Lib.rs

› Data structures
#prompt #template #string-format #chat #model #table #llama-edge

chat-prompts

Chat prompt template

by Xin Liu and 7 contributors

19 releases (5 breaking)

new 0.6.1 Apr 23, 2024
0.5.7 Apr 19, 2024
0.5.3 Mar 25, 2024
0.3.0 Nov 28, 2023

#199 in Data structures

Download history 406/week @ 2024-01-05 3/week @ 2024-01-12 41/week @ 2024-01-19 10/week @ 2024-01-26 8/week @ 2024-02-02 97/week @ 2024-02-16 172/week @ 2024-02-23 172/week @ 2024-03-01 208/week @ 2024-03-08 25/week @ 2024-03-15 147/week @ 2024-03-22 154/week @ 2024-03-29 309/week @ 2024-04-05 93/week @ 2024-04-12 473/week @ 2024-04-19

1,033 downloads per month
Used in llama-core

Apache-2.0

210KB
4K SLoC

Prompt Template Table

You can learn the prompt template name and its associated string format, and models that LlamaEdge supports from the following table.

Prompt type Format Models using this template
llama-2-chat 
<s>[INST] <<SYS>>
{{ system_prompt }}
<</SYS>>

{{ user_msg_1 }} [/INST] {{ model_answer_1 }} </s><s>[INST] {{ user_msg_2 }}   [/INST]
Llama-2-7B-Chat, Llama-2-13B-Chat
chatml 
<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
TinyLlama-1.1B-Chat, i-34B-Chat, OpenHermes-2.5-Mistral-7B, Qwen, Dolphin-2.2-Yi-34B, Dolphin-2.6-Mistral-7B, Samantha-1.2-Mistral-7B, Orca-2-13B, Nous-Hermes-2-Mixtral-8x7B-DPO
openchat 
GPT4 User: {prompt}<|end_of_turn|>GPT4 Assistant:
OpenChat-3.5 series of models
zephyr 
<|system|>
{system_prompt}</s>
<|user|>
{prompt}</s>
<|assistant|>
Zephyr-7B-Alpha
codellama-instruct 
[INST] Write code to solve the following coding problem that obeys the constraints and passes the example test cases. Please wrap your code answer using ```:
{prompt}
[/INST]
Code-llama
mistral-instruct 
<s>[INST] {prompt} [/INST]
Mistral-7B-Instruct-v0.1, Mistral-7B-Instruct-v0.2, Mixtral-8x7B-Instruct-v0.1, miqu-2-70b
stabllm-zephyr 
<|user|>
{prompt}<|endoftext|>
<|assistant|>
stablelm-2-zephyr-1.6b
mistrallite 
<|prompter|>{prompt}</s><|assistant|>
MistralLite-7B
vicuna-1.0-chat 
{system} USER: {prompt} ASSISTANT:
Wizard-Vicuna-13B-Uncensored, Samantha-1.11-CodeLlama-34B, WizardLM-13B-V1.0-Uncensored
vicuna-1.1-chat 
USER: {prompt}
ASSISTANT:
CALM2-7B-Chat, Samantha-1.2-Mistral-7b
wizard-coder 
{system}

### Instruction:
{instruction}

### Response:
WizardCoder-Python-7B-V1.0
deepseek-chat 
User: {user_message_1}

Assistant: {assistant_message_1}<｜end▁of▁sentence｜>User: {user_message_2}

Assistant:
DeepSeek-LLM-7B-Chat
deepseek-coder 
{system}
### Instruction:
{question_1}
### Response:
{answer_1}
<|EOT|>
### Instruction:
{question_2}
### Response:
DeepSeek-Coder-6.7B
solar-instruct 
### User:
{prompt}

### Assistant:
SOLAR-10.7B-Instruct-v1.0
intel-neural 
### System:
{system}
### User:
{usr}
### Assistant:
Intel Nerual series of models
human-assistant 
Human: {input_1}\n\nAssistant:{output_1}Human: {input_2}\n\nAssistant:
Belle-Llama-2-Chat

Dependencies

~16MB
~180K SLoC