Views
No views yet
main1{
2 "model_name_or_path": "allenai/OLMo-2-0425-1B",
3 "revision": "main",
4 "training_text_corpora": [
5 "LeoZotos/bio_textbooks"
6 ],
7 "training_qa_corpora": [],
8 "splits": "all",
9 "lr": 3e-05,
10 "weight_decay": 0.01,
11 "adam_beta2": 0.95,
12 "num_epochs": 1,
13 "batch_size": 4,
14 "gradient_accumulation_steps": 8,
15 "max_steps": null,
16 "warmup_ratio": 0.03,
17 "lr_scheduler_type": "cosine",
18 "max_seq_length": 2048,
19 "min_len": 50,
20 "max_len": 2000,
21 "max_examples": null,
22 "max_qa_examples": null,
23 "mask_prompt_loss": false,
24 "seed": 42,
25 "layer_range": null,
26 "gradient_checkpointing": true,
27 "model_config_path": null,
28 "torch_dtype": null,
29 "eval_every_n_steps": 25,
30 "eval_datasets": [],
31 "eval_num_questions": 100,
32 "eval_permutations": 2,
33 "eval_batch_size": 32,
34 "eval_max_new_tokens": 2048,
35 "eval_temperature": 0.6,
36 "eval_top_p": 0.95,
37 "eval_enable_thinking": null,
38 "eval_export_test_results": false,
39 "eval_keep_text_output": false,
40 "save_local": "/scratch-shared/lzotos/local_models/OLMo-2-0425-1B_bio_cpt",
41 "local_checkpoints_dir": null,
42 "checkpoint_token_interval_m": 10,
43 "push_all_checkpoints_to_hub": false,
44 "push_last_checkpoint_to_hub": true,
45 "max_level": null,
46 "levels": null,
47 "subjects": null
48}