config.json

{
  "architectures": [
    "FlexQwenForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "qwen.FlexQwenConfig",
    "AutoModel": "qwen.FlexQwen",
    "AutoModelForCausalLM": "qwen.FlexQwenForCausalLM",
    "AutoModelForSequenceClassification": "qwen.FlexQwenForSequenceClassification"
  },
  "cls_token_id": 1,
  "context_length": 4096,
  "embedding_dim": 1024,
  "head_dim": 128,
  "hidden_dim": 2048,
  "initializer_range": 0.02,
  "model_type": "flexqwen",
  "moe_hidden_dim": 512,
  "moe_num_experts": 0,
  "moe_num_experts_per_token": -1,
  "num_attention_heads": 8,
  "num_hidden_layers": 32,
  "num_kv_groups": 8,
  "pad_token_id": 3,
  "qk_norm": true,
  "rms_norm_eps": 1e-06,
  "rope_theta": 10000,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.51.3",
  "vocab_size": 64000
}