Qwen3-8B-a16w4 / devkit /vlm_config.json
florianvoss's picture
Upload folder using huggingface_hub
1c6f983 verified
{
"model_name": "Qwen3-8B",
"model_type": "llm-qwen3",
"vm_cfg": null,
"mm_cfg": null,
"lm_cfg": {
"model_type": "qwen3",
"data_type": "bfloat16",
"arch": "qwen",
"gen": "3",
"size": "8b",
"token_cfg": {
"vocab_size": 151936
},
"rope_cfg": {
"rope_theta": 1000000,
"rope_local_base_freq": 1000000,
"rope_scaling": {
"factor": 1.0,
"low_freq_factor": 0,
"high_freq_factor": 0,
"original_max_position_embeddings": 0,
"long_factor": null,
"short_factor": null,
"rope_type": "default"
}
},
"attn_cfg": {
"num_attention_heads": 32,
"num_key_value_heads": 8,
"head_dim": 128,
"swa_enable": false,
"swa_ratio": 0,
"sliding_window": null,
"attention_bias": false,
"attention_dropout": 0.0,
"query_pre_attn_scalar": 0
},
"mlp_cfg": {
"intermediate_size": 12288,
"act": "silu",
"num_layers": 3,
"mlp_bias": false
},
"hidden_size": 4096,
"num_hidden_layers": 36,
"max_position_embeddings": 2048,
"rms_norm_eps": 1e-06,
"rms_norm_unit_offset": false,
"layer_norms": [
"pre_attn",
"post_attn",
"qk_norm"
],
"attn_logit_softcapping": null,
"final_logit_softcapping": null,
"lm_head_num_splits": 3,
"lm_head_split_dim": 50656
},
"pipeline_cfg": {
"system_prompt": null,
"chat_template": null,
"max_num_tokens": 2048,
"input_token_group_size": 128,
"input_token_group_offsets": [
0,
128,
256,
384,
512,
640,
768,
896,
1024,
1152,
1280,
1408,
1536,
1664,
1792,
1920
],
"future_token_mask_size": 128,
"return_logits": false,
"use_strided_kv_cache": false,
"enable_filter_sharing": true
},
"language_model_name": "Qwen3-8B_language"
}