tllm-checkpoint/config.json

{
    "mlp_bias": false,
    "attn_bias": false,
    "rotary_base": 1000000.0,
    "rotary_scaling": null,
    "disable_weight_only_quant_plugin": false,
    "num_labels": 1,
    "use_logn_attn": false,
    "moe": {
        "num_experts": 64,
        "shared_expert_intermediate_size": 20480,
        "top_k": 8,
        "normalization_mode": 0,
        "sparse_mixer_epsilon": 0.01,
        "tp_mode": 0,
        "device_limited_n_group": 0,
        "device_limited_topk_group": 0,
        "device_limited_routed_scaling_factor": 1.0
    },
    "architecture": "Qwen2MoeForCausalLM",
    "dtype": "bfloat16",
    "vocab_size": 151936,
    "hidden_size": 3584,
    "num_hidden_layers": 28,
    "num_attention_heads": 28,
    "hidden_act": "swiglu",
    "logits_dtype": "float32",
    "norm_epsilon": 1e-06,
    "runtime_defaults": null,
    "position_embedding_type": "rope_gpt_neox",
    "num_key_value_heads": 4,
    "intermediate_size": 18944,
    "max_position_embeddings": 32768,
    "mapping": {
        "world_size": 2,
        "gpus_per_node": 8,
        "cp_size": 1,
        "tp_size": 2,
        "pp_size": 1,
        "moe_tp_size": 2,
        "moe_ep_size": 1,
        "auto_parallel": false
    },
    "quantization": {
        "quant_algo": "W4A16",
        "kv_cache_quant_algo": null,
        "group_size": 128,
        "smoothquant_val": 0.5,
        "clamp_val": null,
        "use_meta_recipe": false,
        "has_zero_point": false,
        "pre_quant_scale": false,
        "exclude_modules": null
    },
    "use_parallel_embedding": false,
    "embedding_sharding_dim": 0,
    "head_size": 128,
    "qk_layernorm": false,
    "rotary_embedding_dim": 128,
    "seq_length": 8192,
    "qwen_type": "qwen2_moe",
    "moe_intermediate_size": 2560,
    "moe_shared_expert_intermediate_size": 20480,
    "tie_word_embeddings": false
}