Skip to content

Commit d4a624c

Browse files
author
Judd
committed
more models
1 parent c06972a commit d4a624c

File tree

2 files changed

+174
-2
lines changed

2 files changed

+174
-2
lines changed

‎scripts/model_downloader.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -51,7 +51,7 @@ def show_model(m):
5151
show_variants(info['variants'], info['default'])
5252
print()
5353

54-
for m in all_models.keys():
54+
for m in sorted(all_models.keys()):
5555
show_model(m)
5656

5757
def parse_model_id(model_id: str):

‎scripts/models.json

Lines changed: 173 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,156 @@
11
{
2+
"yi-1.5": {
3+
"brief": "Yi 1.5 is a high-performing, bilingual language model.",
4+
"default": "6b",
5+
"license": "Apache License Version 2.0",
6+
"variants": {
7+
"6b": {
8+
"default": "q8",
9+
"quantized": {
10+
"q8": {
11+
"size": 6441564880,
12+
"url": "chatllm_quantized_yi1.5/yi1.5-6b.bin"
13+
}
14+
}
15+
},
16+
"9b": {
17+
"default": "q8",
18+
"quantized": {
19+
"q8": {
20+
"size": 9383354064,
21+
"url": "chatllm_quantized_yi1.5/yi1.5-9b.bin"
22+
}
23+
}
24+
},
25+
"9b-16k": {
26+
"default": "q4_1",
27+
"quantized": {
28+
"q4_1": {
29+
"size": 5520662224,
30+
"url": "chatllm_quantized_yi1.5/yi1.5-9b-16k_q4_1.bin"
31+
}
32+
}
33+
}
34+
}
35+
},
36+
"deepseek-v2": {
37+
"brief": "A strong, economical, and efficient Mixture-of-Experts language model.",
38+
"default": "light",
39+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
40+
"variants": {
41+
"light": {
42+
"default": "q8",
43+
"quantized": {
44+
"q8": {
45+
"size": 16691737872,
46+
"url": "chatllm_quantized_deepseek/deepseekv2-lite.bin"
47+
}
48+
}
49+
}
50+
}
51+
},
52+
"deepseek-coder-v2": {
53+
"brief": "An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.",
54+
"default": "light",
55+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
56+
"variants": {
57+
"light": {
58+
"default": "q4_1",
59+
"quantized": {
60+
"q8": {
61+
"size": 16691738304,
62+
"url": "chatllm_quantized_deepseek/deepseek-coder-v2-lite-instruct.bin"
63+
},
64+
"q4_1": {
65+
"size": 9820206784,
66+
"url": "chatllm_quantized_deepseek/deepseek-coder-v2-lite-instruct_q4_1.bin"
67+
}
68+
}
69+
}
70+
}
71+
},
72+
"deepseek-coder-v2-base": {
73+
"brief": "An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.",
74+
"default": "light",
75+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
76+
"variants": {
77+
"light": {
78+
"default": "q4_1",
79+
"quantized": {
80+
"q4_1": {
81+
"size": 9820206784,
82+
"url": "chatllm_quantized_deepseek/deepseek-coder-v2-lite-base_q4_1.bin"
83+
}
84+
}
85+
}
86+
}
87+
},
88+
"deepseek-llm": {
89+
"brief": "An advanced language model crafted with 2 trillion bilingual tokens.",
90+
"default": "7b",
91+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
92+
"variants": {
93+
"7b": {
94+
"default": "q8",
95+
"quantized": {
96+
"q8": {
97+
"size": 7345830176,
98+
"url": "chatllm_quantized_deepseek/deepseek-7b.bin"
99+
}
100+
}
101+
}
102+
}
103+
},
104+
"deepseek-coder": {
105+
"brief": "DeepSeek Coder is a capable coding model trained on two trillion code and natural language tokens.",
106+
"default": "1.3b",
107+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
108+
"variants": {
109+
"1.3b": {
110+
"default": "q8",
111+
"quantized": {
112+
"q8": {
113+
"size": 1431733904,
114+
"url": "chatllm_quantized_deepseek/deepseekcoder-1.3b.bin"
115+
}
116+
}
117+
},
118+
"7b": {
119+
"default": "q8",
120+
"quantized": {
121+
"q8": {
122+
"size": 7163394192,
123+
"url": "chatllm_quantized_deepseek/deepseekcoder-7b.bin"
124+
}
125+
}
126+
}
127+
}
128+
},
129+
"deepseek-coder-base": {
130+
"brief": "DeepSeek Coder is a capable coding model trained on two trillion code and natural language tokens.",
131+
"default": "1.3b",
132+
"license": "DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023",
133+
"variants": {
134+
"1.3b": {
135+
"default": "q8",
136+
"quantized": {
137+
"q8": {
138+
"size": 1431733904,
139+
"url": "chatllm_quantized_deepseek/deepseekcoder-1.3b-base.bin"
140+
}
141+
}
142+
},
143+
"6.7b": {
144+
"default": "q8",
145+
"quantized": {
146+
"q8": {
147+
"size": 7163394192,
148+
"url": "chatllm_quantized_deepseek/deepseekcoder-6.7b-base.bin"
149+
}
150+
}
151+
}
152+
}
153+
},
2154
"gemma": {
3155
"brief": "Gemma is a family of lightweight, state-of-the-art open models built by Google DeepMind. Updated to version 1.1.",
4156
"default": "2b",
@@ -25,7 +177,27 @@
25177
"quantized": {
26178
"q4_1": {
27179
"size": 5025629392,
28-
"url": "chatllm_quantized_models/llama3-8b-q4_1.bin"
180+
"url": "chatllm_quantized_llama3/llama3-8b-q4_1.bin"
181+
},
182+
"q8": {
183+
"size": 8538752208,
184+
"url": "chatllm_quantized_llama3/llama3-8b.bin"
185+
}
186+
}
187+
}
188+
}
189+
},
190+
"llama3-chinese-lora": {
191+
"brief": "Llama-3-Chinese-8B-Instruct-LoRA, which is further tuned with 5M instruction data on Llama-3-Chinese-8B.",
192+
"default": "8b",
193+
"license": "",
194+
"variants": {
195+
"8b": {
196+
"default": "q4_0",
197+
"quantized": {
198+
"q4_0": {
199+
"size": 4523754704,
200+
"url": "chatllm_quantized_llama3/llama3-8b-lora-q4_0.bin"
29201
}
30202
}
31203
}

0 commit comments

Comments
 (0)