Hy3-preview-JANGTQ initial release: 79 GB JANGTQ2 (renamed JANGTQ), MTP preserved-disabled
c2d6f85 verified | { | |
| "architectures": [ | |
| "HYV3ForCausalLM" | |
| ], | |
| "bos_token_id": 120000, | |
| "enable_attention_fp32_softmax": false, | |
| "enable_lm_head_fp32": true, | |
| "enable_moe_fp32_combine": false, | |
| "eod_token_id": 120026, | |
| "eos_token_id": 120025, | |
| "expert_hidden_dim": 1536, | |
| "moe_intermediate_size": 1536, | |
| "first_k_dense_replace": 1, | |
| "head_dim": 128, | |
| "hidden_act": "silu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.006, | |
| "intermediate_size": 13312, | |
| "max_position_embeddings": 262144, | |
| "model_type": "hy_v3", | |
| "moe_router_enable_expert_bias": true, | |
| "moe_router_use_sigmoid": true, | |
| "num_attention_heads": 64, | |
| "num_experts": 192, | |
| "num_experts_per_tok": 8, | |
| "num_hidden_layers": 80, | |
| "num_key_value_heads": 8, | |
| "num_shared_experts": 1, | |
| "output_router_logits": true, | |
| "pad_token_id": 120002, | |
| "qk_norm": true, | |
| "rms_norm_eps": 1e-05, | |
| "rope_parameters": { | |
| "rope_theta": 11158840.0, | |
| "rope_type": "default" | |
| }, | |
| "route_norm": true, | |
| "router_scaling_factor": 2.826, | |
| "sep_token_id": 120007, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "5.6.0", | |
| "use_cache": true, | |
| "use_grouped_mm": false, | |
| "vocab_size": 120832, | |
| "num_nextn_predict_layers": 1, | |
| "weight_format": "mxtq", | |
| "mxtq_bits": { | |
| "routed_expert": 2, | |
| "attention": 8, | |
| "shared_expert": 8, | |
| "dense_ffn": 8, | |
| "mtp": 8, | |
| "embed_tokens": 8, | |
| "lm_head": 8, | |
| "norms_router_biases": 16 | |
| }, | |
| "mxtq_seed": 42, | |
| "quantization": { | |
| "bits": 8, | |
| "group_size": 64, | |
| "mode": "affine", | |
| "routed_expert_bits": 2, | |
| "mxtq_bits": { | |
| "routed_expert": 2, | |
| "attention": 8, | |
| "shared_expert": 8, | |
| "dense_ffn": 8, | |
| "mtp": 8, | |
| "embed_tokens": 8, | |
| "lm_head": 8, | |
| "norms_router_biases": 16 | |
| }, | |
| "expert_layout": "per_expert" | |
| }, | |
| "capabilities": { | |
| "reasoning_parser": "qwen3", | |
| "tool_parser": "hunyuan", | |
| "think_in_template": false, | |
| "supports_tools": true, | |
| "supports_thinking": true, | |
| "family": "hy_v3", | |
| "modality": "text", | |
| "cache_type": "kv" | |
| }, | |
| "runtime": { | |
| "bundle_has_mtp": true, | |
| "mtp_layers": 1, | |
| "mtp_mode": "preserved_disabled", | |
| "mtp_status": "MTP tensors are preserved in the bundle, but the first JANG runtime path must use normal autoregressive decode until an accept/reject speculative loop is implemented and tested." | |
| } | |
| } | |