| { |
| "model_type": "pivot", |
| "architectures": [ |
| "PivotModel" |
| ], |
| "auto_map": { |
| "AutoConfig": "configuration_pivot.PivotConfig", |
| "AutoModel": "modeling_pivot.PivotModel" |
| }, |
| "dsbt_config": { |
| "backbone": { |
| "encoder": "huggingface", |
| "model_id": "LiquidAI/LFM2.5-Encoder-350M", |
| "revision": "b886781f7c6f10ca9b7096e21b83e30a073c2f39", |
| "trust_remote_code": true |
| }, |
| "scorer": { |
| "type": "mlp", |
| "hidden_size": 1024 |
| }, |
| "data": { |
| "k_min": 2, |
| "k_max": 16, |
| "max_context_tokens": 512, |
| "max_option_tokens": 64, |
| "noul_true": "true", |
| "noul_false": "false", |
| "split_ratios": [ |
| 0.8, |
| 0.1, |
| 0.1 |
| ], |
| "family_sample_caps": { |
| "casehold": 0.15 |
| } |
| }, |
| "train": { |
| "epochs": 8, |
| "warmup_epochs": 2, |
| "min_pcgrad_epochs": 4, |
| "batch_size": 64, |
| "grad_accum": 1, |
| "lr_encoder": 1e-05, |
| "lr_scorer": 5e-05, |
| "lr_schedule": "cosine", |
| "lr_warmup_ratio": 0.05, |
| "lr_min_ratio": 0.1, |
| "weight_decay": 0.01, |
| "grad_clip": 1.0, |
| "precision": "bf16", |
| "seed": 42, |
| "decoupling": "pcgrad", |
| "log_every": 100, |
| "num_workers": 8, |
| "early_stop_patience": 2, |
| "lora": { |
| "enabled": false |
| } |
| }, |
| "eval": { |
| "k_strata": [ |
| 2, |
| 4, |
| 8, |
| 16 |
| ], |
| "ece_bins": 15, |
| "batch_size": 128, |
| "temperature_scaling": false |
| }, |
| "brier": { |
| "reduction": "mean_over_real_options" |
| }, |
| "serving": { |
| "max_context_tokens": 512, |
| "max_option_tokens": 128 |
| } |
| }, |
| "encoder_config": { |
| "transformers_version": "5.17.0", |
| "architectures": [ |
| "Lfm2BidirectionalForMaskedLM" |
| ], |
| "output_hidden_states": false, |
| "return_dict": true, |
| "dtype": "float32", |
| "chunk_size_feed_forward": 0, |
| "is_encoder_decoder": false, |
| "id2label": { |
| "0": "LABEL_0", |
| "1": "LABEL_1" |
| }, |
| "label2id": { |
| "LABEL_0": 0, |
| "LABEL_1": 1 |
| }, |
| "problem_type": null, |
| "vocab_size": 65536, |
| "hidden_size": 1024, |
| "intermediate_size": 6656, |
| "num_hidden_layers": 16, |
| "num_attention_heads": 16, |
| "num_key_value_heads": 8, |
| "max_position_embeddings": 128000, |
| "initializer_range": 0.02, |
| "norm_eps": 1e-05, |
| "use_cache": false, |
| "pad_token_id": 0, |
| "bos_token_id": 1, |
| "eos_token_id": 7, |
| "tie_word_embeddings": true, |
| "rope_parameters": { |
| "rope_theta": 1000000.0, |
| "rope_type": "default" |
| }, |
| "conv_bias": false, |
| "conv_L_cache": 3, |
| "block_multiple_of": 256, |
| "block_ffn_dim_multiplier": 1.0, |
| "block_auto_adjust_ff_dim": true, |
| "full_attn_idxs": null, |
| "layer_types": [ |
| "conv", |
| "conv", |
| "full_attention", |
| "conv", |
| "conv", |
| "full_attention", |
| "conv", |
| "conv", |
| "full_attention", |
| "conv", |
| "full_attention", |
| "conv", |
| "full_attention", |
| "conv", |
| "full_attention", |
| "conv" |
| ], |
| "_name_or_path": "LiquidAI/LFM2.5-Encoder-350M", |
| "block_dim": 1024, |
| "block_mlp_init_scale": 1.0, |
| "block_norm_eps": 1e-05, |
| "block_out_init_scale": 1.0, |
| "block_use_swiglu": true, |
| "block_use_xavier_init": true, |
| "conv_dim": 1024, |
| "conv_dim_out": 1024, |
| "conv_use_xavier_init": true, |
| "model_type": "lfm2", |
| "num_heads": 16, |
| "use_pos_enc": true, |
| "auto_map": { |
| "AutoModel": "modeling_lfm2_bidirectional.Lfm2BidirectionalModel", |
| "AutoModelForMaskedLM": "modeling_lfm2_bidirectional.Lfm2BidirectionalForMaskedLM" |
| }, |
| "output_attentions": false |
| }, |
| "pivot_training": { |
| "selected_epoch": 7, |
| "global_step": 18008, |
| "method": "CE warmup -> live multiclass Brier + PCGrad", |
| "probability_contract": "softmax(logits.float())" |
| } |
| } |
|
|
