From 88e37e1f2d6d3b0c0e0f91bcf615233a2ccdbd2f Mon Sep 17 00:00:00 2001 From: hahuyhoang411 Date: Mon, 4 Dec 2023 12:34:10 +0700 Subject: [PATCH] update ctxlen --- models/neural-chat-7b/model.json | 4 ++-- models/neuralhermes-7b/model.json | 4 ++-- models/noromaid-20b/model.json | 4 ++-- models/openorca-13b/model.json | 4 ++-- models/phind-34b/model.json | 4 ++-- models/rocket-3b/model.json | 2 +- models/starling-7b/model.json | 4 ++-- models/tiefighter-13b/model.json | 4 ++-- models/wizardcoder-13b/model.json | 4 ++-- models/wizardcoder-34b/model.json | 4 ++-- models/yi-34b/model.json | 4 ++-- models/zephyr-beta-7b/model.json | 4 ++-- 12 files changed, 23 insertions(+), 23 deletions(-) diff --git a/models/neural-chat-7b/model.json b/models/neural-chat-7b/model.json index a31662496..84c8d029f 100644 --- a/models/neural-chat-7b/model.json +++ b/models/neural-chat-7b/model.json @@ -7,13 +7,13 @@ "description": "The Neural Chat 7B model, developed on the foundation of mistralai/Mistral-7B-v0.1, has been fine-tuned using the Open-Orca/SlimOrca dataset and aligned with the Direct Preference Optimization (DPO) algorithm. It has demonstrated substantial improvements in various AI tasks and performance well on the open_llm_leaderboard.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "### System: ", "user_prompt": "### User: ", "ai_prompt": "### Assistant: " }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "Intel, The Bloke", diff --git a/models/neuralhermes-7b/model.json b/models/neuralhermes-7b/model.json index 3ec0c6253..0cc83d7fb 100644 --- a/models/neuralhermes-7b/model.json +++ b/models/neuralhermes-7b/model.json @@ -7,13 +7,13 @@ "description": "NeuralHermes 2.5 has been enhanced using Direct Preference Optimization. This fine-tuning, inspired by the RLHF process of Neural-chat-7b and OpenHermes-2.5-Mistral-7B, has led to improved performance across several benchmarks.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "<|im_start|>system\n", "user_prompt": "<|im_end|>\n<|im_start|>user\n", "ai_prompt": "<|im_end|>\n<|im_start|>assistant\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "Intel, The Bloke", diff --git a/models/noromaid-20b/model.json b/models/noromaid-20b/model.json index f0460568c..ff2b91c1d 100644 --- a/models/noromaid-20b/model.json +++ b/models/noromaid-20b/model.json @@ -7,13 +7,13 @@ "description": "The Noromaid 20b model is designed for role-playing and general use, featuring a unique touch with the no_robots dataset that enhances human-like behavior.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "", "user_prompt": "### Instruction: ", "ai_prompt": "### Response: " }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "NeverSleep, The Bloke", diff --git a/models/openorca-13b/model.json b/models/openorca-13b/model.json index 5027dbd22..13b6abf29 100644 --- a/models/openorca-13b/model.json +++ b/models/openorca-13b/model.json @@ -7,13 +7,13 @@ "description": "Orca 2 is a finetuned version of LLAMA-2, designed primarily for single-turn responses in reasoning, reading comprehension, math problem solving, and text summarization.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "<|im_start|>system\n", "user_prompt": "<|im_end|>\n<|im_start|>user\n", "ai_prompt": "<|im_end|>\n<|im_start|>assistant\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "Microsoft, The Bloke", diff --git a/models/phind-34b/model.json b/models/phind-34b/model.json index 18ed50bcc..761c51346 100644 --- a/models/phind-34b/model.json +++ b/models/phind-34b/model.json @@ -7,13 +7,13 @@ "description": "Phind-CodeLlama-34B-v2 is an AI model fine-tuned on 1.5B tokens of high-quality programming data. It's a SOTA open-source model in coding. This multi-lingual model excels in various programming languages, including Python, C/C++, TypeScript, Java, and is designed to be steerable and user-friendly.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "### System Prompt\n", "user_prompt": "### User Message\n", "ai_prompt": "### Assistant\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "Phind, The Bloke", diff --git a/models/rocket-3b/model.json b/models/rocket-3b/model.json index 4eddb04b0..a70db1dbd 100644 --- a/models/rocket-3b/model.json +++ b/models/rocket-3b/model.json @@ -13,7 +13,7 @@ "ai_prompt": "<|im_end|>\n<|im_start|>assistant\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "pansophic, The Bloke", diff --git a/models/starling-7b/model.json b/models/starling-7b/model.json index 4ff5651e0..cf8aeda18 100644 --- a/models/starling-7b/model.json +++ b/models/starling-7b/model.json @@ -7,13 +7,13 @@ "description": "Starling-RM-7B-alpha is a language model finetuned with Reinforcement Learning from AI Feedback from Openchat 3.5. It stands out for its impressive performance using GPT-4 as a judge, making it one of the top-performing models in its category.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "", "user_prompt": "GPT4 User: ", "ai_prompt": "<|end_of_turn|>\nGPT4 Assistant: " }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "Berkeley-nest, The Bloke", diff --git a/models/tiefighter-13b/model.json b/models/tiefighter-13b/model.json index ea38fadbf..4dd9bb37f 100644 --- a/models/tiefighter-13b/model.json +++ b/models/tiefighter-13b/model.json @@ -7,13 +7,13 @@ "description": "Tiefighter-13B is a highly creative, merged AI model achieved by combining various 'LORAs' on top of an existing merge, particularly focusing on storytelling and improvisation. This model excels in story writing, chatbots, and adventuring, and is designed to perform better with less detailed inputs, leveraging its inherent creativity.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "", "user_prompt": "### Instruction: ", "ai_prompt": "\n### Response: " }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "KoboldAI, The Bloke", diff --git a/models/wizardcoder-13b/model.json b/models/wizardcoder-13b/model.json index 9fe250348..47899534a 100644 --- a/models/wizardcoder-13b/model.json +++ b/models/wizardcoder-13b/model.json @@ -7,13 +7,13 @@ "description": "WizardCoder-Python-13B is a Python coding model major models like ChatGPT-3.5. This model based on the Llama2 architecture, demonstrate high proficiency in specific domains like coding and mathematics.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "", "user_prompt": "### Instruction:\n", "ai_prompt": "### Response:\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "WizardLM, The Bloke", diff --git a/models/wizardcoder-34b/model.json b/models/wizardcoder-34b/model.json index db92a4fbf..f564afc35 100644 --- a/models/wizardcoder-34b/model.json +++ b/models/wizardcoder-34b/model.json @@ -7,13 +7,13 @@ "description": "WizardCoder-Python-34B is a Python coding model major models like ChatGPT-3.5. This model based on the Llama2 architecture, demonstrate high proficiency in specific domains like coding and mathematics.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "", "user_prompt": "### Instruction:\n", "ai_prompt": "### Response:\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "WizardLM, The Bloke", diff --git a/models/yi-34b/model.json b/models/yi-34b/model.json index 9bbe65760..d34c0efed 100644 --- a/models/yi-34b/model.json +++ b/models/yi-34b/model.json @@ -7,13 +7,13 @@ "description": "Yi-34B, a specialized chat model, is known for its diverse and creative responses and excels across various NLP tasks and benchmarks.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "<|im_start|>system\n", "user_prompt": "<|im_end|>\n<|im_start|>user\n", "ai_prompt": "<|im_end|>\n<|im_start|>assistant\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "01-ai, The Bloke", diff --git a/models/zephyr-beta-7b/model.json b/models/zephyr-beta-7b/model.json index 474dba766..70c7c72e8 100644 --- a/models/zephyr-beta-7b/model.json +++ b/models/zephyr-beta-7b/model.json @@ -7,13 +7,13 @@ "description": "The Zephyr-7B-β model marks the second iteration in the Zephyr series, designed to function as an effective assistant. It has been fine-tuned from the mistralai/Mistral-7B-v0.1 base model, utilizing a combination of public and synthetic datasets with the application of Direct Preference Optimization.", "format": "gguf", "settings": { - "ctx_len": 4096, + "ctx_len": 2048, "system_prompt": "<|system|>\n", "user_prompt": "\n<|user|>\n", "ai_prompt": "\n<|assistant|>\n" }, "parameters": { - "max_tokens": 4096 + "max_tokens": 2048 }, "metadata": { "author": "HuggingFaceH4, The Bloke",