Hugging Face's logo

haoyang-amd
/
deepseek-moe-16b-chat-fp8-perchannel

Safetensors
deepseek
custom_code
fp8
Model card Files Community
1
deepseek-moe-16b-chat-fp8-perchannel
File size: 1,717 Bytes 
f66e03b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
 
{
  "_name_or_path": "/group/amdneuralopt/huggingface/pretrained_models/deepseek-ai/deepseek-moe-16b-chat",
  "architectures": [
    "DeepseekForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration_deepseek.DeepseekConfig",
    "AutoModel": "modeling_deepseek.DeepseekModel",
    "AutoModelForCausalLM": "modeling_deepseek.DeepseekForCausalLM"
  },
  "aux_loss_alpha": 0.001,
  "bos_token_id": 100000,
  "eos_token_id": 100001,
  "first_k_dense_replace": 1,
  "hidden_act": "silu",
  "hidden_size": 2048,
  "initializer_range": 0.02,
  "intermediate_size": 10944,
  "max_position_embeddings": 4096,
  "model_type": "deepseek",
  "moe_intermediate_size": 1408,
  "moe_layer_freq": 1,
  "n_routed_experts": 64,
  "n_shared_experts": 2,
  "norm_topk_prob": false,
  "num_attention_heads": 16,
  "num_experts_per_tok": 6,
  "num_hidden_layers": 28,
  "num_key_value_heads": 16,
  "pretraining_tp": 1,
  "quantization_config": {
    "activation_scheme": "static",
    "export": {
      "kv_cache_group": [
        "*k_proj",
        "*v_proj"
      ],
      "min_kv_scale": 1.0,
      "pack_method": "reorder",
      "weight_format": "real_quantized",
      "weight_merge_groups": null
    },
    "ignored_layers": [
      "lm_head"
    ],
    "int4_experts": {
      "bits": 4,
      "group": "column",
      "sym": true
    },
    "kv_cache_scheme": "static",
    "quant_method": "fp8"
  },
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 10000,
  "scoring_func": "softmax",
  "seq_aux": true,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.47.1",
  "use_cache": true,
  "vocab_size": 102400
}