| { | |
| "dataset.debug": false, | |
| "dataset.git_diff": "", | |
| "dataset.git_sha1": "unknown", | |
| "dataset.manual_sample_ids": [], | |
| "dataset.output_dir": "output", | |
| "dataset.path": "output/datasets/ds_EXAONE-3.5-7.8B-Instruct/", | |
| "dataset.run_name": "temp_run", | |
| "dataset.seed": 42, | |
| "dataset_generation.batch_size": 1, | |
| "dataset_generation.debug": false, | |
| "dataset_generation.debug_target": null, | |
| "dataset_generation.ds_prefix": "ds_", | |
| "dataset_generation.git_diff": "", | |
| "dataset_generation.git_sha1": "unknown", | |
| "dataset_generation.max_length": 4096, | |
| "dataset_generation.output_dir": "output", | |
| "dataset_generation.run_name": "temp_run", | |
| "dataset_generation.save_every": 1000, | |
| "dataset_generation.seed": 42, | |
| "inference.debug": false, | |
| "inference.detail_time_stats": false, | |
| "inference.draft_tree_shape": "mc_sim_7b_65", | |
| "inference.git_diff": "", | |
| "inference.git_sha1": "unknown", | |
| "inference.max_new_tokens": 512, | |
| "inference.mode": "speculative", | |
| "inference.output_dir": "output", | |
| "inference.run_name": "temp_run", | |
| "inference.seed": 42, | |
| "modeling.add_noise": true, | |
| "modeling.attention_offset": "random.randrange(0, 3)", | |
| "modeling.attention_wind": "5", | |
| "modeling.ckpt_path": null, | |
| "modeling.debug": false, | |
| "modeling.decoder_key_remap": {}, | |
| "modeling.dtype": "torch.bfloat16", | |
| "modeling.frozen_targets": [], | |
| "modeling.git_diff": "", | |
| "modeling.git_sha1": "unknown", | |
| "modeling.layer_path": "transformer.h", | |
| "modeling.lmhead_path": "lm_head", | |
| "modeling.model_path": "beagle/models/exaone3.5/EXAONE-3.5-7.8B-Instruct/", | |
| "modeling.norm_path": "transformer.ln_f", | |
| "modeling.output_dir": "output", | |
| "modeling.reuse_layer": null, | |
| "modeling.rotary_path": "transformer.rotary", | |
| "modeling.run_name": "temp_run", | |
| "modeling.seed": 42, | |
| "modeling.tokenizer_path": "LGAI-EXAONE/EXAONE-3.5-7.8B-Instruct", | |
| "modeling.use_fc_eagle": false, | |
| "modeling.use_lower_layers": 1, | |
| "modeling.use_state_distill": false, | |
| "training.adam_beta2": 0.95, | |
| "training.bf16": true, | |
| "training.ddp_find_unused_parameters": false, | |
| "training.debug": false, | |
| "training.eval_steps": 100, | |
| "training.eval_strategy": "steps", | |
| "training.filter_out_shorts": false, | |
| "training.git_diff": "", | |
| "training.git_sha1": "23ad81b6b032265fbba2d332e1635539350da0f6", | |
| "training.gradient_accumulation_steps": 4, | |
| "training.learning_rate": 3e-05, | |
| "training.logging_steps": 1, | |
| "training.lr_scheduler_type": "constant_with_warmup", | |
| "training.max_grad_norm": 0.5, | |
| "training.max_length": 4096, | |
| "training.max_steps": -1, | |
| "training.num_train_epochs": 20, | |
| "training.optim": "adamw_torch_fused", | |
| "training.output_dir": "output", | |
| "training.overwrite_output_dir": true, | |
| "training.per_device_eval_batch_size": 1, | |
| "training.per_device_train_batch_size": 1, | |
| "training.project": "beagle", | |
| "training.report_to": "wandb", | |
| "training.resume_from_checkpoint": false, | |
| "training.resume_wandb_runid": null, | |
| "training.run_name": "temp_run", | |
| "training.save_steps": 500, | |
| "training.save_strategy": "steps", | |
| "training.save_total_limit": 2, | |
| "training.save_vram": true, | |
| "training.seed": 42, | |
| "training.warmup_steps": 50 | |
| } |