Text Generation
Transformers
Safetensors
hy_v4
hunyuan
hy4
Mixture of Experts
fp8
quantized
conversational
modelopt
Instructions to use tencent/Hy4-preview-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use tencent/Hy4-preview-FP8 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="tencent/Hy4-preview-FP8") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("tencent/Hy4-preview-FP8", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use tencent/Hy4-preview-FP8 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "tencent/Hy4-preview-FP8" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "tencent/Hy4-preview-FP8", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/tencent/Hy4-preview-FP8
- SGLang
How to use tencent/Hy4-preview-FP8 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "tencent/Hy4-preview-FP8" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "tencent/Hy4-preview-FP8", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "tencent/Hy4-preview-FP8" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "tencent/Hy4-preview-FP8", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use tencent/Hy4-preview-FP8 with Docker Model Runner:
docker model run hf.co/tencent/Hy4-preview-FP8
| { | |
| "architectures": [ | |
| "HYV4ForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bitwise_backward_align": false, | |
| "bos_token_id": 120000, | |
| "dtype": "bfloat16", | |
| "enable_ihc": true, | |
| "enable_lm_head_fp32": true, | |
| "eos_token_id": 120025, | |
| "gated_mla": true, | |
| "gating_type": "elementwise", | |
| "hc_eps": 1e-06, | |
| "hc_magnitude": 2.0, | |
| "hc_mult": 4, | |
| "head_dim": 64, | |
| "hidden_act": "silu", | |
| "hidden_size": 6144, | |
| "index_head_dim": 128, | |
| "index_n_heads": 32, | |
| "index_topk": 2048, | |
| "indexer_types": [ | |
| "full", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full", | |
| "shared", | |
| "shared", | |
| "shared", | |
| "full" | |
| ], | |
| "initializer_range": 0.006, | |
| "intermediate_size": 18432, | |
| "kv_lora_rank": 512, | |
| "layer_types": [ | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention", | |
| "deepseek_sparse_attention" | |
| ], | |
| "learnable_sink": true, | |
| "learnable_sink_init": 0.0, | |
| "max_position_embeddings": 1048576, | |
| "mlp_layer_types": [ | |
| "dense", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse", | |
| "sparse" | |
| ], | |
| "model_type": "hy_v4", | |
| "moe_intermediate_size": 2048, | |
| "mtp_loss_factor": 0.1, | |
| "n_group": 1, | |
| "n_routed_experts": 256, | |
| "n_shared_experts": 1, | |
| "norm_topk_prob": true, | |
| "num_attention_heads": 64, | |
| "num_experts_per_tok": 8, | |
| "num_hidden_layers": 78, | |
| "num_key_value_heads": 8, | |
| "num_nextn_predict_layers": 1, | |
| "pad_token_id": 120002, | |
| "q_lora_rank": 2048, | |
| "qk_head_dim": 256, | |
| "qk_nope_head_dim": 192, | |
| "qk_rope_head_dim": 64, | |
| "rms_norm_eps": 1e-05, | |
| "rope_parameters": { | |
| "rope_theta": 10000000, | |
| "rope_type": "default" | |
| }, | |
| "routed_scaling_factor": 2.827, | |
| "swiglu_limit": 10.0, | |
| "tie_word_embeddings": false, | |
| "topk_group": 1, | |
| "transformers_version": "5.16.2", | |
| "use_cache": true, | |
| "use_dsa": true, | |
| "use_mla": true, | |
| "v_head_dim": 256, | |
| "vocab_size": 120832, | |
| "quantization_config": { | |
| "quant_method": "modelopt", | |
| "quantization": { | |
| "quant_algo": "MXFP8", | |
| "kv_cache_quant_algo": null, | |
| "exclude_modules": [ | |
| "lm_head", | |
| "model.embed_tokens", | |
| "model.hc_head.hc_head_fn", | |
| "model.layers.0.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.0.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.0.self_attn.indexer.weights_proj", | |
| "model.layers.0.self_attn.linear_gate", | |
| "model.layers.1.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.1.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.1.mlp.gate", | |
| "model.layers.1.self_attn.indexer.weights_proj", | |
| "model.layers.1.self_attn.linear_gate", | |
| "model.layers.10.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.10.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.10.mlp.gate", | |
| "model.layers.10.self_attn.linear_gate", | |
| "model.layers.11.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.11.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.11.mlp.gate", | |
| "model.layers.11.self_attn.linear_gate", | |
| "model.layers.12.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.12.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.12.mlp.gate", | |
| "model.layers.12.self_attn.linear_gate", | |
| "model.layers.13.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.13.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.13.mlp.gate", | |
| "model.layers.13.self_attn.indexer.weights_proj", | |
| "model.layers.13.self_attn.linear_gate", | |
| "model.layers.14.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.14.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.14.mlp.gate", | |
| "model.layers.14.self_attn.linear_gate", | |
| "model.layers.15.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.15.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.15.mlp.gate", | |
| "model.layers.15.self_attn.linear_gate", | |
| "model.layers.16.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.16.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.16.mlp.gate", | |
| "model.layers.16.self_attn.linear_gate", | |
| "model.layers.17.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.17.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.17.mlp.gate", | |
| "model.layers.17.self_attn.indexer.weights_proj", | |
| "model.layers.17.self_attn.linear_gate", | |
| "model.layers.18.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.18.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.18.mlp.gate", | |
| "model.layers.18.self_attn.linear_gate", | |
| "model.layers.19.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.19.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.19.mlp.gate", | |
| "model.layers.19.self_attn.linear_gate", | |
| "model.layers.2.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.2.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.2.mlp.gate", | |
| "model.layers.2.self_attn.linear_gate", | |
| "model.layers.20.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.20.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.20.mlp.gate", | |
| "model.layers.20.self_attn.linear_gate", | |
| "model.layers.21.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.21.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.21.mlp.gate", | |
| "model.layers.21.self_attn.indexer.weights_proj", | |
| "model.layers.21.self_attn.linear_gate", | |
| "model.layers.22.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.22.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.22.mlp.gate", | |
| "model.layers.22.self_attn.linear_gate", | |
| "model.layers.23.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.23.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.23.mlp.gate", | |
| "model.layers.23.self_attn.linear_gate", | |
| "model.layers.24.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.24.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.24.mlp.gate", | |
| "model.layers.24.self_attn.linear_gate", | |
| "model.layers.25.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.25.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.25.mlp.gate", | |
| "model.layers.25.self_attn.indexer.weights_proj", | |
| "model.layers.25.self_attn.linear_gate", | |
| "model.layers.26.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.26.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.26.mlp.gate", | |
| "model.layers.26.self_attn.linear_gate", | |
| "model.layers.27.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.27.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.27.mlp.gate", | |
| "model.layers.27.self_attn.linear_gate", | |
| "model.layers.28.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.28.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.28.mlp.gate", | |
| "model.layers.28.self_attn.linear_gate", | |
| "model.layers.29.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.29.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.29.mlp.gate", | |
| "model.layers.29.self_attn.indexer.weights_proj", | |
| "model.layers.29.self_attn.linear_gate", | |
| "model.layers.3.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.3.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.3.mlp.gate", | |
| "model.layers.3.self_attn.linear_gate", | |
| "model.layers.30.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.30.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.30.mlp.gate", | |
| "model.layers.30.self_attn.linear_gate", | |
| "model.layers.31.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.31.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.31.mlp.gate", | |
| "model.layers.31.self_attn.linear_gate", | |
| "model.layers.32.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.32.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.32.mlp.gate", | |
| "model.layers.32.self_attn.linear_gate", | |
| "model.layers.33.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.33.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.33.mlp.gate", | |
| "model.layers.33.self_attn.indexer.weights_proj", | |
| "model.layers.33.self_attn.linear_gate", | |
| "model.layers.34.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.34.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.34.mlp.gate", | |
| "model.layers.34.self_attn.linear_gate", | |
| "model.layers.35.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.35.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.35.mlp.gate", | |
| "model.layers.35.self_attn.linear_gate", | |
| "model.layers.36.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.36.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.36.mlp.gate", | |
| "model.layers.36.self_attn.linear_gate", | |
| "model.layers.37.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.37.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.37.mlp.gate", | |
| "model.layers.37.self_attn.indexer.weights_proj", | |
| "model.layers.37.self_attn.linear_gate", | |
| "model.layers.38.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.38.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.38.mlp.gate", | |
| "model.layers.38.self_attn.linear_gate", | |
| "model.layers.39.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.39.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.39.mlp.gate", | |
| "model.layers.39.self_attn.linear_gate", | |
| "model.layers.4.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.4.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.4.mlp.gate", | |
| "model.layers.4.self_attn.linear_gate", | |
| "model.layers.40.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.40.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.40.mlp.gate", | |
| "model.layers.40.self_attn.linear_gate", | |
| "model.layers.41.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.41.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.41.mlp.gate", | |
| "model.layers.41.self_attn.indexer.weights_proj", | |
| "model.layers.41.self_attn.linear_gate", | |
| "model.layers.42.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.42.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.42.mlp.gate", | |
| "model.layers.42.self_attn.linear_gate", | |
| "model.layers.43.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.43.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.43.mlp.gate", | |
| "model.layers.43.self_attn.linear_gate", | |
| "model.layers.44.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.44.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.44.mlp.gate", | |
| "model.layers.44.self_attn.linear_gate", | |
| "model.layers.45.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.45.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.45.mlp.gate", | |
| "model.layers.45.self_attn.indexer.weights_proj", | |
| "model.layers.45.self_attn.linear_gate", | |
| "model.layers.46.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.46.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.46.mlp.gate", | |
| "model.layers.46.self_attn.linear_gate", | |
| "model.layers.47.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.47.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.47.mlp.gate", | |
| "model.layers.47.self_attn.linear_gate", | |
| "model.layers.48.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.48.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.48.mlp.gate", | |
| "model.layers.48.self_attn.linear_gate", | |
| "model.layers.49.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.49.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.49.mlp.gate", | |
| "model.layers.49.self_attn.indexer.weights_proj", | |
| "model.layers.49.self_attn.linear_gate", | |
| "model.layers.5.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.5.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.5.mlp.gate", | |
| "model.layers.5.self_attn.indexer.weights_proj", | |
| "model.layers.5.self_attn.linear_gate", | |
| "model.layers.50.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.50.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.50.mlp.gate", | |
| "model.layers.50.self_attn.linear_gate", | |
| "model.layers.51.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.51.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.51.mlp.gate", | |
| "model.layers.51.self_attn.linear_gate", | |
| "model.layers.52.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.52.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.52.mlp.gate", | |
| "model.layers.52.self_attn.linear_gate", | |
| "model.layers.53.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.53.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.53.mlp.gate", | |
| "model.layers.53.self_attn.indexer.weights_proj", | |
| "model.layers.53.self_attn.linear_gate", | |
| "model.layers.54.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.54.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.54.mlp.gate", | |
| "model.layers.54.self_attn.linear_gate", | |
| "model.layers.55.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.55.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.55.mlp.gate", | |
| "model.layers.55.self_attn.linear_gate", | |
| "model.layers.56.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.56.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.56.mlp.gate", | |
| "model.layers.56.self_attn.linear_gate", | |
| "model.layers.57.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.57.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.57.mlp.gate", | |
| "model.layers.57.self_attn.indexer.weights_proj", | |
| "model.layers.57.self_attn.linear_gate", | |
| "model.layers.58.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.58.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.58.mlp.gate", | |
| "model.layers.58.self_attn.linear_gate", | |
| "model.layers.59.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.59.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.59.mlp.gate", | |
| "model.layers.59.self_attn.linear_gate", | |
| "model.layers.6.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.6.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.6.mlp.gate", | |
| "model.layers.6.self_attn.linear_gate", | |
| "model.layers.60.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.60.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.60.mlp.gate", | |
| "model.layers.60.self_attn.linear_gate", | |
| "model.layers.61.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.61.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.61.mlp.gate", | |
| "model.layers.61.self_attn.indexer.weights_proj", | |
| "model.layers.61.self_attn.linear_gate", | |
| "model.layers.62.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.62.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.62.mlp.gate", | |
| "model.layers.62.self_attn.linear_gate", | |
| "model.layers.63.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.63.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.63.mlp.gate", | |
| "model.layers.63.self_attn.linear_gate", | |
| "model.layers.64.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.64.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.64.mlp.gate", | |
| "model.layers.64.self_attn.linear_gate", | |
| "model.layers.65.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.65.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.65.mlp.gate", | |
| "model.layers.65.self_attn.indexer.weights_proj", | |
| "model.layers.65.self_attn.linear_gate", | |
| "model.layers.66.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.66.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.66.mlp.gate", | |
| "model.layers.66.self_attn.linear_gate", | |
| "model.layers.67.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.67.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.67.mlp.gate", | |
| "model.layers.67.self_attn.linear_gate", | |
| "model.layers.68.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.68.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.68.mlp.gate", | |
| "model.layers.68.self_attn.linear_gate", | |
| "model.layers.69.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.69.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.69.mlp.gate", | |
| "model.layers.69.self_attn.indexer.weights_proj", | |
| "model.layers.69.self_attn.linear_gate", | |
| "model.layers.7.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.7.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.7.mlp.gate", | |
| "model.layers.7.self_attn.linear_gate", | |
| "model.layers.70.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.70.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.70.mlp.gate", | |
| "model.layers.70.self_attn.linear_gate", | |
| "model.layers.71.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.71.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.71.mlp.gate", | |
| "model.layers.71.self_attn.linear_gate", | |
| "model.layers.72.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.72.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.72.mlp.gate", | |
| "model.layers.72.self_attn.linear_gate", | |
| "model.layers.73.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.73.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.73.mlp.gate", | |
| "model.layers.73.self_attn.indexer.weights_proj", | |
| "model.layers.73.self_attn.linear_gate", | |
| "model.layers.74.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.74.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.74.mlp.gate", | |
| "model.layers.74.self_attn.linear_gate", | |
| "model.layers.75.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.75.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.75.mlp.gate", | |
| "model.layers.75.self_attn.linear_gate", | |
| "model.layers.76.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.76.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.76.mlp.gate", | |
| "model.layers.76.self_attn.linear_gate", | |
| "model.layers.77.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.77.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.77.mlp.gate", | |
| "model.layers.77.self_attn.indexer.weights_proj", | |
| "model.layers.77.self_attn.linear_gate", | |
| "model.layers.8.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.8.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.8.mlp.gate", | |
| "model.layers.8.self_attn.linear_gate", | |
| "model.layers.9.hc_attn_layer.hc_pre.hc_fn", | |
| "model.layers.9.hc_mlp_layer.hc_pre.hc_fn", | |
| "model.layers.9.mlp.gate", | |
| "model.layers.9.self_attn.indexer.weights_proj", | |
| "model.layers.9.self_attn.linear_gate", | |
| "model.mtp_layers.0.eh_proj", | |
| "model.mtp_layers.0.mlp.gate", | |
| "model.mtp_layers.0.self_attn.indexer.weights_proj", | |
| "model.mtp_layers.0.self_attn.linear_gate" | |
| ] | |
| } | |
| }, | |
| "torch_dtype": "bfloat16" | |
| } | |