qwen3-8b-eagle3 / config.json
junuxyz's picture
Add files using upload-large-folder tool
7c80c1e verified
Raw
History Blame Contribute Delete
1.65 kB
{
"architectures": [
"Eagle3Speculator"
],
"draft_vocab_size": 32000,
"eagle_aux_hidden_state_layer_ids": [
2,
18,
33
],
"model_type": "speculators",
"norm_before_residual": true,
"norm_before_fc": false,
"fc_norm": false,
"norm_output": false,
"speculators_model_type": "eagle3",
"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": [
"Qwen3ForCausalLM"
],
"name_or_path": "Qwen/Qwen3-8B"
}
},
"transformer_layer_config": {
"_name_or_path": "Qwen/Qwen3-8B",
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 151643,
"dtype": "bfloat16",
"eos_token_id": 151645,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 4096,
"initializer_range": 0.02,
"intermediate_size": 12288,
"layer_types": [
"full_attention"
],
"max_position_embeddings": 40960,
"max_window_layers": 36,
"model_type": "qwen3",
"num_attention_heads": 32,
"num_hidden_layers": 1,
"num_key_value_heads": 8,
"pad_token_id": null,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"rope_theta": 1000000,
"rope_type": "default"
},
"sliding_window": null,
"tie_word_embeddings": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}
}