From 76f327f771af1b2390976c84dff0f6decef1d56a Mon Sep 17 00:00:00 2001 From: Eyobyb Date: Thu, 10 Aug 2023 14:07:53 +0300 Subject: [PATCH 1/3] change last message getter , remove comment --- apps/slackbot/scrape/prompt_reconstructor.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/apps/slackbot/scrape/prompt_reconstructor.py b/apps/slackbot/scrape/prompt_reconstructor.py index 61007c3c..d0370d10 100644 --- a/apps/slackbot/scrape/prompt_reconstructor.py +++ b/apps/slackbot/scrape/prompt_reconstructor.py @@ -16,6 +16,10 @@ def reconstruct_prompt(self): last_message = self.slack_message last_message_links = get_link_from_slack_client_conversation( last_message) + print('************************************************************************************ ', flush=True) + print(last_message, flush=True) + print(last_message_links, flush=True) + print('************************************************************************************ ', flush=True) # if there is a link inside the question scrape then summerize based # on question and then aggregate to the question From e01aea1482390dd14ed155010e13f444e135ee72 Mon Sep 17 00:00:00 2001 From: Eyobyb Date: Thu, 10 Aug 2023 16:24:59 +0300 Subject: [PATCH 2/3] remove print --- apps/slackbot/scrape/prompt_reconstructor.py | 4 ---- 1 file changed, 4 deletions(-) diff --git a/apps/slackbot/scrape/prompt_reconstructor.py b/apps/slackbot/scrape/prompt_reconstructor.py index d0370d10..61007c3c 100644 --- a/apps/slackbot/scrape/prompt_reconstructor.py +++ b/apps/slackbot/scrape/prompt_reconstructor.py @@ -16,10 +16,6 @@ def reconstruct_prompt(self): last_message = self.slack_message last_message_links = get_link_from_slack_client_conversation( last_message) - print('************************************************************************************ ', flush=True) - print(last_message, flush=True) - print(last_message_links, flush=True) - print('************************************************************************************ ', flush=True) # if there is a link inside the question scrape then summerize based # on question and then aggregate to the question From b122266bbca09725338f304191a26c8142028b25 Mon Sep 17 00:00:00 2001 From: Amir Feizpour Date: Sat, 12 Aug 2023 13:07:24 -0400 Subject: [PATCH 3/3] fixing the sphinx theme and syntax --- ...ge Language Model Applications - Intro.rst | 46 ------------------ .../Self-Improving LLMs.pdf | Bin docs/{LLM Agents => LLM_Agents}/SuhasP.jpg | Bin .../data_generation.rst} | 0 docs/{LLM Agents => LLM_Agents}/gordong.jpeg | Bin .../llm_limitations.rst | 0 .../llm_tools_use.rst | 0 ...e Models and Machine Learning Research.rst | 4 +- ...eflexion - A Reflection-Based Approach.rst | 4 +- ...Models into Template-Based Generators .rst | 4 +- docs/conf.py | 3 +- docs/index.rst | 12 ++--- 12 files changed, 14 insertions(+), 59 deletions(-) delete mode 100644 docs/LLM Foundations/Designing Large Language Model Applications - Intro.rst rename docs/{LLM Agents => LLM_Agents}/Self-Improving LLMs.pdf (100%) rename docs/{LLM Agents => LLM_Agents}/SuhasP.jpg (100%) rename docs/{LLM Agents/data generation.rst => LLM_Agents/data_generation.rst} (100%) rename docs/{LLM Agents => LLM_Agents}/gordong.jpeg (100%) rename docs/{LLM Agents => LLM_Agents}/llm_limitations.rst (100%) rename docs/{LLM Agents => LLM_Agents}/llm_tools_use.rst (100%) diff --git a/docs/LLM Foundations/Designing Large Language Model Applications - Intro.rst b/docs/LLM Foundations/Designing Large Language Model Applications - Intro.rst deleted file mode 100644 index 3152555c..00000000 --- a/docs/LLM Foundations/Designing Large Language Model Applications - Intro.rst +++ /dev/null @@ -1,46 +0,0 @@ - -=========================================== -Designing Large Language Model Applications -=========================================== -*Suhas Pai* - -Summary -------- -Suhas Pai discusses various aspects of large language models and machine learning research, including prompt engineering, model selection, prompting in deployed applications, challenges in deploying GPT-based tasks, estimating token probabilities, determining correctness of answers, token frequency in pre-training data sets, limitations of retrieval models, vocabulary sizes, calibration in practical use cases, and the underlying architecture and training process of large language models. - -Topics: -------- - Prompt Engineering - * Prompt engineering may not be as crucial in the future as models improve and become more aligned with human performance. - * Certain prompt engineering techniques, such as chain of thought prompting, have proven to be impactful and useful. - Model Selection - * Augmented models, which include instruction tuning, reinforcement learning, and multitask training, are more suitable for human use. - * Smaller models may be sufficient for specific use cases. - Prompting in Deployed Applications - * Both hard prompts and soft prompts are used in applications. - * Prompt tuning is still more of an intuitive process rather than a scientific one. - Challenges in Deploying GPT-based Tasks - * Uncertainty estimates, particularly in the finance industry, are important for trust and predictability. - * Model calibration can address the challenge of unreliable outputs, but it can be more challenging for large language models like GPT. - Estimating Token Probabilities and Determining Correctness of Answers - * Heuristics and scoring methods like Blue and ROUGE are not effective in solving these problems. - * Asking for multiple generations and using heuristics or smaller models can improve the accuracy of answers. - * Decoding strategies used in LLMs, especially with external APIs like OpenAI, are limited. - Token Frequency in Pre-training Data Sets - * Tokens with higher frequency in the data set are more likely to produce accurate results when used in prompts. - * The quality of the retrieval model is crucial in determining the accuracy of the language model's responses. - * Using a retrieval or embedding model trained on abstracts could be a more effective approach. - Vocabulary Sizes - * Models like Bloom, which are multilingual, have larger vocabulary sizes to accommodate different languages. - * Additional tokens in the vocabulary could be for special processing, such as code snippets or comments. - Integration between Search Engines and Large Language Models - * Leveraging ideas and optimization techniques from search engines could improve retrieval in language models. - * Optimizing search engines is a challenging task, and finding a relevant metric for ranking is complex. - Calibration in Practical Use Cases - * Calibration is more useful when you have access to your own models and can adjust the output probabilities. - * Calibration becomes more challenging when language models are part of a larger system with external models. - Underlying Architecture and Training Process of Large Language Models - * Large language models are typically built using deep neural networks with multiple layers of interconnected nodes. - * Training involves feeding the model with a large corpus of text data and adjusting the weights of the network. - * The size and diversity of the training data play a crucial role in the performance of large language models. - diff --git a/docs/LLM Agents/Self-Improving LLMs.pdf b/docs/LLM_Agents/Self-Improving LLMs.pdf similarity index 100% rename from docs/LLM Agents/Self-Improving LLMs.pdf rename to docs/LLM_Agents/Self-Improving LLMs.pdf diff --git a/docs/LLM Agents/SuhasP.jpg b/docs/LLM_Agents/SuhasP.jpg similarity index 100% rename from docs/LLM Agents/SuhasP.jpg rename to docs/LLM_Agents/SuhasP.jpg diff --git a/docs/LLM Agents/data generation.rst b/docs/LLM_Agents/data_generation.rst similarity index 100% rename from docs/LLM Agents/data generation.rst rename to docs/LLM_Agents/data_generation.rst diff --git a/docs/LLM Agents/gordong.jpeg b/docs/LLM_Agents/gordong.jpeg similarity index 100% rename from docs/LLM Agents/gordong.jpeg rename to docs/LLM_Agents/gordong.jpeg diff --git a/docs/LLM Agents/llm_limitations.rst b/docs/LLM_Agents/llm_limitations.rst similarity index 100% rename from docs/LLM Agents/llm_limitations.rst rename to docs/LLM_Agents/llm_limitations.rst diff --git a/docs/LLM Agents/llm_tools_use.rst b/docs/LLM_Agents/llm_tools_use.rst similarity index 100% rename from docs/LLM Agents/llm_tools_use.rst rename to docs/LLM_Agents/llm_tools_use.rst diff --git a/docs/Research Papers/Auto GPT - A Breakthrough in Language Models and Machine Learning Research.rst b/docs/Research Papers/Auto GPT - A Breakthrough in Language Models and Machine Learning Research.rst index cf8ed2e1..51a2df2d 100644 --- a/docs/Research Papers/Auto GPT - A Breakthrough in Language Models and Machine Learning Research.rst +++ b/docs/Research Papers/Auto GPT - A Breakthrough in Language Models and Machine Learning Research.rst @@ -1,7 +1,7 @@ -========================================================================= +========================================================================== Auto GPT - A Breakthrough in Language Models and Machine Learning Research -========================================================================= +========================================================================== *Yujing Yang* Summary diff --git a/docs/Research Papers/Reflexion - A Reflection-Based Approach.rst b/docs/Research Papers/Reflexion - A Reflection-Based Approach.rst index a9d0fc52..b4f15e83 100644 --- a/docs/Research Papers/Reflexion - A Reflection-Based Approach.rst +++ b/docs/Research Papers/Reflexion - A Reflection-Based Approach.rst @@ -1,7 +1,7 @@ -====================================== +======================================= Reflexion - A Reflection-Based Approach -====================================== +======================================= *Noah Shinn* Summary diff --git a/docs/Research Papers/TempLM - LM Distilling Language Models into Template-Based Generators .rst b/docs/Research Papers/TempLM - LM Distilling Language Models into Template-Based Generators .rst index 1a7d886f..938e1c29 100644 --- a/docs/Research Papers/TempLM - LM Distilling Language Models into Template-Based Generators .rst +++ b/docs/Research Papers/TempLM - LM Distilling Language Models into Template-Based Generators .rst @@ -1,7 +1,7 @@ -==================================================================== +===================================================================== TempLM - LM Distilling Language Models into Template-Based Generators -==================================================================== +===================================================================== *Kua Chen* Summary diff --git a/docs/conf.py b/docs/conf.py index 33207925..6e911372 100644 --- a/docs/conf.py +++ b/docs/conf.py @@ -47,7 +47,8 @@ # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # -html_theme = 'nature' +html_permalinks_icon = '#' +html_theme = 'sphinxawesome_theme' # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, diff --git a/docs/index.rst b/docs/index.rst index 5d427e8c..04973a6e 100644 --- a/docs/index.rst +++ b/docs/index.rst @@ -26,11 +26,11 @@ CONTENT :maxdepth: 1 :caption: LARGE LANGUAGE MODEL AGENTS - LLM Agents/llm_tools_use.rst - LLM Agents/llm_limitations.rst - LLM Agents/data generation.rst + LLM_Agents/llm_tools_use.rst + LLM_Agents/llm_limitations.rst + LLM_Agents/data_generation.rst - .. toctree:: +.. toctree:: :maxdepth: 1 :caption: LARGE LANGUAGE MODEL COMMERCIALIZATION @@ -45,14 +45,14 @@ CONTENT LLM Engineering/llms_in_enterprise.rst LLM Engineering/Performance Computing on Cloud for Gen AI.rst - .. toctree:: +.. toctree:: :maxdepth: 1 :caption: LARGE LANGUAGE MODEL USE CASES LLM Use Cases/innovation.rst LLM Use Cases/LLM Personal Assistant.rst - .. toctree:: +.. toctree:: :maxdepth: 1 :caption: LARGE LANGUAGE MODEL PROJECTS