Skip to content

Commit 6ab4a1f

Browse files
nsarrazinnsarrazin
authored
feat(huggingchat): add Qwen 2.5 72B (#1481)
* feat(huggingchat): add Qwen 2.5 72B * config: bump temp to 0.6
1 parent 15ef09c commit 6ab4a1f

File tree

1 file changed

+174
-123
lines changed

1 file changed

+174
-123
lines changed

‎chart/env/prod.yaml

Lines changed: 174 additions & 123 deletions
Original file line numberDiff line numberDiff line change
@@ -39,191 +39,242 @@ envVars:
3939
METRICS_ENABLED: "true"
4040
MODELS: >
4141
[
42-
{
43-
"name" : "meta-llama/Meta-Llama-3.1-70B-Instruct",
42+
{
43+
"name": "meta-llama/Meta-Llama-3.1-70B-Instruct",
4444
"id": "meta-llama/Meta-Llama-3.1-70B-Instruct",
45-
"tokenizer": {"tokenizerUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/resolve/main/tokenizer.json", "tokenizerConfigUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/raw/main/tokenizer_config.json"},
45+
"tokenizer": {
46+
"tokenizerUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/resolve/main/tokenizer.json",
47+
"tokenizerConfigUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/raw/main/tokenizer_config.json"
48+
},
4649
"description": "Ideal for everyday use. A fast and extremely capable model matching closed source models' capabilities.",
4750
"modelUrl": "https://huggingface.co/meta-llama/Meta-Llama-3.1-70B-Instruct",
4851
"websiteUrl": "https://llama.meta.com/",
4952
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/meta-logo.png",
5053
"tools": true,
51-
"preprompt" : "",
54+
"preprompt": "",
5255
"parameters": {
53-
"temperature": 0.6,
54-
"top_p": 0.9,
55-
"stop": ["<|endoftext|>", "<|eot_id|>"],
56+
"temperature": 0.6,
57+
"top_p": 0.9,
58+
"stop": ["<|endoftext|>", "<|eot_id|>"],
5659
"max_new_tokens": 1024,
5760
"truncate": 7167
5861
},
5962
"promptExamples": [
6063
{
6164
"title": "Write an email from bullet list",
6265
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
63-
}, {
66+
},
67+
{
6468
"title": "Code a snake game",
6569
"prompt": "Code a basic snake game in python, give explanations for each step."
66-
}, {
70+
},
71+
{
6772
"title": "Assist in a task",
6873
"prompt": "How do I make a delicious lemon cheesecake?"
6974
}
7075
]
71-
},
76+
},
7277
{
73-
"name" : "CohereForAI/c4ai-command-r-plus-08-2024",
74-
"tokenizer": {"tokenizerUrl": "https://huggingface.co/nsarrazin/c4ai-command-r-v01-tokenizer/resolve/main/tokenizer.json", "tokenizerConfigUrl": "https://huggingface.co/nsarrazin/c4ai-command-r-v01-tokenizer/raw/main/tokenizer_config.json"},
78+
"name": "CohereForAI/c4ai-command-r-plus-08-2024",
79+
"tokenizer": {
80+
"tokenizerUrl": "https://huggingface.co/nsarrazin/c4ai-command-r-v01-tokenizer/resolve/main/tokenizer.json",
81+
"tokenizerConfigUrl": "https://huggingface.co/nsarrazin/c4ai-command-r-v01-tokenizer/raw/main/tokenizer_config.json"
82+
},
7583
"description": "Cohere's largest language model, optimized for conversational interaction and tool use. Now with the 2024 update!",
7684
"modelUrl": "https://huggingface.co/CohereForAI/c4ai-command-r-plus-08-2024",
7785
"websiteUrl": "https://docs.cohere.com/docs/command-r-plus",
7886
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/cohere-logo.png",
7987
"tools": true,
8088
"parameters": {
8189
"stop": ["<|END_OF_TURN_TOKEN|>"],
82-
"truncate" : 28672,
83-
"max_new_tokens" : 2048,
84-
"temperature" : 0.3
90+
"truncate": 28672,
91+
"max_new_tokens": 2048,
92+
"temperature": 0.3
8593
},
86-
"promptExamples" : [
94+
"promptExamples": [
8795
{
8896
"title": "Generate a mouse portrait",
8997
"prompt": "Generate the portrait of a scientific mouse in its laboratory."
90-
}, {
98+
},
99+
{
91100
"title": "Review a pull request",
92101
"prompt": "Review this pull request: https://github.com/huggingface/chat-ui/pull/1131/files"
93-
}, {
102+
},
103+
{
104+
"title": "Code a snake game",
105+
"prompt": "Code a basic snake game in python, give explanations for each step."
106+
}
107+
]
108+
},
109+
{
110+
"name": "Qwen/Qwen2.5-72B-Instruct",
111+
"tokenizer": "Qwen/Qwen2.5-72B-Instruct",
112+
"description": "The latest Qwen open model with improved role-playing, long text generation and structured data understanding.",
113+
"modelUrl": "https://huggingface.co/Qwen/Qwen2.5-72B-Instruct",
114+
"websiteUrl": "https://qwenlm.github.io/blog/qwen2.5/",
115+
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/qwen-logo.png",
116+
"preprompt": "",
117+
"parameters": {
118+
"temperature": 0.6,
119+
"stop": ["<|endoftext|>", "<|im_end|>"],
120+
"truncate": 28672,
121+
"max_new_tokens": 3072,
122+
},
123+
"promptExamples": [
124+
{
125+
"title": "Write an email from bullet list",
126+
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
127+
},
128+
{
94129
"title": "Code a snake game",
95130
"prompt": "Code a basic snake game in python, give explanations for each step."
131+
},
132+
{
133+
"title": "Assist in a task",
134+
"prompt": "How do I make a delicious lemon cheesecake?"
96135
}
97136
]
98137
},
99138
{
100-
"name" : "mistralai/Mixtral-8x7B-Instruct-v0.1",
101-
"description" : "A high-quality sparse mixture of experts model with open weights.",
139+
"name": "mistralai/Mixtral-8x7B-Instruct-v0.1",
140+
"description": "A high-quality sparse mixture of experts model with open weights.",
102141
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/mistral-logo.png",
103-
"websiteUrl" : "https://mistral.ai/news/mixtral-of-experts/",
142+
"websiteUrl": "https://mistral.ai/news/mixtral-of-experts/",
104143
"modelUrl": "https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1",
105144
"tokenizer": "mistralai/Mixtral-8x7B-Instruct-v0.1",
106-
"preprompt" : "",
107-
"parameters" : {
108-
"temperature" : 0.6,
109-
"top_p" : 0.95,
110-
"repetition_penalty" : 1.2,
111-
"top_k" : 50,
112-
"truncate" : 24576,
113-
"max_new_tokens" : 8192,
114-
"stop" : ["</s>"]
145+
"preprompt": "",
146+
"parameters": {
147+
"temperature": 0.6,
148+
"top_p": 0.95,
149+
"repetition_penalty": 1.2,
150+
"top_k": 50,
151+
"truncate": 24576,
152+
"max_new_tokens": 8192,
153+
"stop": ["</s>"]
115154
},
116-
"promptExamples" : [
155+
"promptExamples": [
117156
{
118157
"title": "Write an email from bullet list",
119158
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
120-
}, {
159+
},
160+
{
121161
"title": "Code a snake game",
122162
"prompt": "Code a basic snake game in python, give explanations for each step."
123-
}, {
163+
},
164+
{
124165
"title": "Assist in a task",
125166
"prompt": "How do I make a delicious lemon cheesecake?"
126167
}
127168
]
128169
},
129170
{
130-
"name" : "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
131-
"description" : "Nous Hermes' strong flagship model trained on the Mixtral 8x7B.",
132-
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/nous-logo.png",
133-
"websiteUrl" : "https://nousresearch.com/",
134-
"modelUrl": "https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
135-
"tokenizer": "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
136-
"promptExamples": [
137-
{
138-
"title": "Write an email from bullet list",
139-
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
140-
}, {
141-
"title": "Code a snake game",
142-
"prompt": "Code a basic snake game in python, give explanations for each step."
143-
}, {
144-
"title": "Assist in a task",
145-
"prompt": "How do I make a delicious lemon cheesecake?"
146-
}
147-
],
148-
"parameters": {
149-
"temperature": 0.7,
150-
"top_p": 0.95,
151-
"repetition_penalty": 1,
152-
"top_k": 50,
153-
"truncate": 24576,
154-
"max_new_tokens": 2048,
155-
"stop": ["<|im_end|>"]
171+
"name": "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
172+
"description": "Nous Hermes' strong flagship model trained on the Mixtral 8x7B.",
173+
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/nous-logo.png",
174+
"websiteUrl": "https://nousresearch.com/",
175+
"modelUrl": "https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
176+
"tokenizer": "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO",
177+
"promptExamples": [
178+
{
179+
"title": "Write an email from bullet list",
180+
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
181+
},
182+
{
183+
"title": "Code a snake game",
184+
"prompt": "Code a basic snake game in python, give explanations for each step."
185+
},
186+
{
187+
"title": "Assist in a task",
188+
"prompt": "How do I make a delicious lemon cheesecake?"
156189
}
190+
],
191+
"parameters": {
192+
"temperature": 0.7,
193+
"top_p": 0.95,
194+
"repetition_penalty": 1,
195+
"top_k": 50,
196+
"truncate": 24576,
197+
"max_new_tokens": 2048,
198+
"stop": ["<|im_end|>"]
199+
}
200+
},
201+
{
202+
"name": "mistralai/Mistral-7B-Instruct-v0.3",
203+
"displayName": "mistralai/Mistral-7B-Instruct-v0.3",
204+
"description": "A small model with good capabilities in language understanding and commonsense reasoning.",
205+
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/mistral-logo.png",
206+
"websiteUrl": "https://mistral.ai/news/announcing-mistral-7b/",
207+
"modelUrl": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3",
208+
"tokenizer": "mistralai/Mistral-7B-Instruct-v0.3",
209+
"preprompt": "",
210+
"parameters": {
211+
"temperature": 0.3,
212+
"top_p": 0.95,
213+
"repetition_penalty": 1.2,
214+
"top_k": 50,
215+
"truncate": 3072,
216+
"max_new_tokens": 1024,
217+
"stop": ["</s>"]
157218
},
158-
{
159-
"name": "mistralai/Mistral-7B-Instruct-v0.3",
160-
"displayName": "mistralai/Mistral-7B-Instruct-v0.3",
161-
"description": "A small model with good capabilities in language understanding and commonsense reasoning.",
162-
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/mistral-logo.png",
163-
"websiteUrl": "https://mistral.ai/news/announcing-mistral-7b/",
164-
"modelUrl": "https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3",
165-
"tokenizer": "mistralai/Mistral-7B-Instruct-v0.3",
166-
"preprompt": "",
167-
"parameters": {
168-
"temperature": 0.3,
169-
"top_p": 0.95,
170-
"repetition_penalty": 1.2,
171-
"top_k": 50,
172-
"truncate": 3072,
173-
"max_new_tokens": 1024,
174-
"stop": ["</s>"]
219+
"promptExamples": [
220+
{
221+
"title": "Write an email from bullet list",
222+
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
175223
},
176-
"promptExamples": [
177-
{
178-
"title": "Write an email from bullet list",
179-
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
180-
}, {
181-
"title": "Code a snake game",
182-
"prompt": "Code a basic snake game in python, give explanations for each step."
183-
}, {
184-
"title": "Assist in a task",
185-
"prompt": "How do I make a delicious lemon cheesecake?"
186-
}
187-
]
188-
},
189-
{
190-
"name": "microsoft/Phi-3-mini-4k-instruct",
191-
"tokenizer": "microsoft/Phi-3-mini-4k-instruct",
192-
"description" : "One of the best small models (3.8B parameters), super fast for simple tasks.",
193-
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/microsoft-logo.png",
194-
"modelUrl": "https://huggingface.co/microsoft/Phi-3-mini-4k-instruct",
195-
"websiteUrl": "https://azure.microsoft.com/en-us/blog/introducing-phi-3-redefining-whats-possible-with-slms/",
196-
"preprompt": "",
197-
"parameters": {
198-
"stop": ["<|end|>", "<|endoftext|>", "<|assistant|>"],
199-
"temperature": 0.7,
200-
"max_new_tokens": 1024,
201-
"truncate": 3071
224+
{
225+
"title": "Code a snake game",
226+
"prompt": "Code a basic snake game in python, give explanations for each step."
202227
},
203-
"promptExamples": [
204-
{
205-
"title": "Write an email from bullet list",
206-
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
207-
}, {
208-
"title": "Code a snake game",
209-
"prompt": "Code a basic snake game in python, give explanations for each step."
210-
}, {
211-
"title": "Assist in a task",
212-
"prompt": "How do I make a delicious lemon cheesecake?"
213-
}
214-
]
228+
{
229+
"title": "Assist in a task",
230+
"prompt": "How do I make a delicious lemon cheesecake?"
231+
}
232+
]
233+
},
234+
{
235+
"name": "microsoft/Phi-3-mini-4k-instruct",
236+
"tokenizer": "microsoft/Phi-3-mini-4k-instruct",
237+
"description": "One of the best small models (3.8B parameters), super fast for simple tasks.",
238+
"logoUrl": "https://huggingface.co/datasets/huggingchat/models-logo/resolve/main/microsoft-logo.png",
239+
"modelUrl": "https://huggingface.co/microsoft/Phi-3-mini-4k-instruct",
240+
"websiteUrl": "https://azure.microsoft.com/en-us/blog/introducing-phi-3-redefining-whats-possible-with-slms/",
241+
"preprompt": "",
242+
"parameters": {
243+
"stop": ["<|end|>", "<|endoftext|>", "<|assistant|>"],
244+
"temperature": 0.7,
245+
"max_new_tokens": 1024,
246+
"truncate": 3071
215247
},
216-
{
217-
"name" : "llhf/Meta-Llama-3.1-8B-Instruct",
218-
"tokenizer": {"tokenizerUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/resolve/main/tokenizer.json", "tokenizerConfigUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/raw/main/tokenizer_config.json"},
219-
"parameters": {
220-
"temperature": 0.6,
221-
"top_p": 0.9,
222-
"stop": ["<|endoftext|>", "<|eot_id|>"],
248+
"promptExamples": [
249+
{
250+
"title": "Write an email from bullet list",
251+
"prompt": "As a restaurant owner, write a professional email to the supplier to get these products every week: \n\n- Wine (x10)\n- Eggs (x24)\n- Bread (x12)"
223252
},
224-
"unlisted": true
225-
}
253+
{
254+
"title": "Code a snake game",
255+
"prompt": "Code a basic snake game in python, give explanations for each step."
256+
},
257+
{
258+
"title": "Assist in a task",
259+
"prompt": "How do I make a delicious lemon cheesecake?"
260+
}
261+
]
262+
},
263+
{
264+
"name": "llhf/Meta-Llama-3.1-8B-Instruct",
265+
"tokenizer": {
266+
"tokenizerUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/resolve/main/tokenizer.json",
267+
"tokenizerConfigUrl": "https://huggingface.co/nsarrazin/llama3.1-tokenizer/raw/main/tokenizer_config.json"
268+
},
269+
"parameters": {
270+
"temperature": 0.6,
271+
"top_p": 0.9,
272+
"stop": ["<|endoftext|>", "<|eot_id|>"]
273+
},
274+
"unlisted": true
275+
}
226276
]
277+
227278
NODE_ENV: "prod"
228279
NODE_LOG_STRUCTURED_DATA: true
229280
OLD_MODELS: >

0 commit comments

Comments
 (0)