋     | 4461/9474 [00:01<00:01, 3002.66 examples/s]
Dropping Invalid Sequences (<None or >5000) (num_proc=17): 100%|██████████| 9474/9474 [00:02<00:00, 4284.64 examples/s]
[2026-08-11 19:03:00,622] [INFO] [axolotl.utils.data.utils._drop_outside_range:306] [PID:31] Dropped 443 sequences outside valid range ([None, 5000])
[2026-08-11 19:03:00,626] [INFO] [axolotl.utils.trainer.process_datasets_for_packing:254] [PID:31] dropping token_type_ids column if it exists

Drop Samples with Zero Trainable Tokens (num_proc=17):   0%|          | 0/9031 [00:00<?, ? examples/s]
Drop Samples with Zero Trainable Tokens (num_proc=17):   6%|▌         | 532/9031 [00:01<00:27, 305.55 examples/s]
Drop Samples with Zero Trainable Tokens (num_proc=17):  24%|██▎       | 2127/9031 [00:01<00:04, 1471.87 examples/s]
Drop Samples with Zero Trainable Tokens (num_proc=17):  88%|████████▊ | 7969/9031 [00:01<00:00, 6898.04 examples/s]
Drop Samples with Zero Trainable Tokens (num_proc=17): 100%|██████████| 9031/9031 [00:02<00:00, 4297.46 examples/s]

Add position_id column (Sample Packing) (num_proc=17):   0%|          | 0/9031 [00:00<?, ? examples/s][2026-08-11 19:03:02,994] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/multiprocess/popen_fork.py:66: DeprecationWarning: This process (pid=31) is multi-threaded, use of fork() may lead to deadlocks in the child.
  self.pid = os.fork()


Add position_id column (Sample Packing) (num_proc=17):   6%|▌         | 532/9031 [00:02<00:34, 249.39 examples/s]
Add position_id column (Sample Packing) (num_proc=17):  35%|███▌      | 3190/9031 [00:02<00:03, 1898.45 examples/s]
Add position_id column (Sample Packing) (num_proc=17):  94%|█████████▍| 8500/9031 [00:02<00:00, 6099.61 examples/s]
Add position_id column (Sample Packing) (num_proc=17): 100%|██████████| 9031/9031 [00:02<00:00, 3623.92 examples/s]

Saving the dataset (0/17 shards):   0%|          | 0/9031 [00:00<?, ? examples/s]
Saving the dataset (0/17 shards):   6%|▌         | 531/9031 [00:05<01:28, 96.46 examples/s]
Saving the dataset (1/17 shards):   6%|▌         | 531/9031 [00:05<01:28, 96.46 examples/s]
Saving the dataset (1/17 shards):  12%|█▏        | 1062/9031 [00:05<00:35, 222.35 examples/s]
Saving the dataset (2/17 shards):  12%|█▏        | 1062/9031 [00:05<00:35, 222.35 examples/s]
Saving the dataset (3/17 shards):  18%|█▊        | 1594/9031 [00:05<00:33, 222.35 examples/s]
Saving the dataset (4/17 shards):  24%|██▎       | 2126/9031 [00:05<00:31, 222.35 examples/s]
Saving the dataset (5/17 shards):  29%|██▉       | 2657/9031 [00:05<00:28, 222.35 examples/s]
Saving the dataset (6/17 shards):  35%|███▌      | 3189/9031 [00:05<00:26, 222.35 examples/s]
Saving the dataset (7/17 shards):  47%|████▋     | 4251/9031 [00:05<00:21, 222.35 examples/s]
Saving the dataset (8/17 shards):  47%|████▋     | 4251/9031 [00:05<00:21, 222.35 examples/s]
Saving the dataset (9/17 shards):  53%|█████▎    | 4782/9031 [00:05<00:19, 222.35 examples/s]
Saving the dataset (10/17 shards):  59%|█████▉    | 5313/9031 [00:05<00:16, 222.35 examples/s]
Saving the dataset (11/17 shards):  65%|██████▍   | 5844/9031 [00:05<00:14, 222.35 examples/s]
Saving the dataset (12/17 shards):  71%|███████   | 6375/9031 [00:05<00:11, 222.35 examples/s]
Saving the dataset (13/17 shards):  76%|███████▋  | 6906/9031 [00:05<00:09, 222.35 examples/s]
Saving the dataset (14/17 shards):  82%|████████▏ | 7437/9031 [00:05<00:07, 222.35 examples/s]
Saving the dataset (15/17 shards):  88%|████████▊ | 7968/9031 [00:05<00:04, 222.35 examples/s]
Saving the dataset (15/17 shards):  94%|█████████▍| 8500/9031 [00:05<00:00, 2684.59 examples/s]
Saving the dataset (16/17 shards):  94%|█████████▍| 8500/9031 [00:05<00:00, 2684.59 examples/s]
Saving the dataset (17/17 shards): 100%|██████████| 9031/9031 [00:05<00:00, 2684.59 examples/s]
Saving the dataset (17/17 shards): 100%|██████████| 9031/9031 [00:06<00:00, 1500.92 examples/s]
[2026-08-11 19:03:11,569] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:418] [PID:31] total_num_tokens: 17_945_275
[2026-08-11 19:03:11,619] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:440] [PID:31] total_supervised_tokens: 17_669_047
[2026-08-11 19:03:11,772] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_fork.py:66: DeprecationWarning: This process (pid=31) is multi-threaded, use of fork() may lead to deadlocks in the child.
  self.pid = os.fork()

[2026-08-11 19:03:14,907] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.3644943237304688
[2026-08-11 19:03:16,260] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.352339267730713
[2026-08-11 19:03:17,635] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.3746399879455566
[2026-08-11 19:03:19,068] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.432831048965454
[2026-08-11 19:03:19,109] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:443] [PID:31] gather_len_batches: [240]
[2026-08-11 19:03:19,110] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:497] [PID:31] data_loader_len: 24
[2026-08-11 19:03:19,110] [INFO] [axolotl.utils.trainer.calc_sample_packing_eff_est:506] [PID:31] sample_packing_eff_est across ranks: [0.9969597222222222]
[2026-08-11 19:03:19,110] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:518] [PID:31] sample_packing_eff_est: 1.0
[2026-08-11 19:03:19,110] [DEBUG] [axolotl.utils.trainer.calculate_total_num_steps:523] [PID:31] total_num_steps: 96
[2026-08-11 19:03:19,111] [INFO] [axolotl.utils.data.sft._prepare_standard_dataset:121] [PID:31] Maximum number of steps set at 96
[2026-08-11 19:03:19,121] [DEBUG] [axolotl.train.setup_model_and_tokenizer:70] [PID:31] loading tokenizer... c4tdr0ut/heli-pretrain
[2026-08-11 19:03:20,183] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:31] EOS: 2 / </s>
[2026-08-11 19:03:20,183] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:31] BOS: 1 / <s>
[2026-08-11 19:03:20,184] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:31] PAD: 0 / <unk>
[2026-08-11 19:03:20,184] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:31] UNK: 0 / <unk>
[2026-08-11 19:03:20,184] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:31] Loading model
[2026-08-11 19:03:20,301] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:31] Patched OptimState8bit for torch.compile compatibility
[2026-08-11 19:03:20,301] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:31] Patched OptimState4bit for torch.compile compatibility
[2026-08-11 19:03:20,302] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:31] Patched OptimStateFp8 for torch.compile compatibility
[2026-08-11 19:03:20,310] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:31] Patched Trainer.evaluation_loop with nanmean loss calculation
[2026-08-11 19:03:20,311] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:31] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation
[2026-08-11 19:03:20,312] [INFO] [axolotl.loaders.patch_manager._apply_multipack_patches:912] [PID:31] Applying multipack dataloader patch for sample packing...
[2026-08-11 19:03:20,322] [INFO] [axolotl.monkeypatch.attention.flash_attn_4.fa4_usable:140] [PID:31] Flash Attention 4 is available for your GPU and offers faster training. To enable: pip install --pre flash-attn-4
[2026-08-11 19:03:24,678] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/torch/jit/_script.py:1488: DeprecationWarning: `torch.jit.script` is deprecated. Please switch to `torch.compile` or `torch.export`.
  warnings.warn(

[2026-08-11 19:03:24,784] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/fla/ops/nsa/parallel.py:20: ImportWarning: Flash Attention is not installed. Please install it via `pip install flash-attn --no-build-isolation`
  warnings.warn(

[2026-08-11 19:03:25,005] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:145] [PID:31] Applying LIGER to mistral with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'swiglu': True}
[2026-08-11 19:04:14,342] [WARNING] [kernels._versions.resolve_version_spec_as_ref:83] [PID:31] You are using version 1 of 'kernels-community/flash-attn2', but version 3 is available.

Fetching ... files: 0it [00:00, ?it/s]
Fetching ... files: 1it [00:00,  8.89it/s]
Fetching ... files: 2it [00:05,  3.05s/it]
Fetching ... files: 23it [00:05,  4.40it/s]

Loading weights:   0%|          | 0/291 [00:00<?, ?it/s]
Loading weights:   1%|          | 2/291 [00:00<00:17, 16.34it/s]
Loading weights:   5%|▌         | 15/291 [00:00<00:03, 70.76it/s]
Loading weights:  11%|█         | 31/291 [00:00<00:02, 100.08it/s]
Loading weights:  14%|█▍        | 42/291 [00:00<00:02, 98.29it/s] 
Loading weights:  20%|█▉        | 58/291 [00:00<00:02, 115.07it/s]
Loading weights:  24%|██▍       | 70/291 [00:00<00:01, 111.24it/s]
Loading weights:  29%|██▉       | 85/291 [00:00<00:01, 116.40it/s]
Loading weights:  33%|███▎      | 97/291 [00:00<00:01, 108.51it/s]
Loading weights:  38%|███▊      | 112/291 [00:01<00:01, 116.78it/s]
Loading weights:  43%|████▎     | 124/291 [00:01<00:01, 112.18it/s]
Loading weights:  48%|████▊     | 139/291 [00:01<00:01, 117.28it/s]
Loading weights:  52%|█████▏    | 151/291 [00:01<00:01, 111.38it/s]
Loading weights:  57%|█████▋    | 166/291 [00:01<00:01, 117.88it/s]
Loading weights:  61%|██████    | 178/291 [00:01<00:01, 111.62it/s]
Loading weights:  65%|██████▌   | 190/291 [00:01<00:00, 112.22it/s]
Loading weights:  69%|██████▉   | 202/291 [00:01<00:00, 111.04it/s]
Loading weights:  74%|███████▎  | 214/291 [00:01<00:00, 107.60it/s]
Loading weights:  79%|███████▊  | 229/291 [00:02<00:00, 115.35it/s]
Loading weights:  83%|████████▎ | 241/291 [00:02<00:00, 110.60it/s]
Loading weights:  88%|████████▊ | 256/291 [00:02<00:00, 116.51it/s]
Loading weights:  92%|█████████▏| 268/291 [00:02<00:00, 111.35it/s]
Loading weights:  97%|█████████▋| 283/291 [00:02<00:00, 116.71it/s]
Loading weights: 100%|██████████| 291/291 [00:02<00:00, 110.96it/s]
[2026-08-11 19:04:23,372] [INFO] [axolotl.loaders.model._configure_embedding_dtypes:482] [PID:31] Converting modules to torch.bfloat16
[2026-08-11 19:04:23,375] [DEBUG] [axolotl.loaders.model.log_gpu_memory_usage:127] [PID:31] Memory usage after model load 14.251GB (+14.251GB allocated, +14.502GB reserved)
[2026-08-11 19:04:24,900] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/torch/jit/_script.py:1488: DeprecationWarning: `torch.jit.script` is deprecated. Please switch to `torch.compile` or `torch.export`.
  warnings.warn(

[2026-08-11 19:04:25,089] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /workspace/axolotl-venv/lib/python3.12/site-packages/torch/jit/_script.py:1640: DeprecationWarning: `torch.jit.interface` is deprecated. Please use `torch.compile` instead.
  warnings.warn(

[2026-08-11 19:04:25,356] [WARNING] [accelerate.utils.other.check_os_kernel:520] [PID:31] [RANK 0] Detected kernel version 4.19.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
[2026-08-11 19:04:27,407] [INFO] [axolotl.train.save_initial_configs:486] [PID:31] Pre-saving tokenizer to ./finetune-model-output...
[2026-08-11 19:04:27,448] [INFO] [axolotl.train.save_initial_configs:491] [PID:31] Pre-saving model config to ./finetune-model-output...
[2026-08-11 19:04:27,453] [INFO] [axolotl.train.execute_training:234] [PID:31] Starting trainer...
[2026-08-11 19:04:28,034] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_fork.py:66: DeprecationWarning: This process (pid=31) is multi-threaded, use of fork() may lead to deadlocks in the child.
  self.pid = os.fork()

[2026-08-11 19:04:30,659] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.315068244934082
[2026-08-11 19:04:31,972] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.3127715587615967
[2026-08-11 19:04:33,341] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.368100643157959
[2026-08-11 19:04:34,639] [DEBUG] [axolotl.utils.samplers.multipack.__len__:467] [PID:31] generate_batches time: 1.297865867614746
[2026-08-11 19:04:34,639] [INFO] [axolotl.utils.samplers.multipack.calc_min_len:443] [PID:31] gather_len_batches: [240]

  0%|          | 0/96 [00:00<?, ?it/s][2026-08-11 19:04:34,682] [INFO] [axolotl.core.trainers.base.evaluate:464] [PID:31] Running evaluation step...


  0%|          | 0/90 [00:00<?, ?it/s]

  2%|▏         | 2/90 [00:00<00:28,  3.05it/s]

  4%|▍         | 4/90 [00:01<00:26,  3.30it/s]

  6%|▌         | 5/90 [00:01<00:29,  2.88it/s]

  7%|▋         | 6/90 [00:02<00:33,  2.51it/s]

  8%|▊         | 7/90 [00:02<00:28,  2.95it/s]

  9%|▉         | 8/90 [00:02<00:32,  2.54it/s]

 10%|█         | 9/90 [00:03<00:30,  2.65it/s]

 11%|█         | 10/90 [00:03<00:27,  2.90it/s]

 12%|█▏        | 11/90 [00:03<00:27,  2.83it/s]

 13%|█▎        | 12/90 [00:04<00:30,  2.53it/s]

 14%|█▍        | 13/90 [00:04<00:33,  2.32it/s]

 16%|█▌        | 14/90 [00:05<00:34,  2.20it/s]

 17%|█▋        | 15/90 [00:05<00:35,  2.11it/s]

 18%|█▊        | 16/90 [00:06<00:27,  2.64it/s]

 19%|█▉        | 17/90 [00:06<00:30,  2.39it/s]

 20%|██        | 18/90 [00:07<00:32,  2.22it/s]

 21%|██        | 19/90 [00:07<00:33,  2.14it/s]

 22%|██▏       | 20/90 [00:08<00:33,  2.06it/s]

 23%|██▎       | 21/90 [00:08<00:32,  2.14it/s]

 26%|██▌       | 23/90 [00:09<00:25,  2.64it/s]

 27%|██▋       | 24/90 [00:09<00:27,  2.40it/s]

 28%|██▊       | 25/90 [00:10<00:28,  2.26it/s]

 29%|██▉       | 26/90 [00:10<00:23,  2.78it/s]

 30%|███       | 27/90 [00:10<00:22,  2.80it/s]

 32%|███▏      | 29/90 [00:11<00:18,  3.30it/s]

 33%|███▎      | 30/90 [00:11<00:19,  3.02it/s]

 34%|███▍      | 31/90 [00:12<00:22,  2.59it/s]

 36%|███▌      | 32/90 [00:12<00:22,  2.54it/s]

 37%|███▋      | 33/90 [00:13<00:24,  2.35it/s]

 38%|███▊      | 34/90 [00:13<00:25,  2.20it/s]

 39%|███▉      | 35/90 [00:13<00:22,  2.40it/s]

 40%|████      | 36/90 [00:14<00:23,  2.34it/s]

 41%|████      | 37/90 [00:14<00:19,  2.65it/s]

 42%|████▏     | 38/90 [00:15<00:20,  2.56it/s]

 43%|████▎     | 39/90 [00:15<00:17,  2.93it/s]

 44%|████▍     | 40/90 [00:15<00:16,  3.02it/s]

 46%|████▌     | 41/90 [00:15<00:17,  2.83it/s]

 47%|████▋     | 42/90 [00:16<00:15,  3.19it/s]

 48%|████▊     | 43/90 [00:16<00:16,  2.85it/s]

 49%|████▉     | 44/90 [00:16<00:14,  3.13it/s]

 50%|█████     | 45/90 [00:17<00:14,  3.04it/s]

 51%|█████     | 46/90 [00:17<00:11,  3.72it/s]

 52%|█████▏    | 47/90 [00:17<00:11,  3.79it/s]

 53%|█████▎    | 48/90 [00:17<00:12,  3.43it/s]

 54%|█████▍    | 49/90 [00:18<00:12,  3.34it/s]

 56%|█████▌    | 50/90 [00:18<00:11,  3.57it/s]

 57%|█████▋    | 51/90 [00:18<00:13,  2.89it/s]

 58%|█████▊    | 52/90 [00:19<00:12,  3.01it/s]

 59%|█████▉    | 53/90 [00:19<00:14,  2.58it/s]

 60%|██████    | 54/90 [00:20<00:12,  2.88it/s]

 61%|██████    | 55/90 [00:20<00:13,  2.51it/s]

 62%|██████▏   | 56/90 [00:20<00:12,  2.62it/s]

 63%|██████▎   | 57/90 [00:21<00:10,  3.28it/s]

 64%|██████▍   | 58/90 [00:21<00:08,  3.80it/s]

 66%|██████▌   | 59/90 [00:21<00:10,  2.95it/s]

 67%|██████▋   | 60/90 [00:22<00:11,  2.54it/s]

 68%|██████▊   | 61/90 [00:22<00:11,  2.60it/s]

 69%|██████▉   | 62/90 [00:23<00:11,  2.45it/s]

 70%|███████   | 63/90 [00:23<00:09,  2.80it/s]

 71%|███████   | 64/90 [00:23<00:07,  3.39it/s]

 72%|███████▏  | 65/90 [00:23<00:09,  2.75it/s]

 73%|███████▎  | 66/90 [00:24<00:09,  2.47it/s]

 74%|███████▍  | 67/90 [00:24<00:08,  2.61it/s]

 76%|███████▌  | 68/90 [00:25<00:09,  2.37it/s]

 77%|███████▋  | 69/90 [00:25<00:08,  2.61it/s]

 78%|███████▊  | 70/90 [00:25<00:06,  2.94it/s]

 79%|███████▉  | 71/90 [00:26<00:06,  2.75it/s]

 80%|████████  | 72/90 [00:26<00:05,  3.12it/s]

 81%|████████  | 73/90 [00:26<00:05,  3.31it/s]

 82%|████████▏ | 74/90 [00:27<00:05,  2.71it/s]

 83%|████████▎ | 75/90 [00:27<00:06,  2.42it/s]

 84%|████████▍ | 76/90 [00:28<00:05,  2.69it/s]

 86%|████████▌ | 77/90 [00:28<00:04,  3.10it/s]

 87%|████████▋ | 78/90 [00:28<00:04,  2.84it/s]

 88%|████████▊ | 79/90 [00:29<00:04,  2.47it/s]

 89%|████████▉ | 80/90 [00:29<00:03,  2.73it/s]

 90%|█████████ | 81/90 [00:30<00:03,  2.37it/s]

 91%|█████████ | 82/90 [00:30<00:03,  2.27it/s]

 92%|█████████▏| 83/90 [00:31<00:03,  2.15it/s]

 93%|█████████▎| 84/90 [00:31<00:02,  2.07it/s]

 94%|█████████▍| 85/90 [00:32<00:02,  2.04it/s]

 96%|█████████▌| 86/90 [00:32<00:02,  1.99it/s]

 97%|█████████▋| 87/90 [00:32<00:01,  2.61it/s]

 98%|█████████▊| 88/90 [00:33<00:00,  2.35it/s]

 99%|█████████▉| 89/90 [00:33<00:00,  2.63it/s]

100%|██████████| 90/90 [00:33<00:00,  2.69it/s]
                                      

                                               
{'eval_loss': '1.762', 'eval_runtime': '36.99', 'eval_samples_per_second': '9.787', 'eval_steps_per_second': '2.46', 'eval_ppl': '5.824', 'memory/max_active (GiB)': '19.79', 'memory/max_allocated (GiB)': '19.79', 'memory/device_reserved (GiB)': '26.67', 'epoch': 0}

  0%|          | 0/96 [00:36<?, ?it/s]

100%|██████████| 90/90 [00:34<00:00,  2.69it/s]

                                               
  1%|          | 1/96 [02:01<3:12:52, 121.82s/it]
                                                 
{'loss': '1.668', 'grad_norm': '30.88', 'learning_rate': '2e-05', 'ppl': '5.3', 'memory/max_active (GiB)': '57.85', 'memory/max_allocated (GiB)': '57.85', 'memory/device_reserved (GiB)': '60.99', 'tokens/trainable': 742568, 'tokens/total': 749902, 'epoch': '0.04167'}

  1%|          | 1/96 [02:01<3:12:52, 121.82s/it]
  2%|▏         | 2/96 [03:20<2:31:24, 96.64s/it] 
                                                
{'loss': '1.606', 'grad_norm': '28.88', 'learning_rate': '2e-05', 'ppl': '4.985', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9394', 'tokens/trainable': 1484702, 'tokens/total': 1499819, 'epoch': '0.08333'}

  2%|▏         | 2/96 [03:20<2:31:24, 96.64s/it]
  3%|▎         | 3/96 [04:38<2:16:08, 87.83s/it]
                                                
{'loss': '1.242', 'grad_norm': '19.12', 'learning_rate': '2e-05', 'ppl': '3.461', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9349', 'tokens/trainable': 2207855, 'tokens/total': 2249708, 'epoch': '0.125'}

  3%|▎         | 3/96 [04:38<2:16:08, 87.83s/it]
  4%|▍         | 4/96 [05:56<2:08:53, 84.06s/it]
                                                
{'loss': '1.165', 'grad_norm': '33.5', 'learning_rate': '2e-05', 'ppl': '3.206', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9482', 'tokens/trainable': 2950021, 'tokens/total': 2999475, 'epoch': '0.1667'}

  4%|▍         | 4/96 [05:56<2:08:53, 84.06s/it]
  5%|▌         | 5/96 [07:14<2:03:55, 81.71s/it]
                                                
{'loss': '1.185', 'grad_norm': '17', 'learning_rate': '2e-05', 'ppl': '3.272', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9495', 'tokens/trainable': 3686352, 'tokens/total': 3749133, 'epoch': '0.2083'}

  5%|▌         | 5/96 [07:14<2:03:55, 81.71s/it]
  6%|▋         | 6/96 [08:31<2:00:32, 80.36s/it]
                                                
{'loss': '1.06', 'grad_norm': '8.812', 'learning_rate': '2e-05', 'ppl': '2.885', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9440', 'tokens/trainable': 4420112, 'tokens/total': 4498747, 'epoch': '0.25'}

  6%|▋         | 6/96 [08:31<2:00:32, 80.36s/it]
  7%|▋         | 7/96 [09:49<1:57:59, 79.55s/it]
                                                
{'loss': '0.9709', 'grad_norm': '6.844', 'learning_rate': '2e-05', 'ppl': '2.64', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '65.01', 'tokens/train_per_sec_per_gpu': '9429', 'tokens/trainable': 5154407, 'tokens/total': 5248232, 'epoch': '0.2917'}

  7%|▋         | 7/96 [09:49<1:57:59, 79.55s/it]
  8%|▊         | 8/96 [11:07<1:56:02, 79.12s/it]
                                                
{'loss': '0.9857', 'grad_norm': '4.281', 'learning_rate': '2e-05', 'ppl': '2.68', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9504', 'tokens/trainable': 5897628, 'tokens/total': 5997795, 'epoch': '0.3333'}

  8%|▊         | 8/96 [11:07<1:56:02, 79.12s/it]
  9%|▉         | 9/96 [12:25<1:54:02, 78.65s/it]
                                                
{'loss': '0.9362', 'grad_norm': '3.953', 'learning_rate': '2e-05', 'ppl': '2.55', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9489', 'tokens/trainable': 6634055, 'tokens/total': 6747342, 'epoch': '0.375'}

  9%|▉         | 9/96 [12:25<1:54:02, 78.65s/it]
 10%|█         | 10/96 [13:43<1:52:20, 78.37s/it]
                                                 
{'loss': '0.9151', 'grad_norm': '4.656', 'learning_rate': '2e-05', 'ppl': '2.497', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9548', 'tokens/trainable': 7376513, 'tokens/total': 7497031, 'epoch': '0.4167'}

 10%|█         | 10/96 [13:43<1:52:20, 78.37s/it]
 11%|█▏        | 11/96 [15:01<1:50:50, 78.24s/it]
                                                 
{'loss': '0.9457', 'grad_norm': '5.469', 'learning_rate': '2e-05', 'ppl': '2.575', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9519', 'tokens/trainable': 8118382, 'tokens/total': 8246701, 'epoch': '0.4583'}

 11%|█▏        | 11/96 [15:01<1:50:50, 78.24s/it]
 12%|█▎        | 12/96 [16:18<1:49:20, 78.10s/it]
                                                 
{'loss': '0.8496', 'grad_norm': '2.969', 'learning_rate': '2e-05', 'ppl': '2.339', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9560', 'tokens/trainable': 8862049, 'tokens/total': 8995758, 'epoch': '0.5'}

 12%|█▎        | 12/96 [16:18<1:49:20, 78.10s/it]
 14%|█▎        | 13/96 [17:36<1:47:44, 77.88s/it]
                                                 
{'loss': '0.8735', 'grad_norm': '3.25', 'learning_rate': '2e-05', 'ppl': '2.395', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9489', 'tokens/trainable': 9596186, 'tokens/total': 9745292, 'epoch': '0.5417'}

 14%|█▎        | 13/96 [17:36<1:47:44, 77.88s/it]
 15%|█▍        | 14/96 [18:53<1:46:18, 77.79s/it]
                                                 
{'loss': '0.8467', 'grad_norm': '2.234', 'learning_rate': '2e-05', 'ppl': '2.332', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9480', 'tokens/trainable': 10331619, 'tokens/total': 10494275, 'epoch': '0.5833'}

 15%|█▍        | 14/96 [18:53<1:46:18, 77.79s/it]
 16%|█▌        | 15/96 [20:11<1:44:48, 77.64s/it]
                                                 
{'loss': '0.9049', 'grad_norm': '2.453', 'learning_rate': '2e-05', 'ppl': '2.472', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9656', 'tokens/trainable': 11077979, 'tokens/total': 11243240, 'epoch': '0.625'}

 16%|█▌        | 15/96 [20:11<1:44:48, 77.64s/it]
 17%|█▋        | 16/96 [21:28<1:43:34, 77.68s/it]
                                                 
{'loss': '0.8693', 'grad_norm': '2.672', 'learning_rate': '2e-05', 'ppl': '2.385', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9484', 'tokens/trainable': 11815398, 'tokens/total': 11991989, 'epoch': '0.6667'}

 17%|█▋        | 16/96 [21:28<1:43:34, 77.68s/it]
 18%|█▊        | 17/96 [22:46<1:42:22, 77.76s/it]
                                                 
{'loss': '0.8343', 'grad_norm': '2.891', 'learning_rate': '2e-05', 'ppl': '2.303', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9529', 'tokens/trainable': 12558179, 'tokens/total': 12741571, 'epoch': '0.7083'}

 18%|█▊        | 17/96 [22:46<1:42:22, 77.76s/it]
 19%|█▉        | 18/96 [24:04<1:40:54, 77.62s/it]
                                                 
{'loss': '0.8577', 'grad_norm': '2.438', 'learning_rate': '2e-05', 'ppl': '2.358', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9480', 'tokens/trainable': 13291084, 'tokens/total': 13490669, 'epoch': '0.75'}

 19%|█▉        | 18/96 [24:04<1:40:54, 77.62s/it]
 20%|█▉        | 19/96 [25:21<1:39:36, 77.62s/it]
                                                 
{'loss': '0.8067', 'grad_norm': '2.438', 'learning_rate': '2e-05', 'ppl': '2.241', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9488', 'tokens/trainable': 14027534, 'tokens/total': 14239739, 'epoch': '0.7917'}

 20%|█▉        | 19/96 [25:21<1:39:36, 77.62s/it]
 21%|██        | 20/96 [26:39<1:38:27, 77.74s/it]
                                                 
{'loss': '0.8145', 'grad_norm': '2.422', 'learning_rate': '2e-05', 'ppl': '2.258', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9524', 'tokens/trainable': 14770394, 'tokens/total': 14988473, 'epoch': '0.8333'}

 21%|██        | 20/96 [26:39<1:38:27, 77.74s/it]
 22%|██▏       | 21/96 [27:57<1:36:59, 77.59s/it]
                                                 
{'loss': '0.7839', 'grad_norm': '2.406', 'learning_rate': '2e-05', 'ppl': '2.19', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9443', 'tokens/trainable': 15499968, 'tokens/total': 15736924, 'epoch': '0.875'}

 22%|██▏       | 21/96 [27:57<1:36:59, 77.59s/it]
 23%|██▎       | 22/96 [29:15<1:35:57, 77.80s/it]
                                                 
{'loss': '0.8035', 'grad_norm': '1.961', 'learning_rate': '2e-05', 'ppl': '2.233', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9450', 'tokens/trainable': 16239694, 'tokens/total': 16485262, 'epoch': '0.9167'}

 23%|██▎       | 22/96 [29:15<1:35:57, 77.80s/it]
 24%|██▍       | 23/96 [30:32<1:34:36, 77.76s/it]
                                                 
{'loss': '0.8221', 'grad_norm': '2.031', 'learning_rate': '2e-05', 'ppl': '2.275', 'memory/max_active (GiB)': '58.05', 'memory/max_allocated (GiB)': '58.05', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9388', 'tokens/trainable': 16968820, 'tokens/total': 17233354, 'epoch': '0.9583'}

 24%|██▍       | 23/96 [30:33<1:34:36, 77.76s/it]
 25%|██▌       | 24/96 [31:46<1:31:53, 76.58s/it]
                                                 
{'loss': '0.7618', 'grad_norm': '2.062', 'learning_rate': '2e-05', 'ppl': '2.142', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '67.02', 'tokens/train_per_sec_per_gpu': '9486', 'tokens/trainable': 17669050, 'tokens/total': 17945280, 'epoch': '1'}

 25%|██▌       | 24/96 [31:46<1:31:53, 76.58s/it][2026-08-11 19:36:21,497] [INFO] [axolotl.core.trainers.base.evaluate:464] [PID:31] Running evaluation step...


  0%|          | 0/90 [00:00<?, ?it/s]

  2%|▏         | 2/90 [00:00<00:31,  2.80it/s]

  3%|▎         | 3/90 [00:00<00:23,  3.72it/s]

  4%|▍         | 4/90 [00:01<00:31,  2.74it/s]

  6%|▌         | 5/90 [00:01<00:33,  2.51it/s]

  7%|▋         | 6/90 [00:02<00:36,  2.28it/s]

  8%|▊         | 7/90 [00:02<00:29,  2.78it/s]

  9%|▉         | 8/90 [00:03<00:33,  2.41it/s]

 10%|█         | 9/90 [00:03<00:31,  2.58it/s]

 11%|█         | 10/90 [00:03<00:27,  2.87it/s]

 12%|█▏        | 11/90 [00:04<00:27,  2.83it/s]

 13%|█▎        | 12/90 [00:04<00:30,  2.58it/s]

 14%|█▍        | 13/90 [00:05<00:33,  2.30it/s]

 16%|█▌        | 14/90 [00:05<00:34,  2.20it/s]

 17%|█▋        | 15/90 [00:06<00:35,  2.09it/s]

 18%|█▊        | 16/90 [00:06<00:28,  2.62it/s]

 19%|█▉        | 17/90 [00:06<00:30,  2.37it/s]

 20%|██        | 18/90 [00:07<00:32,  2.21it/s]

 21%|██        | 19/90 [00:07<00:33,  2.12it/s]

 22%|██▏       | 20/90 [00:08<00:34,  2.06it/s]

 23%|██▎       | 21/90 [00:08<00:26,  2.57it/s]

 24%|██▍       | 22/90 [00:08<00:23,  2.86it/s]

 26%|██▌       | 23/90 [00:09<00:26,  2.48it/s]

 27%|██▋       | 24/90 [00:09<00:28,  2.28it/s]

 28%|██▊       | 25/90 [00:10<00:30,  2.16it/s]

 29%|██▉       | 26/90 [00:10<00:23,  2.75it/s]

 30%|███       | 27/90 [00:10<00:21,  2.93it/s]

 32%|███▏      | 29/90 [00:11<00:18,  3.28it/s]

 33%|███▎      | 30/90 [00:11<00:20,  2.98it/s]

 34%|███▍      | 31/90 [00:12<00:22,  2.59it/s]

 36%|███▌      | 32/90 [00:12<00:22,  2.53it/s]

 37%|███▋      | 33/90 [00:13<00:24,  2.33it/s]

 38%|███▊      | 34/90 [00:13<00:25,  2.19it/s]

 39%|███▉      | 35/90 [00:13<00:22,  2.41it/s]

 40%|████      | 36/90 [00:14<00:23,  2.34it/s]

 41%|████      | 37/90 [00:14<00:20,  2.65it/s]

 42%|████▏     | 38/90 [00:15<00:20,  2.56it/s]

 43%|████▎     | 39/90 [00:15<00:17,  2.93it/s]

 44%|████▍     | 40/90 [00:15<00:16,  3.02it/s]

 46%|████▌     | 41/90 [00:16<00:16,  2.95it/s]

 47%|████▋     | 42/90 [00:16<00:14,  3.27it/s]

 48%|████▊     | 43/90 [00:16<00:16,  2.86it/s]

 49%|████▉     | 44/90 [00:16<00:14,  3.16it/s]

 50%|█████     | 45/90 [00:17<00:14,  3.05it/s]

 51%|█████     | 46/90 [00:17<00:11,  3.74it/s]

 52%|█████▏    | 47/90 [00:17<00:11,  3.79it/s]

 53%|█████▎    | 48/90 [00:18<00:12,  3.46it/s]

 54%|█████▍    | 49/90 [00:18<00:12,  3.37it/s]

 56%|█████▌    | 50/90 [00:18<00:10,  3.72it/s]

 57%|█████▋    | 51/90 [00:19<00:13,  2.89it/s]

 58%|█████▊    | 52/90 [00:19<00:12,  3.00it/s]

 59%|█████▉    | 53/90 [00:19<00:14,  2.57it/s]

 60%|██████    | 54/90 [00:20<00:12,  2.89it/s]

 61%|██████    | 55/90 [00:20<00:14,  2.50it/s]

 62%|██████▏   | 56/90 [00:21<00:12,  2.62it/s]

 63%|██████▎   | 57/90 [00:21<00:10,  3.29it/s]

 64%|██████▍   | 58/90 [00:21<00:08,  3.78it/s]

 66%|██████▌   | 59/90 [00:21<00:10,  2.94it/s]

 67%|██████▋   | 60/90 [00:22<00:11,  2.53it/s]

 68%|██████▊   | 61/90 [00:22<00:10,  2.68it/s]

 69%|██████▉   | 62/90 [00:23<00:11,  2.47it/s]

 70%|███████   | 63/90 [00:23<00:09,  2.80it/s]

 71%|███████   | 64/90 [00:23<00:07,  3.30it/s]

 72%|███████▏  | 65/90 [00:24<00:09,  2.77it/s]

 73%|███████▎  | 66/90 [00:24<00:09,  2.47it/s]

 74%|███████▍  | 67/90 [00:24<00:08,  2.62it/s]

 76%|███████▌  | 68/90 [00:25<00:09,  2.37it/s]

 77%|███████▋  | 69/90 [00:25<00:08,  2.62it/s]

 78%|███████▊  | 70/90 [00:25<00:06,  2.94it/s]

 79%|███████▉  | 71/90 [00:26<00:06,  2.79it/s]

 80%|████████  | 72/90 [00:26<00:05,  3.16it/s]

 81%|████████  | 73/90 [00:26<00:05,  3.31it/s]

 82%|████████▏ | 74/90 [00:27<00:05,  2.72it/s]

 83%|████████▎ | 75/90 [00:27<00:06,  2.42it/s]

 84%|████████▍ | 76/90 [00:28<00:05,  2.69it/s]

 86%|████████▌ | 77/90 [00:28<00:04,  3.10it/s]

 87%|████████▋ | 78/90 [00:28<00:04,  2.83it/s]

 88%|████████▊ | 79/90 [00:29<00:04,  2.46it/s]

 89%|████████▉ | 80/90 [00:29<00:03,  2.70it/s]

 90%|█████████ | 81/90 [00:30<00:03,  2.38it/s]

 91%|█████████ | 82/90 [00:30<00:03,  2.27it/s]

 92%|█████████▏| 83/90 [00:31<00:03,  2.16it/s]

 93%|█████████▎| 84/90 [00:31<00:02,  2.07it/s]

 94%|█████████▍| 85/90 [00:32<00:02,  2.04it/s]

 96%|█████████▌| 86/90 [00:32<00:02,  1.99it/s]

 97%|█████████▋| 87/90 [00:32<00:01,  2.61it/s]

 98%|█████████▊| 88/90 [00:33<00:00,  2.35it/s]

 99%|█████████▉| 89/90 [00:33<00:00,  2.62it/s]

100%|██████████| 90/90 [00:33<00:00,  2.69it/s]
                                                 

                                               
{'eval_loss': '0.8244', 'eval_runtime': '35.13', 'eval_samples_per_second': '10.31', 'eval_steps_per_second': '2.591', 'eval_ppl': '2.281', 'memory/max_active (GiB)': '20.04', 'memory/max_allocated (GiB)': '20.04', 'memory/device_reserved (GiB)': '71.9', 'epoch': '1'}

 25%|██▌       | 24/96 [32:21<1:31:53, 76.58s/it]

100%|██████████| 90/90 [00:34<00:00,  2.69it/s]

                                               [2026-08-11 19:36:56,637] [INFO] [axolotl.core.trainers.base._save:928] [PID:31] Saving model checkpoint to ./finetune-model-output/checkpoint-24


Writing model shards:   0%|          | 0/1 [00:00<?, ?it/s]

Writing model shards: 100%|██████████| 1/1 [00:09<00:00,  9.37s/it]
Writing model shards: 100%|██████████| 1/1 [00:09<00:00,  9.37s/it]
[2026-08-11 19:37:30,375] [WARNING] [py.warnings._showwarnmsg:112] [PID:31] /root/.local/share/uv/python/cpython-3.12.13-linux-x86_64-gnu/lib/python3.12/multiprocessing/popen_fork.py:66: DeprecationWarning: This process (pid=31) is multi-threaded, use of fork() may lead to deadlocks in the child.
  self.pid = os.fork()


 26%|██▌       | 25/96 [34:17<1:57:04, 98.93s/it]
                                                 
{'loss': '0.645', 'grad_norm': '3.094', 'learning_rate': '2e-05', 'ppl': '1.906', 'memory/max_active (GiB)': '58.06', 'memory/max_allocated (GiB)': '58.06', 'memory/device_reserved (GiB)': '66.16', 'tokens/train_per_sec_per_gpu': '4903', 'tokens/trainable': 18409932, 'tokens/total': 18695242, 'epoch': '1.042'}

 26%|██▌       | 25/96 [34:17<1:57:04, 98.93s/it]