Small update and reorg of text

#198
by thomwolf HF staff - opened
Files changed (1) hide show
  1. src/assets/text_content.py +50 -46
src/assets/text_content.py CHANGED
@@ -59,18 +59,31 @@ CHANGELOG_TEXT = f"""
59
  TITLE = """<h1 align="center" id="space-title">🤗 Open LLM Leaderboard</h1>"""
60
 
61
  INTRODUCTION_TEXT = f"""
62
- 📐 The 🤗 Open LLM Leaderboard aims to track, rank and evaluate LLMs and chatbots as they are released.
63
 
64
- 🤗 Anyone from the community can submit a model for automated evaluation on the 🤗 GPU cluster, as long as it is a 🤗 Transformers model with weights on the Hub. We also support evaluation of models with delta-weights for non-commercial licensed models, such as the original LLaMa release.
65
 
66
- Other cool benchmarks for LLMs are developed at HuggingFace, go check them out: 🙋🤖 [human and GPT4 evals](https://huggingface.co/spaces/HuggingFaceH4/human_eval_llm_leaderboard), 🖥️ [performance benchmarks](https://huggingface.co/spaces/optimum/llm-perf-leaderboard)
 
 
 
 
67
  """
68
 
69
  LLM_BENCHMARKS_TEXT = f"""
70
  # Context
71
  With the plethora of large language models (LLMs) and chatbots being released week upon week, often with grandiose claims of their performance, it can be hard to filter out the genuine progress that is being made by the open-source community and which model is the current state of the art.
72
 
73
- 📈 We evaluate models on 4 key benchmarks from the <a href="https://github.com/EleutherAI/lm-evaluation-harness" target="_blank"> Eleuther AI Language Model Evaluation Harness </a>, a unified framework to test generative language models on a large number of different evaluation tasks.
 
 
 
 
 
 
 
 
 
74
 
75
  - <a href="https://arxiv.org/abs/1803.05457" target="_blank"> AI2 Reasoning Challenge </a> (25-shot) - a set of grade-school science questions.
76
  - <a href="https://arxiv.org/abs/1905.07830" target="_blank"> HellaSwag </a> (10-shot) - a test of commonsense inference, which is easy for humans (~95%) but challenging for SOTA models.
@@ -80,38 +93,13 @@ With the plethora of large language models (LLMs) and chatbots being released we
80
  For all these evaluations, a higher score is a better score.
81
  We chose these benchmarks as they test a variety of reasoning and general knowledge across a wide variety of fields in 0-shot and few-shot settings.
82
 
83
- # Some good practices before submitting a model
84
-
85
- ### 1) Make sure you can load your model and tokenizer using AutoClasses:
86
- ```python
87
- from transformers import AutoConfig, AutoModel, AutoTokenizer
88
- config = AutoConfig.from_pretrained("your model name", revision=revision)
89
- model = AutoModel.from_pretrained("your model name", revision=revision)
90
- tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
91
- ```
92
- If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
93
-
94
- Note: make sure your model is public!
95
- Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!
96
-
97
- ### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
98
- It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
99
-
100
- ### 3) Make sure your model has an open license!
101
- This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
102
-
103
- ### 4) Fill up your model card
104
- When we add extra information about models to the leaderboard, it will be automatically taken from the model card
105
-
106
- # Reproducibility and details
107
-
108
- ### Details and logs
109
  You can find:
110
  - detailed numerical results in the `results` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/results
111
  - details on the input/outputs for the models in the `details` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/details
112
  - community queries and running status in the `requests` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/requests
113
 
114
- ### Reproducibility
115
  To reproduce our results, here is the commands you can run, using [this version](https://github.com/EleutherAI/lm-evaluation-harness/tree/b281b0921b636bc36ad05c0b0b0763bd6dd43463) of the Eleuther AI Harness:
116
  `python main.py --model=hf-causal --model_args="pretrained=<your_model>,use_accelerate=True,revision=<your_model_revision>"`
117
  ` --tasks=<task_list> --num_fewshot=<n_few_shot> --batch_size=2 --output_path=<output_path>`
@@ -125,29 +113,45 @@ The tasks and few shots parameters are:
125
  - TruthfulQA: 0-shot, *truthfulqa-mc* (`mc2`)
126
  - MMLU: 5-shot, *hendrycksTest-abstract_algebra,hendrycksTest-anatomy,hendrycksTest-astronomy,hendrycksTest-business_ethics,hendrycksTest-clinical_knowledge,hendrycksTest-college_biology,hendrycksTest-college_chemistry,hendrycksTest-college_computer_science,hendrycksTest-college_mathematics,hendrycksTest-college_medicine,hendrycksTest-college_physics,hendrycksTest-computer_security,hendrycksTest-conceptual_physics,hendrycksTest-econometrics,hendrycksTest-electrical_engineering,hendrycksTest-elementary_mathematics,hendrycksTest-formal_logic,hendrycksTest-global_facts,hendrycksTest-high_school_biology,hendrycksTest-high_school_chemistry,hendrycksTest-high_school_computer_science,hendrycksTest-high_school_european_history,hendrycksTest-high_school_geography,hendrycksTest-high_school_government_and_politics,hendrycksTest-high_school_macroeconomics,hendrycksTest-high_school_mathematics,hendrycksTest-high_school_microeconomics,hendrycksTest-high_school_physics,hendrycksTest-high_school_psychology,hendrycksTest-high_school_statistics,hendrycksTest-high_school_us_history,hendrycksTest-high_school_world_history,hendrycksTest-human_aging,hendrycksTest-human_sexuality,hendrycksTest-international_law,hendrycksTest-jurisprudence,hendrycksTest-logical_fallacies,hendrycksTest-machine_learning,hendrycksTest-management,hendrycksTest-marketing,hendrycksTest-medical_genetics,hendrycksTest-miscellaneous,hendrycksTest-moral_disputes,hendrycksTest-moral_scenarios,hendrycksTest-nutrition,hendrycksTest-philosophy,hendrycksTest-prehistory,hendrycksTest-professional_accounting,hendrycksTest-professional_law,hendrycksTest-professional_medicine,hendrycksTest-professional_psychology,hendrycksTest-public_relations,hendrycksTest-security_studies,hendrycksTest-sociology,hendrycksTest-us_foreign_policy,hendrycksTest-virology,hendrycksTest-world_religions* (average of all the results `acc`)
127
 
128
- ### Quantization
129
  To get more information about quantization, see:
130
  - 8 bits: [blog post](https://huggingface.co/blog/hf-bitsandbytes-integration), [paper](https://arxiv.org/abs/2208.07339)
131
  - 4 bits: [blog post](https://huggingface.co/blog/4bit-transformers-bitsandbytes), [paper](https://arxiv.org/abs/2305.14314)
132
 
133
- ### Icons
134
- {ModelType.PT.to_str(" : ")} model
135
- {ModelType.FT.to_str(" : ")} model
136
- {ModelType.IFT.to_str(" : ")} model
137
- {ModelType.RL.to_str(" : ")} model
138
- If there is no icon, we have not uploaded the information on the model yet, feel free to open an issue with the model information!
 
 
 
 
 
 
 
 
 
 
 
139
 
 
 
 
 
 
 
 
 
140
 
141
- # In case of model failure
 
 
 
142
  If your model is displayed in the `FAILED` category, its execution stopped.
143
  Make sure you have followed the above steps first.
144
  If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task).
145
-
146
- """
147
-
148
- EVALUATION_QUEUE_TEXT = f"""
149
- # Evaluation Queue for the 🤗 Open LLM Leaderboard
150
- These models will be automatically evaluated on the 🤗 cluster.
151
  """
152
 
153
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
@@ -216,4 +220,4 @@ CITATION_BUTTON_TEXT = r"""
216
  eprint={2109.07958},
217
  archivePrefix={arXiv},
218
  primaryClass={cs.CL}
219
- }"""
 
59
  TITLE = """<h1 align="center" id="space-title">🤗 Open LLM Leaderboard</h1>"""
60
 
61
  INTRODUCTION_TEXT = f"""
62
+ 📐 The 🤗 Open LLM Leaderboard aims to track, rank and evaluate open LLMs and chatbots.
63
 
64
+ 🤗 Submit a model for automated evaluation on the 🤗 GPU cluster on the "Submit" page!
65
 
66
+ The leaderboard's backend runs the great [Eleuther AI Language Model Evaluation Harness](https://github.com/EleutherAI/lm-evaluation-harness) to compute numbers. Read more details and reproducibility on the "About" page!
67
+
68
+ Other cool benchmarks for LLMs are developed at HuggingFace: 🙋🤖 [human and GPT4 evals](https://huggingface.co/spaces/HuggingFaceH4/human_eval_llm_leaderboard), 🖥️ [performance benchmarks](https://huggingface.co/spaces/optimum/llm-perf-leaderboard)
69
+
70
+ And also in other labs, check out the [AlpacaEval Leaderboard](https://tatsu-lab.github.io/alpaca_eval/) and [MT Bench](https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard) among other great ressources.
71
  """
72
 
73
  LLM_BENCHMARKS_TEXT = f"""
74
  # Context
75
  With the plethora of large language models (LLMs) and chatbots being released week upon week, often with grandiose claims of their performance, it can be hard to filter out the genuine progress that is being made by the open-source community and which model is the current state of the art.
76
 
77
+ ## Icons
78
+ {ModelType.PT.to_str(" : ")} model
79
+ {ModelType.FT.to_str(" : ")} model
80
+ {ModelType.IFT.to_str(" : ")} model
81
+ {ModelType.RL.to_str(" : ")} model
82
+ If there is no icon, we have not uploaded the information on the model yet, feel free to open an issue with the model information!
83
+
84
+ ## How it works
85
+
86
+ 📈 We evaluate models on 4 key benchmarks using the <a href="https://github.com/EleutherAI/lm-evaluation-harness" target="_blank"> Eleuther AI Language Model Evaluation Harness </a>, a unified framework to test generative language models on a large number of different evaluation tasks.
87
 
88
  - <a href="https://arxiv.org/abs/1803.05457" target="_blank"> AI2 Reasoning Challenge </a> (25-shot) - a set of grade-school science questions.
89
  - <a href="https://arxiv.org/abs/1905.07830" target="_blank"> HellaSwag </a> (10-shot) - a test of commonsense inference, which is easy for humans (~95%) but challenging for SOTA models.
 
93
  For all these evaluations, a higher score is a better score.
94
  We chose these benchmarks as they test a variety of reasoning and general knowledge across a wide variety of fields in 0-shot and few-shot settings.
95
 
96
+ ## Details and logs
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
97
  You can find:
98
  - detailed numerical results in the `results` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/results
99
  - details on the input/outputs for the models in the `details` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/details
100
  - community queries and running status in the `requests` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/requests
101
 
102
+ ## Reproducibility
103
  To reproduce our results, here is the commands you can run, using [this version](https://github.com/EleutherAI/lm-evaluation-harness/tree/b281b0921b636bc36ad05c0b0b0763bd6dd43463) of the Eleuther AI Harness:
104
  `python main.py --model=hf-causal --model_args="pretrained=<your_model>,use_accelerate=True,revision=<your_model_revision>"`
105
  ` --tasks=<task_list> --num_fewshot=<n_few_shot> --batch_size=2 --output_path=<output_path>`
 
113
  - TruthfulQA: 0-shot, *truthfulqa-mc* (`mc2`)
114
  - MMLU: 5-shot, *hendrycksTest-abstract_algebra,hendrycksTest-anatomy,hendrycksTest-astronomy,hendrycksTest-business_ethics,hendrycksTest-clinical_knowledge,hendrycksTest-college_biology,hendrycksTest-college_chemistry,hendrycksTest-college_computer_science,hendrycksTest-college_mathematics,hendrycksTest-college_medicine,hendrycksTest-college_physics,hendrycksTest-computer_security,hendrycksTest-conceptual_physics,hendrycksTest-econometrics,hendrycksTest-electrical_engineering,hendrycksTest-elementary_mathematics,hendrycksTest-formal_logic,hendrycksTest-global_facts,hendrycksTest-high_school_biology,hendrycksTest-high_school_chemistry,hendrycksTest-high_school_computer_science,hendrycksTest-high_school_european_history,hendrycksTest-high_school_geography,hendrycksTest-high_school_government_and_politics,hendrycksTest-high_school_macroeconomics,hendrycksTest-high_school_mathematics,hendrycksTest-high_school_microeconomics,hendrycksTest-high_school_physics,hendrycksTest-high_school_psychology,hendrycksTest-high_school_statistics,hendrycksTest-high_school_us_history,hendrycksTest-high_school_world_history,hendrycksTest-human_aging,hendrycksTest-human_sexuality,hendrycksTest-international_law,hendrycksTest-jurisprudence,hendrycksTest-logical_fallacies,hendrycksTest-machine_learning,hendrycksTest-management,hendrycksTest-marketing,hendrycksTest-medical_genetics,hendrycksTest-miscellaneous,hendrycksTest-moral_disputes,hendrycksTest-moral_scenarios,hendrycksTest-nutrition,hendrycksTest-philosophy,hendrycksTest-prehistory,hendrycksTest-professional_accounting,hendrycksTest-professional_law,hendrycksTest-professional_medicine,hendrycksTest-professional_psychology,hendrycksTest-public_relations,hendrycksTest-security_studies,hendrycksTest-sociology,hendrycksTest-us_foreign_policy,hendrycksTest-virology,hendrycksTest-world_religions* (average of all the results `acc`)
115
 
116
+ ## Quantization
117
  To get more information about quantization, see:
118
  - 8 bits: [blog post](https://huggingface.co/blog/hf-bitsandbytes-integration), [paper](https://arxiv.org/abs/2208.07339)
119
  - 4 bits: [blog post](https://huggingface.co/blog/4bit-transformers-bitsandbytes), [paper](https://arxiv.org/abs/2305.14314)
120
 
121
+ """
122
+
123
+ EVALUATION_QUEUE_TEXT = f"""
124
+ # Evaluation Queue for the 🤗 Open LLM Leaderboard
125
+
126
+ Models added here will be automatically evaluated on the 🤗 cluster.
127
+
128
+ ## Some good practices before submitting a model
129
+
130
+ ### 1) Make sure you can load your model and tokenizer using AutoClasses:
131
+ ```python
132
+ from transformers import AutoConfig, AutoModel, AutoTokenizer
133
+ config = AutoConfig.from_pretrained("your model name", revision=revision)
134
+ model = AutoModel.from_pretrained("your model name", revision=revision)
135
+ tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
136
+ ```
137
+ If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
138
 
139
+ Note: make sure your model is public!
140
+ Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!
141
+
142
+ ### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
143
+ It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
144
+
145
+ ### 3) Make sure your model has an open license!
146
+ This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
147
 
148
+ ### 4) Fill up your model card
149
+ When we add extra information about models to the leaderboard, it will be automatically taken from the model card
150
+
151
+ ## In case of model failure
152
  If your model is displayed in the `FAILED` category, its execution stopped.
153
  Make sure you have followed the above steps first.
154
  If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task).
 
 
 
 
 
 
155
  """
156
 
157
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
 
220
  eprint={2109.07958},
221
  archivePrefix={arXiv},
222
  primaryClass={cs.CL}
223
+ }"""