Hugging Face's logo

robbiemu
/
salamandra-2b

Text Generation
Transformers
GGUF
llama
Inference Endpoints
imatrix
Model card Files Community
salamandra-2b
File size: 2,560 Bytes 
5dadba4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
 
quantizations:
  - IQ2_XS
  - IQ3_M
  - IQ3_S
  - IQ3_XS
  - IQ3_XXS
  - IQ4_NL
  - IQ4_XS
  - Q3_K_L
  - Q3_K_M
  - Q3_K_S
  - Q4_K_M
  - Q4_K_S
  - Q5_K_M
  - Q5_K_S
  - Q6_K
  - Q8_0
  - TQ1_0
  - TQ2_0

allowed_quantization_types:
  - name: Q4_0
    size: 4.34G
    ppl: +0.4685
    details: Llama-3-8B
  - name: Q4_1
    size: 4.78G
    ppl: +0.4511
    details: Llama-3-8B
  - name: Q5_0
    size: 5.21G
    ppl: +0.1316
    details: Llama-3-8B
  - name: Q5_1
    size: 5.65G
    ppl: +0.1062
    details: Llama-3-8B
  - name: IQ2_XXS
    size: "2.06 bpw"
    type: quantization
  - name: IQ2_XS
    size: "2.31 bpw"
    type: quantization
  - name: IQ2_S
    size: "2.5 bpw"
    type: quantization
  - name: IQ2_M
    size: "2.7 bpw"
    type: quantization
  - name: IQ1_S
    size: "1.56 bpw"
    type: quantization
  - name: IQ1_M
    size: "1.75 bpw"
    type: quantization
  - name: TQ1_0
    size: "1.69 bpw"
    type: ternarization
  - name: TQ2_0
    size: "2.06 bpw"
    type: ternarization
  - name: Q2_K
    size: 2.96G
    ppl: +3.5199
    details: Llama-3-8B
  - name: Q2_K_S
    size: 2.96G
    ppl: +3.1836
    details: Llama-3-8B
  - name: IQ3_XXS
    size: "3.06 bpw"
    type: quantization
  - name: IQ3_S
    size: "3.44 bpw"
    type: quantization
  - name: IQ3_M
    size: "3.66 bpw"
    type: quantization mix
  - name: Q3_K
    alias: Q3_K_M
  - name: IQ3_XS
    size: "3.3 bpw"
    type: quantization
  - name: Q3_K_S
    size: 3.41G
    ppl: +1.6321
    details: Llama-3-8B
  - name: Q3_K_M
    size: 3.74G
    ppl: +0.6569
    details: Llama-3-8B
  - name: Q3_K_L
    size: 4.03G
    ppl: +0.5562
    details: Llama-3-8B
  - name: IQ4_NL
    size: "4.50 bpw"
    type: non-linear quantization
  - name: IQ4_XS
    size: "4.25 bpw"
    type: non-linear quantization
  - name: Q4_K
    alias: Q4_K_M
  - name: Q4_K_S
    size: 4.37G
    ppl: +0.2689
    details: Llama-3-8B
  - name: Q4_K_M
    size: 4.58G
    ppl: +0.1754
    details: Llama-3-8B
  - name: Q5_K
    alias: Q5_K_M
  - name: Q5_K_S
    size: 5.21G
    ppl: +0.1049
    details: Llama-3-8B
  - name: Q5_K_M
    size: 5.33G
    ppl: +0.0569
    details: Llama-3-8B
  - name: Q6_K
    size: 6.14G
    ppl: +0.0217
    details: Llama-3-8B
  - name: Q8_0
    size: 7.96G
    ppl: +0.0026
    details: Llama-3-8B
  - name: F16
    size: 14.00G
    ppl: +0.0020
    details: Mistral-7B
  - name: BF16
    size: 14.00G
    ppl: -0.0050
    details: Mistral-7B
  - name: F32
    size: 26.00G
    details: 7B
  - name: COPY
    description: Only copy tensors, no quantizing