From f728dbddd0737f42173eba65aa0b423335218bd5 Mon Sep 17 00:00:00 2001 From: HanishKVC Date: Sun, 28 Apr 2024 13:07:43 +0530 Subject: [PATCH] ChatON: Add simpcfg based config file matching chaton_meta.json Add missing begin and end fields wrt deepseek-coder assistant in chaton_meta.json. Idea is to avoid json library dependency by adding a simple text based config file support. --- examples/chaton_meta.json | 4 +- examples/chaton_meta.simpcfg | 225 +++++++++++++++++++++++++++++++++++ 2 files changed, 228 insertions(+), 1 deletion(-) create mode 100644 examples/chaton_meta.simpcfg diff --git a/examples/chaton_meta.json b/examples/chaton_meta.json index 5e3c6f8e8..994301386 100644 --- a/examples/chaton_meta.json +++ b/examples/chaton_meta.json @@ -165,8 +165,10 @@ "end": "" }, "assistant": { + "begin": "", "prefix": "### Response:\n", - "suffix": "\n<|EOT|>\n" + "suffix": "\n<|EOT|>\n", + "end": "" }, "reverse-prompt": "<|EOT|>", "systemuser-system-has-suffix": true, diff --git a/examples/chaton_meta.simpcfg b/examples/chaton_meta.simpcfg new file mode 100644 index 000000000..2a209fd21 --- /dev/null +++ b/examples/chaton_meta.simpcfg @@ -0,0 +1,225 @@ +"llama2" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "[INST] <>\n" + "system-suffix": "\n<> " + "system-end": "[/INST]\n\n" + + "user-begin": "" + "user-prefix": "[INST] " + "user-suffix": " [/INST]\n\n" + "user-end": "" + + "assistant-begin": "" + "assistant-prefix": "" + "assistant-suffix": "" + "assistant-end": "" + + "reverse-prompt": "", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": false, + "systemuser-1st-user-has-begin": false, + "systemuser-1st-user-has-prefix": false + +"llama3" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "<|start_header_id|>system<|end_header_id|>\n" + "system-suffix": "<|eot_id|>\n\n" + "system-end": "" + + "user-begin": "" + "user-prefix": "<|start_header_id|>user<|end_header_id|>\n" + "user-suffix": "<|eot_id|>\n\n" + "user-end": "" + + "assistant-begin": "" + "assistant-prefix": "<|start_header_id|>assistant<|end_header_id|>\n" + "assistant-suffix": "<|eot_id|>\n\n" + "assistant-end": "" + + "reverse-prompt": "<|eot_id|>", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": true, + "systemuser-1st-user-has-begin": true, + "systemuser-1st-user-has-prefix": true + +"chatml" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "<|im_start|>system\n" + "system-suffix": "<|im_end|>\n" + "system-end": "" + + "user-begin": "" + "user-prefix": "<|im_start|>user\n" + "user-suffix": "<|im_end|>\n" + "user-end": "" + + "assistant-begin": "" + "assistant-prefix": "<|im_start|>assistant\n" + "assistant-suffix": "<|im_end|>\n" + "assistant-end": "" + + "reverse-prompt": "<|im_start|>user\n", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": true, + "systemuser-1st-user-has-begin": true, + "systemuser-1st-user-has-prefix": true + +"zephyr" + "global-alt-end": "<|endoftext|>\n" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "<|system|>\n" + "system-suffix": "" + "system-end": "\n" + + "user-begin": "" + "user-prefix": "<|user|>\n" + "user-suffix": "" + "user-end": "\n" + + "assistant-end-alt": "<|endoftext|>\n" + "assistant-begin": "" + "assistant-prefix": "<|assistant|>\n" + "assistant-suffix": "" + "assistant-end": "\n" + + "reverse-prompt": "", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": true, + "systemuser-1st-user-has-begin": true, + "systemuser-1st-user-has-prefix": true + +"gemma" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "user\n" + "system-suffix": "\n" + "system-end": "" + + "user-begin": "" + "user-prefix": "user\n" + "user-suffix": "\n" + "user-end": "" + + "assistant-begin": "" + "assistant-prefix": "model\n" + "assistant-suffix": "\n" + "assistant-end": "" + + "reverse-prompt": "", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": false, + "systemuser-1st-user-has-begin": true, + "systemuser-1st-user-has-prefix": true + +"deepseek-coder" + "global-begin": "" + "global-end": "" + + "system-begin": "\n<|begin▁of▁sentence|>" + "system-prefix": "" + "system-suffix": "\n" + "system-end": "" + + "user-begin": "" + "user-prefix": "### Instruction:\n" + "user-suffix": "\n" + "user-end": "" + + "assistant-prefix": "### Response:\n" + "assistant-suffix": "\n<|EOT|>\n + + "reverse-prompt": "<|EOT|>" + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": false, + "systemuser-1st-user-has-begin": false, + "systemuser-1st-user-has-prefix": true + +"deepseek" + "global-begin": "" + "global-end": "" + + "system-begin": "<|begin▁of▁sentence|>" + "system-prefix": "" + "system-suffix": "\n\n" + "system-end": "" + + "user-begin": "<|begin▁of▁sentence|>" + "user-prefix": "User: " + "user-suffix": "\n\n" + "user-end": "" + + "assistant-begin": "" + "assistant-prefix": "Assistant: " + "assistant-suffix": " <|end▁of▁sentence|>\n" + "assistant-end": "" + + "reverse-prompt": "<|end▁of▁sentence|>", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": false, + "systemuser-1st-user-has-begin": false, + "systemuser-1st-user-has-prefix": true + +"monarch" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "system\n" + "system-suffix": "" + "system-end": "\n" + + "user-begin": "" + "user-prefix": "user\n" + "user-suffix": "" + "user-end": "\n" + + "assistant-begin": "" + "assistant-prefix": "assistant\n" + "assistant-suffix": "" + "assistant-end": "\n" + + "reverse-prompt": "", + "systemuser-system-has-suffix": true, + "systemuser-system-has-end": true, + "systemuser-1st-user-has-begin": true, + "systemuser-1st-user-has-prefix": true + +"mistral" + "global-begin": "" + "global-end": "" + + "system-begin": "" + "system-prefix": "[INST] " + "system-suffix": " [/INST]" + "system-end": "\n" + + "user-begin": "" + "user-prefix": "[INST] " + "user-suffix": " [/INST]" + "user-end": "\n" + + "assistant-begin": "" + "assistant-prefix": "" + "assistant-suffix": "" + "assistant-end": " \n" + + "reverse-prompt": "", + "systemuser-system-has-suffix": false, + "systemuser-system-has-end": true, + "systemuser-1st-user-has-begin": false, + "systemuser-1st-user-has-prefix": false +