Nanbeige4.2-3B-EAGLE3 / config.json
NullSense's picture
EAGLE-3 draft model for Nanbeige4.2-3B: epoch-3 head, looped-arch plugin fixes, training curves
b3d9511 verified
Raw
History Blame Contribute Delete
1.79 kB
{
"architectures": [
"Eagle3DraftModel"
],
"auto_map": {
"": "config.Eagle3SpeculatorConfig"
},
"draft_vocab_size": 32000,
"dtype": "bfloat16",
"eagle_aux_hidden_state_layer_ids": [
3,
23,
41
],
"embed_requires_grad": false,
"fc_norm": false,
"norm_before_fc": true,
"norm_before_residual": true,
"norm_output": true,
"speculators_config": {
"algorithm": "eagle3",
"default_proposal_method": "greedy",
"proposal_methods": [
{
"accept_tolerance": 0.0,
"proposal_type": "greedy",
"speculative_tokens": 3,
"verifier_accept_k": 1
}
],
"verifier": {
"architectures": [
"NanbeigeForCausalLM"
],
"name_or_path": "NullSense/Nanbeige4.2-3B-FP8-Dynamic"
}
},
"speculators_model_type": "eagle3",
"speculators_version": "0.6.0.dev0",
"target_hidden_size": null,
"tie_word_embeddings": false,
"transformer_layer_config": {
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 1,
"eos_token_id": 2,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 3072,
"initializer_range": 0.02,
"intermediate_size": 10752,
"layer_types": [
"sliding_attention"
],
"max_position_embeddings": 262144,
"mlp_bias": false,
"model_type": "llama",
"num_attention_heads": 48,
"num_hidden_layers": 1,
"num_key_value_heads": 8,
"pad_token_id": null,
"pretraining_tp": 1,
"rms_norm_eps": 1e-05,
"rope_parameters": {
"rope_theta": 10000.0,
"rope_type": "default"
},
"sliding_window": 2048,
"tie_word_embeddings": false,
"use_cache": true,
"use_sliding_window": true,
"vocab_size": 166144
},
"transformers_version": "5.14.1"
}