Spaces:
Runtime error
Runtime error
Sai Vinay G
commited on
Commit
•
39b62ef
1
Parent(s):
bf07f8e
updates
Browse files- .gitignore +2 -0
- app.py +14 -3
- requirements.txt +0 -0
- src/assets/css_html_js.py +15 -1
- src/assets/hardcoded_evals.py +1 -0
- src/assets/text_content.py +41 -5
- src/auto_leaderboard/get_model_metadata.py +1 -1
- src/auto_leaderboard/load_results.py +14 -20
- src/auto_leaderboard/model_metadata_type.py +555 -496
- src/init.py +0 -0
- src/utils_display.py +2 -2
.gitignore
CHANGED
@@ -6,6 +6,8 @@ __pycache__/
|
|
6 |
*ipynb
|
7 |
.vscode/
|
8 |
|
|
|
|
|
9 |
gpt_4_evals/
|
10 |
human_evals/
|
11 |
eval-queue/
|
|
|
6 |
*ipynb
|
7 |
.vscode/
|
8 |
|
9 |
+
.tmp/
|
10 |
+
|
11 |
gpt_4_evals/
|
12 |
human_evals/
|
13 |
eval-queue/
|
app.py
CHANGED
@@ -18,6 +18,8 @@ from src.assets.css_html_js import custom_css, get_window_url_params
|
|
18 |
from src.utils_display import AutoEvalColumn, EvalQueueColumn, fields, styled_error, styled_warning, styled_message
|
19 |
from src.init import get_all_requested_models, load_all_info_from_hub
|
20 |
|
|
|
|
|
21 |
# clone / pull the lmeh eval data
|
22 |
H4_TOKEN = os.environ.get("H4_TOKEN", None)
|
23 |
|
@@ -91,7 +93,7 @@ def get_leaderboard_df():
|
|
91 |
|
92 |
df = pd.DataFrame.from_records(all_data)
|
93 |
df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
|
94 |
-
df = df[COLS]
|
95 |
|
96 |
# filter out if any of the benchmarks have not been produced
|
97 |
df = df[has_no_nan_values(df, BENCHMARK_COLS)]
|
@@ -183,6 +185,9 @@ def add_new_eval(
|
|
183 |
precision = precision.split(" ")[0]
|
184 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
185 |
|
|
|
|
|
|
|
186 |
# check the model actually exists before adding the eval
|
187 |
if revision == "":
|
188 |
revision = "main"
|
@@ -325,7 +330,13 @@ with demo:
|
|
325 |
)
|
326 |
filter_columns = gr.Radio(
|
327 |
label="⏚ Filter model types",
|
328 |
-
choices = [
|
|
|
|
|
|
|
|
|
|
|
|
|
329 |
value="all",
|
330 |
elem_id="filter-columns"
|
331 |
)
|
@@ -423,4 +434,4 @@ with demo:
|
|
423 |
scheduler = BackgroundScheduler()
|
424 |
scheduler.add_job(restart_space, "interval", seconds=3600)
|
425 |
scheduler.start()
|
426 |
-
demo.queue(concurrency_count=40).launch()
|
|
|
18 |
from src.utils_display import AutoEvalColumn, EvalQueueColumn, fields, styled_error, styled_warning, styled_message
|
19 |
from src.init import get_all_requested_models, load_all_info_from_hub
|
20 |
|
21 |
+
pd.set_option('display.precision', 1)
|
22 |
+
|
23 |
# clone / pull the lmeh eval data
|
24 |
H4_TOKEN = os.environ.get("H4_TOKEN", None)
|
25 |
|
|
|
93 |
|
94 |
df = pd.DataFrame.from_records(all_data)
|
95 |
df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
|
96 |
+
df = df[COLS].round(decimals=2)
|
97 |
|
98 |
# filter out if any of the benchmarks have not been produced
|
99 |
df = df[has_no_nan_values(df, BENCHMARK_COLS)]
|
|
|
185 |
precision = precision.split(" ")[0]
|
186 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
187 |
|
188 |
+
if model_type is None or model_type == "":
|
189 |
+
return styled_error("Please select a model type.")
|
190 |
+
|
191 |
# check the model actually exists before adding the eval
|
192 |
if revision == "":
|
193 |
revision = "main"
|
|
|
330 |
)
|
331 |
filter_columns = gr.Radio(
|
332 |
label="⏚ Filter model types",
|
333 |
+
choices = [
|
334 |
+
"all",
|
335 |
+
ModelType.PT.to_str(),
|
336 |
+
ModelType.FT.to_str(),
|
337 |
+
ModelType.IFT.to_str(),
|
338 |
+
ModelType.RL.to_str(),
|
339 |
+
],
|
340 |
value="all",
|
341 |
elem_id="filter-columns"
|
342 |
)
|
|
|
434 |
scheduler = BackgroundScheduler()
|
435 |
scheduler.add_job(restart_space, "interval", seconds=3600)
|
436 |
scheduler.start()
|
437 |
+
demo.queue(concurrency_count=40).launch()
|
requirements.txt
CHANGED
File without changes
|
src/assets/css_html_js.py
CHANGED
@@ -2,31 +2,40 @@ custom_css = """
|
|
2 |
#changelog-text {
|
3 |
font-size: 16px !important;
|
4 |
}
|
|
|
5 |
#changelog-text h2 {
|
6 |
font-size: 18px !important;
|
7 |
}
|
|
|
8 |
.markdown-text {
|
9 |
font-size: 16px !important;
|
10 |
}
|
|
|
11 |
#models-to-add-text {
|
12 |
font-size: 18px !important;
|
13 |
}
|
|
|
14 |
#citation-button span {
|
15 |
font-size: 16px !important;
|
16 |
}
|
|
|
17 |
#citation-button textarea {
|
18 |
font-size: 16px !important;
|
19 |
}
|
|
|
20 |
#citation-button > label > button {
|
21 |
margin: 6px;
|
22 |
transform: scale(1.3);
|
23 |
}
|
|
|
24 |
#leaderboard-table {
|
25 |
margin-top: 15px
|
26 |
}
|
|
|
27 |
#leaderboard-table-lite {
|
28 |
margin-top: 15px
|
29 |
}
|
|
|
30 |
#search-bar-table-box > div:first-child {
|
31 |
background: none;
|
32 |
border: none;
|
@@ -35,11 +44,13 @@ custom_css = """
|
|
35 |
#search-bar {
|
36 |
padding: 0px;
|
37 |
}
|
|
|
38 |
/* Hides the final AutoEvalColumn */
|
39 |
#llm-benchmark-tab-table table td:last-child,
|
40 |
#llm-benchmark-tab-table table th:last-child {
|
41 |
display: none;
|
42 |
}
|
|
|
43 |
/* Limit the width of the first AutoEvalColumn so that names don't expand too much */
|
44 |
table td:first-child,
|
45 |
table th:first-child {
|
@@ -47,9 +58,11 @@ table th:first-child {
|
|
47 |
overflow: auto;
|
48 |
white-space: nowrap;
|
49 |
}
|
|
|
50 |
.tab-buttons button {
|
51 |
font-size: 20px;
|
52 |
}
|
|
|
53 |
#scale-logo {
|
54 |
border-style: none !important;
|
55 |
box-shadow: none;
|
@@ -58,6 +71,7 @@ table th:first-child {
|
|
58 |
margin-right: auto;
|
59 |
max-width: 600px;
|
60 |
}
|
|
|
61 |
#scale-logo .download {
|
62 |
display: none;
|
63 |
}
|
@@ -69,4 +83,4 @@ get_window_url_params = """
|
|
69 |
url_params = Object.fromEntries(params);
|
70 |
return url_params;
|
71 |
}
|
72 |
-
"""
|
|
|
2 |
#changelog-text {
|
3 |
font-size: 16px !important;
|
4 |
}
|
5 |
+
|
6 |
#changelog-text h2 {
|
7 |
font-size: 18px !important;
|
8 |
}
|
9 |
+
|
10 |
.markdown-text {
|
11 |
font-size: 16px !important;
|
12 |
}
|
13 |
+
|
14 |
#models-to-add-text {
|
15 |
font-size: 18px !important;
|
16 |
}
|
17 |
+
|
18 |
#citation-button span {
|
19 |
font-size: 16px !important;
|
20 |
}
|
21 |
+
|
22 |
#citation-button textarea {
|
23 |
font-size: 16px !important;
|
24 |
}
|
25 |
+
|
26 |
#citation-button > label > button {
|
27 |
margin: 6px;
|
28 |
transform: scale(1.3);
|
29 |
}
|
30 |
+
|
31 |
#leaderboard-table {
|
32 |
margin-top: 15px
|
33 |
}
|
34 |
+
|
35 |
#leaderboard-table-lite {
|
36 |
margin-top: 15px
|
37 |
}
|
38 |
+
|
39 |
#search-bar-table-box > div:first-child {
|
40 |
background: none;
|
41 |
border: none;
|
|
|
44 |
#search-bar {
|
45 |
padding: 0px;
|
46 |
}
|
47 |
+
|
48 |
/* Hides the final AutoEvalColumn */
|
49 |
#llm-benchmark-tab-table table td:last-child,
|
50 |
#llm-benchmark-tab-table table th:last-child {
|
51 |
display: none;
|
52 |
}
|
53 |
+
|
54 |
/* Limit the width of the first AutoEvalColumn so that names don't expand too much */
|
55 |
table td:first-child,
|
56 |
table th:first-child {
|
|
|
58 |
overflow: auto;
|
59 |
white-space: nowrap;
|
60 |
}
|
61 |
+
|
62 |
.tab-buttons button {
|
63 |
font-size: 20px;
|
64 |
}
|
65 |
+
|
66 |
#scale-logo {
|
67 |
border-style: none !important;
|
68 |
box-shadow: none;
|
|
|
71 |
margin-right: auto;
|
72 |
max-width: 600px;
|
73 |
}
|
74 |
+
|
75 |
#scale-logo .download {
|
76 |
display: none;
|
77 |
}
|
|
|
83 |
url_params = Object.fromEntries(params);
|
84 |
return url_params;
|
85 |
}
|
86 |
+
"""
|
src/assets/hardcoded_evals.py
CHANGED
@@ -38,3 +38,4 @@ baseline = {
|
|
38 |
AutoEvalColumn.dummy.name: "baseline",
|
39 |
AutoEvalColumn.model_type.name: "",
|
40 |
}
|
|
|
|
38 |
AutoEvalColumn.dummy.name: "baseline",
|
39 |
AutoEvalColumn.model_type.name: "",
|
40 |
}
|
41 |
+
|
src/assets/text_content.py
CHANGED
@@ -1,42 +1,57 @@
|
|
|
|
|
|
1 |
CHANGELOG_TEXT = f"""
|
2 |
## [2023-06-19]
|
3 |
- Added model type column
|
4 |
- Hid revision and 8bit columns since all models are the same atm
|
|
|
5 |
## [2023-06-16]
|
6 |
- Refactored code base
|
7 |
- Added new columns: number of parameters, hub likes, license
|
|
|
8 |
## [2023-06-13]
|
9 |
- Adjust description for TruthfulQA
|
|
|
10 |
## [2023-06-12]
|
11 |
- Add Human & GPT-4 Evaluations
|
|
|
12 |
## [2023-06-05]
|
13 |
- Increase concurrent thread count to 40
|
14 |
- Search models on ENTER
|
|
|
15 |
## [2023-06-02]
|
16 |
- Add a typeahead search bar
|
17 |
- Use webhooks to automatically spawn a new Space when someone opens a PR
|
18 |
- Start recording `submitted_time` for eval requests
|
19 |
- Limit AutoEvalColumn max-width
|
|
|
20 |
## [2023-05-30]
|
21 |
- Add a citation button
|
22 |
- Simplify Gradio layout
|
|
|
23 |
## [2023-05-29]
|
24 |
- Auto-restart every hour for the latest results
|
25 |
- Sync with the internal version (minor style changes)
|
|
|
26 |
## [2023-05-24]
|
27 |
- Add a baseline that has 25.0 for all values
|
28 |
- Add CHANGELOG
|
|
|
29 |
## [2023-05-23]
|
30 |
- Fix a CSS issue that made the leaderboard hard to read in dark mode
|
|
|
31 |
## [2023-05-22]
|
32 |
- Display a success/error message after submitting evaluation requests
|
33 |
- Reject duplicate submission
|
34 |
- Do not display results that have incomplete results
|
35 |
- Display different queues for jobs that are RUNNING, PENDING, FINISHED status
|
|
|
36 |
## [2023-05-15]
|
37 |
- Fix a typo: from "TruthQA" to "QA"
|
|
|
38 |
## [2023-05-10]
|
39 |
- Fix a bug that prevented auto-refresh
|
|
|
40 |
## [2023-05-10]
|
41 |
- Release the leaderboard to public
|
42 |
"""
|
@@ -45,22 +60,28 @@ TITLE = """<h1 align="center" id="space-title">🤗 Open LLM Leaderboard</h1>"""
|
|
45 |
|
46 |
INTRODUCTION_TEXT = f"""
|
47 |
📐 The 🤗 Open LLM Leaderboard aims to track, rank and evaluate LLMs and chatbots as they are released.
|
|
|
48 |
🤗 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.
|
|
|
49 |
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)
|
50 |
-
🟢: Base pretrained model – 🔶: Instruction finetuned model – 🟦: Model finetuned with RL (read more details in "About" tab)
|
51 |
"""
|
52 |
|
53 |
LLM_BENCHMARKS_TEXT = f"""
|
54 |
# Context
|
55 |
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.
|
|
|
56 |
📈 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.
|
|
|
57 |
- <a href="https://arxiv.org/abs/1803.05457" target="_blank"> AI2 Reasoning Challenge </a> (25-shot) - a set of grade-school science questions.
|
58 |
- <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.
|
59 |
- <a href="https://arxiv.org/abs/2009.03300" target="_blank"> MMLU </a> (5-shot) - a test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more.
|
60 |
- <a href="https://arxiv.org/abs/2109.07958" target="_blank"> TruthfulQA </a> (0-shot) - a test to measure a model’s propensity to reproduce falsehoods commonly found online. Note: TruthfulQA in the Harness is actually a minima a 6-shots task, as it is prepended by 6 examples systematically, even when launched using 0 for the number of few-shot examples.
|
|
|
61 |
For all these evaluations, a higher score is a better score.
|
62 |
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.
|
|
|
63 |
# Some good practices before submitting a model
|
|
|
64 |
### 1) Make sure you can load your model and tokenizer using AutoClasses:
|
65 |
```python
|
66 |
from transformers import AutoConfig, AutoModel, AutoTokenizer
|
@@ -69,44 +90,59 @@ model = AutoModel.from_pretrained("your model name", revision=revision)
|
|
69 |
tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
|
70 |
```
|
71 |
If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
|
|
|
72 |
Note: make sure your model is public!
|
73 |
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!
|
|
|
74 |
### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
|
75 |
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`!
|
|
|
76 |
### 3) Make sure your model has an open license!
|
77 |
This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
|
|
|
78 |
### 4) Fill up your model card
|
79 |
When we add extra information about models to the leaderboard, it will be automatically taken from the model card
|
|
|
80 |
# Reproducibility and details
|
|
|
81 |
### Details and logs
|
82 |
You can find:
|
83 |
- detailed numerical results in the `results` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/results
|
84 |
- details on the input/outputs for the models in the `details` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/details
|
85 |
- community queries and running status in the `requests` Hugging Face dataset: https://huggingface.co/datasets/open-llm-leaderboard/requests
|
|
|
86 |
### Reproducibility
|
87 |
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:
|
88 |
`python main.py --model=hf-causal --model_args="pretrained=<your_model>,use_accelerate=True,revision=<your_model_revision>"`
|
89 |
` --tasks=<task_list> --num_fewshot=<n_few_shot> --batch_size=2 --output_path=<output_path>`
|
|
|
90 |
The total batch size we get for models which fit on one A100 node is 16 (8 GPUs * 2). If you don't use parallelism, adapt your batch size to fit.
|
91 |
*You can expect results to vary slightly for different batch sizes because of padding.*
|
|
|
92 |
The tasks and few shots parameters are:
|
93 |
- ARC: 25-shot, *arc-challenge* (`acc_norm`)
|
94 |
- HellaSwag: 10-shot, *hellaswag* (`acc_norm`)
|
95 |
- TruthfulQA: 0-shot, *truthfulqa-mc* (`mc2`)
|
96 |
-
- 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* (
|
|
|
97 |
### Quantization
|
98 |
To get more information about quantization, see:
|
99 |
- 8 bits: [blog post](https://huggingface.co/blog/hf-bitsandbytes-integration), [paper](https://arxiv.org/abs/2208.07339)
|
100 |
- 4 bits: [blog post](https://huggingface.co/blog/4bit-transformers-bitsandbytes), [paper](https://arxiv.org/abs/2305.14314)
|
|
|
101 |
### Icons
|
102 |
-
|
103 |
-
|
104 |
-
|
|
|
105 |
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!
|
|
|
|
|
106 |
# In case of model failure
|
107 |
If your model is displayed in the `FAILED` category, its execution stopped.
|
108 |
Make sure you have followed the above steps first.
|
109 |
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).
|
|
|
110 |
"""
|
111 |
|
112 |
EVALUATION_QUEUE_TEXT = f"""
|
|
|
1 |
+
from ..auto_leaderboard.model_metadata_type import ModelType
|
2 |
+
|
3 |
CHANGELOG_TEXT = f"""
|
4 |
## [2023-06-19]
|
5 |
- Added model type column
|
6 |
- Hid revision and 8bit columns since all models are the same atm
|
7 |
+
|
8 |
## [2023-06-16]
|
9 |
- Refactored code base
|
10 |
- Added new columns: number of parameters, hub likes, license
|
11 |
+
|
12 |
## [2023-06-13]
|
13 |
- Adjust description for TruthfulQA
|
14 |
+
|
15 |
## [2023-06-12]
|
16 |
- Add Human & GPT-4 Evaluations
|
17 |
+
|
18 |
## [2023-06-05]
|
19 |
- Increase concurrent thread count to 40
|
20 |
- Search models on ENTER
|
21 |
+
|
22 |
## [2023-06-02]
|
23 |
- Add a typeahead search bar
|
24 |
- Use webhooks to automatically spawn a new Space when someone opens a PR
|
25 |
- Start recording `submitted_time` for eval requests
|
26 |
- Limit AutoEvalColumn max-width
|
27 |
+
|
28 |
## [2023-05-30]
|
29 |
- Add a citation button
|
30 |
- Simplify Gradio layout
|
31 |
+
|
32 |
## [2023-05-29]
|
33 |
- Auto-restart every hour for the latest results
|
34 |
- Sync with the internal version (minor style changes)
|
35 |
+
|
36 |
## [2023-05-24]
|
37 |
- Add a baseline that has 25.0 for all values
|
38 |
- Add CHANGELOG
|
39 |
+
|
40 |
## [2023-05-23]
|
41 |
- Fix a CSS issue that made the leaderboard hard to read in dark mode
|
42 |
+
|
43 |
## [2023-05-22]
|
44 |
- Display a success/error message after submitting evaluation requests
|
45 |
- Reject duplicate submission
|
46 |
- Do not display results that have incomplete results
|
47 |
- Display different queues for jobs that are RUNNING, PENDING, FINISHED status
|
48 |
+
|
49 |
## [2023-05-15]
|
50 |
- Fix a typo: from "TruthQA" to "QA"
|
51 |
+
|
52 |
## [2023-05-10]
|
53 |
- Fix a bug that prevented auto-refresh
|
54 |
+
|
55 |
## [2023-05-10]
|
56 |
- Release the leaderboard to public
|
57 |
"""
|
|
|
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.
|
77 |
- <a href="https://arxiv.org/abs/2009.03300" target="_blank"> MMLU </a> (5-shot) - a test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more.
|
78 |
- <a href="https://arxiv.org/abs/2109.07958" target="_blank"> TruthfulQA </a> (0-shot) - a test to measure a model’s propensity to reproduce falsehoods commonly found online. Note: TruthfulQA in the Harness is actually a minima a 6-shots task, as it is prepended by 6 examples systematically, even when launched using 0 for the number of few-shot examples.
|
79 |
+
|
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
|
|
|
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>`
|
118 |
+
|
119 |
The total batch size we get for models which fit on one A100 node is 16 (8 GPUs * 2). If you don't use parallelism, adapt your batch size to fit.
|
120 |
*You can expect results to vary slightly for different batch sizes because of padding.*
|
121 |
+
|
122 |
The tasks and few shots parameters are:
|
123 |
- ARC: 25-shot, *arc-challenge* (`acc_norm`)
|
124 |
- HellaSwag: 10-shot, *hellaswag* (`acc_norm`)
|
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"""
|
src/auto_leaderboard/get_model_metadata.py
CHANGED
@@ -53,4 +53,4 @@ def get_model_size(model_name, model_info):
|
|
53 |
|
54 |
def apply_metadata(leaderboard_data: List[dict]):
|
55 |
get_model_type(leaderboard_data)
|
56 |
-
get_model_infos_from_hub(leaderboard_data)
|
|
|
53 |
|
54 |
def apply_metadata(leaderboard_data: List[dict]):
|
55 |
get_model_type(leaderboard_data)
|
56 |
+
get_model_infos_from_hub(leaderboard_data)
|
src/auto_leaderboard/load_results.py
CHANGED
@@ -26,7 +26,7 @@ class EvalResult:
|
|
26 |
model: str
|
27 |
revision: str
|
28 |
results: dict
|
29 |
-
precision: str = "
|
30 |
model_type: str = ""
|
31 |
weight_type: str = ""
|
32 |
|
@@ -44,9 +44,7 @@ class EvalResult:
|
|
44 |
data_dict[AutoEvalColumn.model.name] = make_clickable_model(base_model)
|
45 |
data_dict[AutoEvalColumn.dummy.name] = base_model
|
46 |
data_dict[AutoEvalColumn.revision.name] = self.revision
|
47 |
-
data_dict[AutoEvalColumn.average.name] =
|
48 |
-
sum([v for k, v in self.results.items()]) / 4.0, 1
|
49 |
-
)
|
50 |
|
51 |
for benchmark in BENCHMARKS:
|
52 |
if benchmark not in self.results.keys():
|
@@ -76,28 +74,29 @@ def parse_eval_result(json_filepath: str) -> Tuple[str, list[dict]]:
|
|
76 |
model = config.get("model_args", None)
|
77 |
|
78 |
model_sha = config.get("model_sha", "")
|
79 |
-
eval_sha = config.get("lighteval_sha", "")
|
80 |
model_split = model.split("/", 1)
|
81 |
|
|
|
|
|
82 |
model = model_split[-1]
|
83 |
|
84 |
if len(model_split) == 1:
|
85 |
org = None
|
86 |
model = model_split[0]
|
87 |
-
result_key = f"{model}_{model_sha}_{
|
88 |
else:
|
89 |
org = model_split[0]
|
90 |
model = model_split[1]
|
91 |
-
result_key = f"{org}_{model}_{model_sha}_{
|
92 |
|
93 |
eval_results = []
|
94 |
for benchmark, metric in zip(BENCHMARKS, METRICS):
|
95 |
accs = np.array([v[metric] for k, v in data["results"].items() if benchmark in k])
|
96 |
if accs.size == 0:
|
97 |
continue
|
98 |
-
mean_acc =
|
99 |
eval_results.append(EvalResult(
|
100 |
-
eval_name=result_key, org=org, model=model, revision=model_sha, results={benchmark: mean_acc}, #todo model_type=, weight_type=
|
101 |
))
|
102 |
|
103 |
return result_key, eval_results
|
@@ -112,20 +111,15 @@ def get_eval_results(is_public) -> List[EvalResult]:
|
|
112 |
continue
|
113 |
|
114 |
# Sort the files by date
|
|
|
115 |
try:
|
116 |
files.sort(key=lambda x: dateutil.parser.parse(x.split("_", 1)[-1][:-5]))
|
117 |
except dateutil.parser._parser.ParserError:
|
118 |
-
|
119 |
-
|
120 |
-
up_to_date = files[-1]
|
121 |
-
|
122 |
-
if len(files) > 1:
|
123 |
-
print(root)
|
124 |
-
print(files)
|
125 |
-
print(up_to_date)
|
126 |
-
print("===")
|
127 |
|
128 |
-
|
|
|
|
|
129 |
|
130 |
eval_results = {}
|
131 |
for json_filepath in json_filepaths:
|
@@ -144,4 +138,4 @@ def get_eval_results(is_public) -> List[EvalResult]:
|
|
144 |
def get_eval_results_dicts(is_public=True) -> List[Dict]:
|
145 |
eval_results = get_eval_results(is_public)
|
146 |
|
147 |
-
return [e.to_dict() for e in eval_results]
|
|
|
26 |
model: str
|
27 |
revision: str
|
28 |
results: dict
|
29 |
+
precision: str = ""
|
30 |
model_type: str = ""
|
31 |
weight_type: str = ""
|
32 |
|
|
|
44 |
data_dict[AutoEvalColumn.model.name] = make_clickable_model(base_model)
|
45 |
data_dict[AutoEvalColumn.dummy.name] = base_model
|
46 |
data_dict[AutoEvalColumn.revision.name] = self.revision
|
47 |
+
data_dict[AutoEvalColumn.average.name] = sum([v for k, v in self.results.items()]) / 4.0
|
|
|
|
|
48 |
|
49 |
for benchmark in BENCHMARKS:
|
50 |
if benchmark not in self.results.keys():
|
|
|
74 |
model = config.get("model_args", None)
|
75 |
|
76 |
model_sha = config.get("model_sha", "")
|
|
|
77 |
model_split = model.split("/", 1)
|
78 |
|
79 |
+
precision = config.get("model_dtype")
|
80 |
+
|
81 |
model = model_split[-1]
|
82 |
|
83 |
if len(model_split) == 1:
|
84 |
org = None
|
85 |
model = model_split[0]
|
86 |
+
result_key = f"{model}_{model_sha}_{precision}"
|
87 |
else:
|
88 |
org = model_split[0]
|
89 |
model = model_split[1]
|
90 |
+
result_key = f"{org}_{model}_{model_sha}_{precision}"
|
91 |
|
92 |
eval_results = []
|
93 |
for benchmark, metric in zip(BENCHMARKS, METRICS):
|
94 |
accs = np.array([v[metric] for k, v in data["results"].items() if benchmark in k])
|
95 |
if accs.size == 0:
|
96 |
continue
|
97 |
+
mean_acc = np.mean(accs) * 100.0
|
98 |
eval_results.append(EvalResult(
|
99 |
+
eval_name=result_key, org=org, model=model, revision=model_sha, results={benchmark: mean_acc}, precision=precision, #todo model_type=, weight_type=
|
100 |
))
|
101 |
|
102 |
return result_key, eval_results
|
|
|
111 |
continue
|
112 |
|
113 |
# Sort the files by date
|
114 |
+
# store results by precision maybe?
|
115 |
try:
|
116 |
files.sort(key=lambda x: dateutil.parser.parse(x.split("_", 1)[-1][:-5]))
|
117 |
except dateutil.parser._parser.ParserError:
|
118 |
+
files = [files[-1]]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
|
120 |
+
#up_to_date = files[-1]
|
121 |
+
for file in files:
|
122 |
+
json_filepaths.append(os.path.join(root, file))
|
123 |
|
124 |
eval_results = {}
|
125 |
for json_filepath in json_filepaths:
|
|
|
138 |
def get_eval_results_dicts(is_public=True) -> List[Dict]:
|
139 |
eval_results = get_eval_results(is_public)
|
140 |
|
141 |
+
return [e.to_dict() for e in eval_results]
|
src/auto_leaderboard/model_metadata_type.py
CHANGED
@@ -1,5 +1,8 @@
|
|
1 |
from dataclasses import dataclass
|
2 |
from enum import Enum
|
|
|
|
|
|
|
3 |
from typing import Dict, List
|
4 |
|
5 |
from ..utils_display import AutoEvalColumn
|
@@ -9,512 +12,568 @@ class ModelInfo:
|
|
9 |
name: str
|
10 |
symbol: str # emoji
|
11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
class ModelType(Enum):
|
14 |
PT = ModelInfo(name="pretrained", symbol="🟢")
|
15 |
-
|
16 |
-
|
|
|
|
|
|
|
|
|
17 |
|
18 |
|
19 |
TYPE_METADATA: Dict[str, ModelType] = {
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
237 |
-
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
-
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
267 |
-
|
268 |
-
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
-
|
278 |
-
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
|
284 |
-
|
285 |
-
|
286 |
-
|
287 |
-
|
288 |
-
|
289 |
-
|
290 |
-
|
291 |
-
|
292 |
-
|
293 |
-
|
294 |
-
|
295 |
-
|
296 |
-
|
297 |
-
|
298 |
-
|
299 |
-
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
-
|
307 |
-
|
308 |
-
|
309 |
-
|
310 |
-
|
311 |
-
|
312 |
-
|
313 |
-
|
314 |
-
|
315 |
-
|
316 |
-
|
317 |
-
|
318 |
-
|
319 |
-
|
320 |
-
|
321 |
-
|
322 |
-
|
323 |
-
|
324 |
-
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
|
336 |
-
|
337 |
-
|
338 |
-
|
339 |
-
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
349 |
-
|
350 |
-
|
351 |
-
|
352 |
-
|
353 |
-
|
354 |
-
|
355 |
-
|
356 |
-
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
|
372 |
-
|
373 |
-
|
374 |
-
|
375 |
-
|
376 |
-
|
377 |
-
|
378 |
-
|
379 |
-
|
380 |
-
|
381 |
-
|
382 |
-
|
383 |
-
|
384 |
-
|
385 |
-
|
386 |
-
|
387 |
-
|
388 |
-
|
389 |
-
|
390 |
-
|
391 |
-
|
392 |
-
|
393 |
-
|
394 |
-
|
395 |
-
|
396 |
-
|
397 |
-
|
398 |
-
|
399 |
-
|
400 |
-
|
401 |
-
|
402 |
-
|
403 |
-
|
404 |
-
|
405 |
-
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
|
410 |
-
|
411 |
-
|
412 |
-
|
413 |
-
|
414 |
-
|
415 |
-
|
416 |
-
|
417 |
-
|
418 |
-
|
419 |
-
|
420 |
-
|
421 |
-
|
422 |
-
|
423 |
-
|
424 |
-
|
425 |
-
|
426 |
-
|
427 |
-
|
428 |
-
|
429 |
-
|
430 |
-
|
431 |
-
|
432 |
-
|
433 |
-
|
434 |
-
|
435 |
-
|
436 |
-
|
437 |
-
|
438 |
-
|
439 |
-
|
440 |
-
|
441 |
-
|
442 |
-
|
443 |
-
|
444 |
-
|
445 |
-
|
446 |
-
|
447 |
-
|
448 |
-
|
449 |
-
|
450 |
-
|
451 |
-
|
452 |
-
|
453 |
-
|
454 |
-
|
455 |
-
|
456 |
-
|
457 |
-
|
458 |
-
|
459 |
-
|
460 |
-
|
461 |
-
|
462 |
-
|
463 |
-
|
464 |
-
|
465 |
-
|
466 |
-
|
467 |
-
|
468 |
-
|
469 |
-
|
470 |
-
|
471 |
-
|
472 |
-
|
473 |
-
|
474 |
-
|
475 |
-
|
476 |
-
|
477 |
-
|
478 |
-
|
479 |
-
|
480 |
-
|
481 |
-
|
482 |
-
|
483 |
-
|
484 |
-
|
485 |
-
|
486 |
-
|
487 |
-
|
488 |
-
|
489 |
-
|
490 |
-
|
491 |
-
|
492 |
-
|
493 |
-
|
494 |
-
|
495 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
496 |
}
|
497 |
|
498 |
|
499 |
def get_model_type(leaderboard_data: List[dict]):
|
500 |
for model_data in leaderboard_data:
|
501 |
# Todo @clefourrier once requests are connected with results
|
502 |
-
is_delta = False # (model_data["weight_type"] != "Original")
|
503 |
# Stored information
|
504 |
-
|
505 |
-
|
506 |
-
|
507 |
-
|
508 |
-
|
509 |
-
model_data[AutoEvalColumn.
|
510 |
-
|
511 |
-
|
512 |
-
|
513 |
-
|
514 |
-
|
515 |
-
|
516 |
-
|
517 |
-
|
518 |
-
|
519 |
-
|
520 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from dataclasses import dataclass
|
2 |
from enum import Enum
|
3 |
+
import glob
|
4 |
+
import json
|
5 |
+
import os
|
6 |
from typing import Dict, List
|
7 |
|
8 |
from ..utils_display import AutoEvalColumn
|
|
|
12 |
name: str
|
13 |
symbol: str # emoji
|
14 |
|
15 |
+
model_type_symbols = {
|
16 |
+
"fine-tuned": "🔶",
|
17 |
+
"pretrained": "🟢",
|
18 |
+
"RL-tuned": "🟦",
|
19 |
+
"instruction-tuned": "⭕",
|
20 |
+
}
|
21 |
|
22 |
class ModelType(Enum):
|
23 |
PT = ModelInfo(name="pretrained", symbol="🟢")
|
24 |
+
FT = ModelInfo(name="fine-tuned", symbol="🔶")
|
25 |
+
IFT = ModelInfo(name="instruction-tuned", symbol="⭕")
|
26 |
+
RL = ModelInfo(name="RL-tuned", symbol="🟦")
|
27 |
+
|
28 |
+
def to_str(self, separator = " "):
|
29 |
+
return f"{self.value.symbol}{separator}{self.value.name}"
|
30 |
|
31 |
|
32 |
TYPE_METADATA: Dict[str, ModelType] = {
|
33 |
+
'notstoic/PygmalionCoT-7b': ModelType.IFT,
|
34 |
+
'aisquared/dlite-v1-355m': ModelType.IFT,
|
35 |
+
'aisquared/dlite-v1-1_5b': ModelType.IFT,
|
36 |
+
'aisquared/dlite-v1-774m': ModelType.IFT,
|
37 |
+
'aisquared/dlite-v1-124m': ModelType.IFT,
|
38 |
+
'aisquared/chopt-2_7b': ModelType.IFT,
|
39 |
+
'aisquared/dlite-v2-124m': ModelType.IFT,
|
40 |
+
'aisquared/dlite-v2-774m': ModelType.IFT,
|
41 |
+
'aisquared/dlite-v2-1_5b': ModelType.IFT,
|
42 |
+
'aisquared/chopt-1_3b': ModelType.IFT,
|
43 |
+
'aisquared/dlite-v2-355m': ModelType.IFT,
|
44 |
+
'augtoma/qCammel-13': ModelType.IFT,
|
45 |
+
'Aspik101/Llama-2-7b-hf-instruct-pl-lora_unload': ModelType.IFT,
|
46 |
+
'Aspik101/vicuna-7b-v1.3-instruct-pl-lora_unload': ModelType.IFT,
|
47 |
+
'TheBloke/alpaca-lora-65B-HF': ModelType.FT,
|
48 |
+
'TheBloke/tulu-7B-fp16': ModelType.IFT,
|
49 |
+
'TheBloke/guanaco-7B-HF': ModelType.FT,
|
50 |
+
'TheBloke/koala-7B-HF': ModelType.FT,
|
51 |
+
'TheBloke/wizardLM-7B-HF': ModelType.IFT,
|
52 |
+
'TheBloke/airoboros-13B-HF': ModelType.IFT,
|
53 |
+
'TheBloke/koala-13B-HF': ModelType.FT,
|
54 |
+
'TheBloke/Wizard-Vicuna-7B-Uncensored-HF': ModelType.FT,
|
55 |
+
'TheBloke/dromedary-65b-lora-HF': ModelType.IFT,
|
56 |
+
'TheBloke/wizardLM-13B-1.0-fp16': ModelType.IFT,
|
57 |
+
'TheBloke/WizardLM-13B-V1-1-SuperHOT-8K-fp16': ModelType.FT,
|
58 |
+
'TheBloke/Wizard-Vicuna-30B-Uncensored-fp16': ModelType.FT,
|
59 |
+
'TheBloke/wizard-vicuna-13B-HF': ModelType.IFT,
|
60 |
+
'TheBloke/UltraLM-13B-fp16': ModelType.IFT,
|
61 |
+
'TheBloke/OpenAssistant-FT-7-Llama-30B-HF': ModelType.FT,
|
62 |
+
'TheBloke/vicuna-13B-1.1-HF': ModelType.IFT,
|
63 |
+
'TheBloke/guanaco-13B-HF': ModelType.FT,
|
64 |
+
'TheBloke/guanaco-65B-HF': ModelType.FT,
|
65 |
+
'TheBloke/airoboros-7b-gpt4-fp16': ModelType.IFT,
|
66 |
+
'TheBloke/llama-30b-supercot-SuperHOT-8K-fp16': ModelType.IFT,
|
67 |
+
'TheBloke/Llama-2-13B-fp16': ModelType.PT,
|
68 |
+
'TheBloke/llama-2-70b-Guanaco-QLoRA-fp16': ModelType.FT,
|
69 |
+
'TheBloke/landmark-attention-llama7b-fp16': ModelType.IFT,
|
70 |
+
'TheBloke/Planner-7B-fp16': ModelType.IFT,
|
71 |
+
'TheBloke/Wizard-Vicuna-13B-Uncensored-HF': ModelType.FT,
|
72 |
+
'TheBloke/gpt4-alpaca-lora-13B-HF': ModelType.IFT,
|
73 |
+
'TheBloke/gpt4-x-vicuna-13B-HF': ModelType.IFT,
|
74 |
+
'TheBloke/gpt4-alpaca-lora_mlp-65B-HF': ModelType.IFT,
|
75 |
+
'TheBloke/tulu-13B-fp16': ModelType.IFT,
|
76 |
+
'TheBloke/VicUnlocked-alpaca-65B-QLoRA-fp16': ModelType.IFT,
|
77 |
+
'TheBloke/Llama-2-70B-fp16': ModelType.IFT,
|
78 |
+
'TheBloke/WizardLM-30B-fp16': ModelType.IFT,
|
79 |
+
'TheBloke/robin-13B-v2-fp16': ModelType.FT,
|
80 |
+
'TheBloke/robin-33B-v2-fp16': ModelType.FT,
|
81 |
+
'TheBloke/Vicuna-13B-CoT-fp16': ModelType.IFT,
|
82 |
+
'TheBloke/Vicuna-33B-1-3-SuperHOT-8K-fp16': ModelType.IFT,
|
83 |
+
'TheBloke/Wizard-Vicuna-30B-Superhot-8K-fp16': ModelType.FT,
|
84 |
+
'TheBloke/Nous-Hermes-13B-SuperHOT-8K-fp16': ModelType.IFT,
|
85 |
+
'TheBloke/GPlatty-30B-SuperHOT-8K-fp16': ModelType.FT,
|
86 |
+
'TheBloke/CAMEL-33B-Combined-Data-SuperHOT-8K-fp16': ModelType.IFT,
|
87 |
+
'TheBloke/Chinese-Alpaca-33B-SuperHOT-8K-fp16': ModelType.IFT,
|
88 |
+
'jphme/orca_mini_v2_ger_7b': ModelType.IFT,
|
89 |
+
'Ejafa/vicuna_7B_vanilla_1.1': ModelType.FT,
|
90 |
+
'kevinpro/Vicuna-13B-CoT': ModelType.IFT,
|
91 |
+
'AlekseyKorshuk/pygmalion-6b-vicuna-chatml': ModelType.FT,
|
92 |
+
'AlekseyKorshuk/chatml-pyg-v1': ModelType.FT,
|
93 |
+
'concedo/Vicuzard-30B-Uncensored': ModelType.FT,
|
94 |
+
'concedo/OPT-19M-ChatSalad': ModelType.FT,
|
95 |
+
'concedo/Pythia-70M-ChatSalad': ModelType.FT,
|
96 |
+
'digitous/13B-HyperMantis': ModelType.IFT,
|
97 |
+
'digitous/Adventien-GPTJ': ModelType.FT,
|
98 |
+
'digitous/Alpacino13b': ModelType.IFT,
|
99 |
+
'digitous/GPT-R': ModelType.IFT,
|
100 |
+
'digitous/Javelin-R': ModelType.IFT,
|
101 |
+
'digitous/Javalion-GPTJ': ModelType.IFT,
|
102 |
+
'digitous/Javalion-R': ModelType.IFT,
|
103 |
+
'digitous/Skegma-GPTJ': ModelType.FT,
|
104 |
+
'digitous/Alpacino30b': ModelType.IFT,
|
105 |
+
'digitous/Janin-GPTJ': ModelType.FT,
|
106 |
+
'digitous/Janin-R': ModelType.FT,
|
107 |
+
'digitous/Javelin-GPTJ': ModelType.FT,
|
108 |
+
'SaylorTwift/gpt2_test': ModelType.PT,
|
109 |
+
'anton-l/gpt-j-tiny-random': ModelType.FT,
|
110 |
+
'Andron00e/YetAnother_Open-Llama-3B-LoRA-OpenOrca': ModelType.FT,
|
111 |
+
'Lazycuber/pyg-instruct-wizardlm': ModelType.FT,
|
112 |
+
'Lazycuber/Janemalion-6B': ModelType.FT,
|
113 |
+
'IDEA-CCNL/Ziya-LLaMA-13B-Pretrain-v1': ModelType.FT,
|
114 |
+
'IDEA-CCNL/Ziya-LLaMA-13B-v1': ModelType.IFT,
|
115 |
+
'dsvv-cair/alpaca-cleaned-llama-30b-bf16': ModelType.FT,
|
116 |
+
'gpt2-medium': ModelType.PT,
|
117 |
+
'camel-ai/CAMEL-13B-Combined-Data': ModelType.IFT,
|
118 |
+
'camel-ai/CAMEL-13B-Role-Playing-Data': ModelType.FT,
|
119 |
+
'camel-ai/CAMEL-33B-Combined-Data': ModelType.IFT,
|
120 |
+
'PygmalionAI/pygmalion-6b': ModelType.FT,
|
121 |
+
'PygmalionAI/metharme-1.3b': ModelType.IFT,
|
122 |
+
'PygmalionAI/pygmalion-1.3b': ModelType.FT,
|
123 |
+
'PygmalionAI/pygmalion-350m': ModelType.FT,
|
124 |
+
'PygmalionAI/pygmalion-2.7b': ModelType.FT,
|
125 |
+
'medalpaca/medalpaca-7b': ModelType.FT,
|
126 |
+
'lilloukas/Platypus-30B': ModelType.IFT,
|
127 |
+
'lilloukas/GPlatty-30B': ModelType.FT,
|
128 |
+
'mncai/chatdoctor': ModelType.FT,
|
129 |
+
'chaoyi-wu/MedLLaMA_13B': ModelType.FT,
|
130 |
+
'LoupGarou/WizardCoder-Guanaco-15B-V1.0': ModelType.IFT,
|
131 |
+
'LoupGarou/WizardCoder-Guanaco-15B-V1.1': ModelType.FT,
|
132 |
+
'hakurei/instruct-12b': ModelType.IFT,
|
133 |
+
'hakurei/lotus-12B': ModelType.FT,
|
134 |
+
'shibing624/chinese-llama-plus-13b-hf': ModelType.IFT,
|
135 |
+
'shibing624/chinese-alpaca-plus-7b-hf': ModelType.IFT,
|
136 |
+
'shibing624/chinese-alpaca-plus-13b-hf': ModelType.IFT,
|
137 |
+
'mosaicml/mpt-7b-instruct': ModelType.IFT,
|
138 |
+
'mosaicml/mpt-30b-chat': ModelType.IFT,
|
139 |
+
'mosaicml/mpt-7b-storywriter': ModelType.FT,
|
140 |
+
'mosaicml/mpt-30b-instruct': ModelType.IFT,
|
141 |
+
'mosaicml/mpt-7b-chat': ModelType.IFT,
|
142 |
+
'mosaicml/mpt-30b': ModelType.PT,
|
143 |
+
'Corianas/111m': ModelType.IFT,
|
144 |
+
'Corianas/Quokka_1.3b': ModelType.IFT,
|
145 |
+
'Corianas/256_5epoch': ModelType.FT,
|
146 |
+
'Corianas/Quokka_256m': ModelType.IFT,
|
147 |
+
'Corianas/Quokka_590m': ModelType.IFT,
|
148 |
+
'Corianas/gpt-j-6B-Dolly': ModelType.FT,
|
149 |
+
'Corianas/Quokka_2.7b': ModelType.IFT,
|
150 |
+
'cyberagent/open-calm-7b': ModelType.FT,
|
151 |
+
'Aspik101/Nous-Hermes-13b-pl-lora_unload': ModelType.IFT,
|
152 |
+
'THUDM/chatglm2-6b': ModelType.IFT,
|
153 |
+
'MetaIX/GPT4-X-Alpasta-30b': ModelType.IFT,
|
154 |
+
'NYTK/PULI-GPTrio': ModelType.PT,
|
155 |
+
'EleutherAI/pythia-1.3b': ModelType.PT,
|
156 |
+
'EleutherAI/pythia-2.8b-deduped': ModelType.PT,
|
157 |
+
'EleutherAI/gpt-neo-125m': ModelType.PT,
|
158 |
+
'EleutherAI/pythia-160m': ModelType.PT,
|
159 |
+
'EleutherAI/gpt-neo-2.7B': ModelType.PT,
|
160 |
+
'EleutherAI/pythia-1b-deduped': ModelType.PT,
|
161 |
+
'EleutherAI/pythia-6.7b': ModelType.PT,
|
162 |
+
'EleutherAI/pythia-70m-deduped': ModelType.PT,
|
163 |
+
'EleutherAI/gpt-neox-20b': ModelType.PT,
|
164 |
+
'EleutherAI/pythia-1.4b-deduped': ModelType.PT,
|
165 |
+
'EleutherAI/pythia-2.7b': ModelType.PT,
|
166 |
+
'EleutherAI/pythia-6.9b-deduped': ModelType.PT,
|
167 |
+
'EleutherAI/pythia-70m': ModelType.PT,
|
168 |
+
'EleutherAI/gpt-j-6b': ModelType.PT,
|
169 |
+
'EleutherAI/pythia-12b-deduped': ModelType.PT,
|
170 |
+
'EleutherAI/gpt-neo-1.3B': ModelType.PT,
|
171 |
+
'EleutherAI/pythia-410m-deduped': ModelType.PT,
|
172 |
+
'EleutherAI/pythia-160m-deduped': ModelType.PT,
|
173 |
+
'EleutherAI/polyglot-ko-12.8b': ModelType.PT,
|
174 |
+
'EleutherAI/pythia-12b': ModelType.PT,
|
175 |
+
'roneneldan/TinyStories-33M': ModelType.PT,
|
176 |
+
'roneneldan/TinyStories-28M': ModelType.PT,
|
177 |
+
'roneneldan/TinyStories-1M': ModelType.PT,
|
178 |
+
'roneneldan/TinyStories-8M': ModelType.PT,
|
179 |
+
'roneneldan/TinyStories-3M': ModelType.PT,
|
180 |
+
'jerryjalapeno/nart-100k-7b': ModelType.FT,
|
181 |
+
'lmsys/vicuna-13b-v1.3': ModelType.IFT,
|
182 |
+
'lmsys/vicuna-7b-v1.3': ModelType.IFT,
|
183 |
+
'lmsys/vicuna-13b-v1.1': ModelType.IFT,
|
184 |
+
'lmsys/vicuna-13b-delta-v1.1': ModelType.IFT,
|
185 |
+
'lmsys/vicuna-7b-delta-v1.1': ModelType.IFT,
|
186 |
+
'abhiramtirumala/DialoGPT-sarcastic-medium': ModelType.FT,
|
187 |
+
'haonan-li/bactrian-x-llama-13b-merged': ModelType.IFT,
|
188 |
+
'Gryphe/MythoLogic-13b': ModelType.IFT,
|
189 |
+
'Gryphe/MythoBoros-13b': ModelType.IFT,
|
190 |
+
'pillowtalks-ai/delta13b': ModelType.FT,
|
191 |
+
'wannaphong/openthaigpt-0.1.0-beta-full-model_for_open_llm_leaderboard': ModelType.FT,
|
192 |
+
'bigscience/bloom-7b1': ModelType.PT,
|
193 |
+
'bigcode/tiny_starcoder_py': ModelType.PT,
|
194 |
+
'bigcode/starcoderplus': ModelType.FT,
|
195 |
+
'bigcode/gpt_bigcode-santacoder': ModelType.PT,
|
196 |
+
'bigcode/starcoder': ModelType.PT,
|
197 |
+
'Open-Orca/OpenOrca-Preview1-13B': ModelType.IFT,
|
198 |
+
'microsoft/DialoGPT-large': ModelType.FT,
|
199 |
+
'microsoft/DialoGPT-small': ModelType.FT,
|
200 |
+
'microsoft/DialoGPT-medium': ModelType.FT,
|
201 |
+
'microsoft/CodeGPT-small-py': ModelType.FT,
|
202 |
+
'Tincando/fiction_story_generator': ModelType.FT,
|
203 |
+
'Pirr/pythia-13b-deduped-green_devil': ModelType.FT,
|
204 |
+
'Aeala/GPT4-x-AlpacaDente2-30b': ModelType.FT,
|
205 |
+
'Aeala/GPT4-x-AlpacaDente-30b': ModelType.FT,
|
206 |
+
'Aeala/GPT4-x-Alpasta-13b': ModelType.FT,
|
207 |
+
'Aeala/VicUnlocked-alpaca-30b': ModelType.IFT,
|
208 |
+
'Tap-M/Luna-AI-Llama2-Uncensored': ModelType.FT,
|
209 |
+
'illuin/test-custom-llama': ModelType.FT,
|
210 |
+
'dvruette/oasst-llama-13b-2-epochs': ModelType.FT,
|
211 |
+
'dvruette/oasst-gpt-neox-20b-1000-steps': ModelType.FT,
|
212 |
+
'dvruette/llama-13b-pretrained-dropout': ModelType.PT,
|
213 |
+
'dvruette/llama-13b-pretrained': ModelType.PT,
|
214 |
+
'dvruette/llama-13b-pretrained-sft-epoch-1': ModelType.PT,
|
215 |
+
'dvruette/llama-13b-pretrained-sft-do2': ModelType.PT,
|
216 |
+
'dvruette/oasst-gpt-neox-20b-3000-steps': ModelType.FT,
|
217 |
+
'dvruette/oasst-pythia-12b-pretrained-sft': ModelType.PT,
|
218 |
+
'dvruette/oasst-pythia-6.9b-4000-steps': ModelType.FT,
|
219 |
+
'dvruette/gpt-neox-20b-full-precision': ModelType.FT,
|
220 |
+
'dvruette/oasst-llama-13b-1000-steps': ModelType.FT,
|
221 |
+
'openlm-research/open_llama_7b_700bt_preview': ModelType.PT,
|
222 |
+
'openlm-research/open_llama_7b': ModelType.PT,
|
223 |
+
'openlm-research/open_llama_7b_v2': ModelType.PT,
|
224 |
+
'openlm-research/open_llama_3b': ModelType.PT,
|
225 |
+
'openlm-research/open_llama_13b': ModelType.PT,
|
226 |
+
'openlm-research/open_llama_3b_v2': ModelType.PT,
|
227 |
+
'PocketDoc/Dans-PileOfSets-Mk1-llama-13b-merged': ModelType.IFT,
|
228 |
+
'GeorgiaTechResearchInstitute/galpaca-30b': ModelType.IFT,
|
229 |
+
'GeorgiaTechResearchInstitute/starcoder-gpteacher-code-instruct': ModelType.IFT,
|
230 |
+
'databricks/dolly-v2-7b': ModelType.IFT,
|
231 |
+
'databricks/dolly-v2-3b': ModelType.IFT,
|
232 |
+
'databricks/dolly-v2-12b': ModelType.IFT,
|
233 |
+
'Rachneet/gpt2-xl-alpaca': ModelType.FT,
|
234 |
+
'Locutusque/gpt2-conversational-or-qa': ModelType.FT,
|
235 |
+
'psyche/kogpt': ModelType.FT,
|
236 |
+
'NbAiLab/nb-gpt-j-6B-alpaca': ModelType.IFT,
|
237 |
+
'Mikael110/llama-2-7b-guanaco-fp16': ModelType.FT,
|
238 |
+
'Mikael110/llama-2-13b-guanaco-fp16': ModelType.FT,
|
239 |
+
'Fredithefish/CrimsonPajama': ModelType.IFT,
|
240 |
+
'Fredithefish/RedPajama-INCITE-Chat-3B-ShareGPT-11K': ModelType.FT,
|
241 |
+
'Fredithefish/ScarletPajama-3B-HF': ModelType.FT,
|
242 |
+
'Fredithefish/RedPajama-INCITE-Chat-3B-Instruction-Tuning-with-GPT-4': ModelType.IFT,
|
243 |
+
'acrastt/RedPajama-INCITE-Chat-Instruct-3B-V1': ModelType.IFT,
|
244 |
+
'eachadea/vicuna-13b-1.1': ModelType.FT,
|
245 |
+
'eachadea/vicuna-7b-1.1': ModelType.FT,
|
246 |
+
'eachadea/vicuna-13b': ModelType.FT,
|
247 |
+
'openaccess-ai-collective/wizard-mega-13b': ModelType.IFT,
|
248 |
+
'openaccess-ai-collective/manticore-13b': ModelType.IFT,
|
249 |
+
'openaccess-ai-collective/manticore-30b-chat-pyg-alpha': ModelType.IFT,
|
250 |
+
'openaccess-ai-collective/minotaur-13b': ModelType.IFT,
|
251 |
+
'openaccess-ai-collective/minotaur-13b-fixed': ModelType.IFT,
|
252 |
+
'openaccess-ai-collective/hippogriff-30b-chat': ModelType.IFT,
|
253 |
+
'openaccess-ai-collective/manticore-13b-chat-pyg': ModelType.IFT,
|
254 |
+
'pythainlp/wangchanglm-7.5B-sft-enth': ModelType.IFT,
|
255 |
+
'pythainlp/wangchanglm-7.5B-sft-en-sharded': ModelType.IFT,
|
256 |
+
'euclaise/gpt-neox-122m-minipile-digits': ModelType.FT,
|
257 |
+
'stabilityai/StableBeluga1-Delta': ModelType.IFT,
|
258 |
+
'stabilityai/stablelm-tuned-alpha-7b': ModelType.IFT,
|
259 |
+
'stabilityai/StableBeluga2': ModelType.IFT,
|
260 |
+
'stabilityai/StableBeluga-13B': ModelType.IFT,
|
261 |
+
'stabilityai/StableBeluga-7B': ModelType.IFT,
|
262 |
+
'stabilityai/stablelm-base-alpha-7b': ModelType.PT,
|
263 |
+
'stabilityai/stablelm-base-alpha-3b': ModelType.PT,
|
264 |
+
'stabilityai/stablelm-tuned-alpha-3b': ModelType.IFT,
|
265 |
+
'alibidaran/medical_transcription_generator': ModelType.FT,
|
266 |
+
'CalderaAI/30B-Lazarus': ModelType.IFT,
|
267 |
+
'CalderaAI/13B-BlueMethod': ModelType.IFT,
|
268 |
+
'CalderaAI/13B-Ouroboros': ModelType.IFT,
|
269 |
+
'KoboldAI/OPT-13B-Erebus': ModelType.FT,
|
270 |
+
'KoboldAI/GPT-J-6B-Janeway': ModelType.FT,
|
271 |
+
'KoboldAI/GPT-J-6B-Shinen': ModelType.FT,
|
272 |
+
'KoboldAI/fairseq-dense-2.7B': ModelType.PT,
|
273 |
+
'KoboldAI/OPT-6B-nerys-v2': ModelType.FT,
|
274 |
+
'KoboldAI/GPT-NeoX-20B-Skein': ModelType.FT,
|
275 |
+
'KoboldAI/PPO_Pygway-6b-Mix': ModelType.FT,
|
276 |
+
'KoboldAI/fairseq-dense-6.7B': ModelType.PT,
|
277 |
+
'KoboldAI/fairseq-dense-125M': ModelType.PT,
|
278 |
+
'KoboldAI/OPT-13B-Nerybus-Mix': ModelType.FT,
|
279 |
+
'KoboldAI/OPT-2.7B-Erebus': ModelType.FT,
|
280 |
+
'KoboldAI/OPT-350M-Nerys-v2': ModelType.FT,
|
281 |
+
'KoboldAI/OPT-2.7B-Nerys-v2': ModelType.FT,
|
282 |
+
'KoboldAI/OPT-2.7B-Nerybus-Mix': ModelType.FT,
|
283 |
+
'KoboldAI/OPT-13B-Nerys-v2': ModelType.FT,
|
284 |
+
'KoboldAI/GPT-NeoX-20B-Erebus': ModelType.FT,
|
285 |
+
'KoboldAI/OPT-6.7B-Erebus': ModelType.FT,
|
286 |
+
'KoboldAI/fairseq-dense-355M': ModelType.PT,
|
287 |
+
'KoboldAI/OPT-6.7B-Nerybus-Mix': ModelType.FT,
|
288 |
+
'KoboldAI/GPT-J-6B-Adventure': ModelType.FT,
|
289 |
+
'KoboldAI/OPT-350M-Erebus': ModelType.FT,
|
290 |
+
'KoboldAI/GPT-J-6B-Skein': ModelType.FT,
|
291 |
+
'KoboldAI/OPT-30B-Erebus': ModelType.FT,
|
292 |
+
'klosax/pythia-160m-deduped-step92k-193bt': ModelType.PT,
|
293 |
+
'klosax/open_llama_3b_350bt_preview': ModelType.PT,
|
294 |
+
'klosax/openllama-3b-350bt': ModelType.PT,
|
295 |
+
'klosax/pythia-70m-deduped-step44k-92bt': ModelType.PT,
|
296 |
+
'klosax/open_llama_13b_600bt_preview': ModelType.PT,
|
297 |
+
'klosax/open_llama_7b_400bt_preview': ModelType.PT,
|
298 |
+
'kfkas/Llama-2-ko-7b-Chat': ModelType.IFT,
|
299 |
+
'WeOpenML/Alpaca-7B-v1': ModelType.IFT,
|
300 |
+
'WeOpenML/PandaLM-Alpaca-7B-v1': ModelType.IFT,
|
301 |
+
'TFLai/gpt2-turkish-uncased': ModelType.FT,
|
302 |
+
'ehartford/WizardLM-13B-Uncensored': ModelType.IFT,
|
303 |
+
'ehartford/dolphin-llama-13b': ModelType.IFT,
|
304 |
+
'ehartford/Wizard-Vicuna-30B-Uncensored': ModelType.FT,
|
305 |
+
'ehartford/WizardLM-30B-Uncensored': ModelType.IFT,
|
306 |
+
'ehartford/Wizard-Vicuna-13B-Uncensored': ModelType.FT,
|
307 |
+
'ehartford/WizardLM-7B-Uncensored': ModelType.IFT,
|
308 |
+
'ehartford/based-30b': ModelType.FT,
|
309 |
+
'ehartford/Wizard-Vicuna-7B-Uncensored': ModelType.FT,
|
310 |
+
'wahaha1987/llama_7b_sharegpt94k_fastchat': ModelType.FT,
|
311 |
+
'wahaha1987/llama_13b_sharegpt94k_fastchat': ModelType.FT,
|
312 |
+
'OpenAssistant/oasst-sft-1-pythia-12b': ModelType.FT,
|
313 |
+
'OpenAssistant/stablelm-7b-sft-v7-epoch-3': ModelType.IFT,
|
314 |
+
'OpenAssistant/oasst-sft-4-pythia-12b-epoch-3.5': ModelType.FT,
|
315 |
+
'OpenAssistant/pythia-12b-sft-v8-2.5k-steps': ModelType.IFT,
|
316 |
+
'OpenAssistant/pythia-12b-sft-v8-7k-steps': ModelType.IFT,
|
317 |
+
'OpenAssistant/pythia-12b-pre-v8-12.5k-steps': ModelType.IFT,
|
318 |
+
'OpenAssistant/llama2-13b-orca-8k-3319': ModelType.IFT,
|
319 |
+
'junelee/wizard-vicuna-13b': ModelType.FT,
|
320 |
+
'BreadAi/gpt-YA-1-1_160M': ModelType.PT,
|
321 |
+
'BreadAi/MuseCan': ModelType.PT,
|
322 |
+
'BreadAi/MusePy-1-2': ModelType.PT,
|
323 |
+
'BreadAi/DiscordPy': ModelType.PT,
|
324 |
+
'BreadAi/PM_modelV2': ModelType.PT,
|
325 |
+
'BreadAi/gpt-Youtube': ModelType.PT,
|
326 |
+
'BreadAi/StoryPy': ModelType.FT,
|
327 |
+
'julianweng/Llama-2-7b-chat-orcah': ModelType.FT,
|
328 |
+
'AGI-inc/lora_moe_7b_baseline': ModelType.FT,
|
329 |
+
'AGI-inc/lora_moe_7b': ModelType.FT,
|
330 |
+
'togethercomputer/GPT-NeoXT-Chat-Base-20B': ModelType.IFT,
|
331 |
+
'togethercomputer/RedPajama-INCITE-Chat-7B-v0.1': ModelType.IFT,
|
332 |
+
'togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1': ModelType.IFT,
|
333 |
+
'togethercomputer/RedPajama-INCITE-7B-Base': ModelType.PT,
|
334 |
+
'togethercomputer/RedPajama-INCITE-7B-Instruct': ModelType.IFT,
|
335 |
+
'togethercomputer/RedPajama-INCITE-Base-3B-v1': ModelType.PT,
|
336 |
+
'togethercomputer/Pythia-Chat-Base-7B': ModelType.IFT,
|
337 |
+
'togethercomputer/RedPajama-INCITE-Base-7B-v0.1': ModelType.PT,
|
338 |
+
'togethercomputer/GPT-JT-6B-v1': ModelType.IFT,
|
339 |
+
'togethercomputer/GPT-JT-6B-v0': ModelType.IFT,
|
340 |
+
'togethercomputer/RedPajama-INCITE-Chat-3B-v1': ModelType.IFT,
|
341 |
+
'togethercomputer/RedPajama-INCITE-7B-Chat': ModelType.IFT,
|
342 |
+
'togethercomputer/RedPajama-INCITE-Instruct-3B-v1': ModelType.IFT,
|
343 |
+
'Writer/camel-5b-hf': ModelType.IFT,
|
344 |
+
'Writer/palmyra-base': ModelType.PT,
|
345 |
+
'MBZUAI/LaMini-GPT-1.5B': ModelType.IFT,
|
346 |
+
'MBZUAI/lamini-cerebras-111m': ModelType.IFT,
|
347 |
+
'MBZUAI/lamini-neo-1.3b': ModelType.IFT,
|
348 |
+
'MBZUAI/lamini-cerebras-1.3b': ModelType.IFT,
|
349 |
+
'MBZUAI/lamini-cerebras-256m': ModelType.IFT,
|
350 |
+
'MBZUAI/LaMini-GPT-124M': ModelType.IFT,
|
351 |
+
'MBZUAI/lamini-neo-125m': ModelType.IFT,
|
352 |
+
'TehVenom/DiffMerge-DollyGPT-Pygmalion': ModelType.FT,
|
353 |
+
'TehVenom/PPO_Shygmalion-6b': ModelType.FT,
|
354 |
+
'TehVenom/Dolly_Shygmalion-6b-Dev_V8P2': ModelType.FT,
|
355 |
+
'TehVenom/Pygmalion_AlpacaLora-7b': ModelType.FT,
|
356 |
+
'TehVenom/PPO_Pygway-V8p4_Dev-6b': ModelType.FT,
|
357 |
+
'TehVenom/Dolly_Malion-6b': ModelType.FT,
|
358 |
+
'TehVenom/PPO_Shygmalion-V8p4_Dev-6b': ModelType.FT,
|
359 |
+
'TehVenom/ChanMalion': ModelType.FT,
|
360 |
+
'TehVenom/GPT-J-Pyg_PPO-6B': ModelType.IFT,
|
361 |
+
'TehVenom/Pygmalion-13b-Merged': ModelType.FT,
|
362 |
+
'TehVenom/Metharme-13b-Merged': ModelType.IFT,
|
363 |
+
'TehVenom/Dolly_Shygmalion-6b': ModelType.FT,
|
364 |
+
'TehVenom/GPT-J-Pyg_PPO-6B-Dev-V8p4': ModelType.IFT,
|
365 |
+
'georgesung/llama2_7b_chat_uncensored': ModelType.FT,
|
366 |
+
'vicgalle/gpt2-alpaca': ModelType.IFT,
|
367 |
+
'vicgalle/alpaca-7b': ModelType.FT,
|
368 |
+
'vicgalle/gpt2-alpaca-gpt4': ModelType.IFT,
|
369 |
+
'facebook/opt-350m': ModelType.PT,
|
370 |
+
'facebook/opt-125m': ModelType.PT,
|
371 |
+
'facebook/xglm-4.5B': ModelType.PT,
|
372 |
+
'facebook/opt-2.7b': ModelType.PT,
|
373 |
+
'facebook/opt-6.7b': ModelType.PT,
|
374 |
+
'facebook/galactica-30b': ModelType.PT,
|
375 |
+
'facebook/opt-13b': ModelType.PT,
|
376 |
+
'facebook/opt-66b': ModelType.PT,
|
377 |
+
'facebook/xglm-7.5B': ModelType.PT,
|
378 |
+
'facebook/xglm-564M': ModelType.PT,
|
379 |
+
'facebook/opt-30b': ModelType.PT,
|
380 |
+
'golaxy/gogpt-7b': ModelType.FT,
|
381 |
+
'golaxy/gogpt2-7b': ModelType.FT,
|
382 |
+
'golaxy/gogpt-7b-bloom': ModelType.FT,
|
383 |
+
'golaxy/gogpt-3b-bloom': ModelType.FT,
|
384 |
+
'psmathur/orca_mini_v2_7b': ModelType.IFT,
|
385 |
+
'psmathur/orca_mini_7b': ModelType.IFT,
|
386 |
+
'psmathur/orca_mini_3b': ModelType.IFT,
|
387 |
+
'psmathur/orca_mini_v2_13b': ModelType.IFT,
|
388 |
+
'gpt2-xl': ModelType.PT,
|
389 |
+
'lxe/Cerebras-GPT-2.7B-Alpaca-SP': ModelType.FT,
|
390 |
+
'Monero/Manticore-13b-Chat-Pyg-Guanaco': ModelType.FT,
|
391 |
+
'Monero/WizardLM-Uncensored-SuperCOT-StoryTelling-30b': ModelType.IFT,
|
392 |
+
'Monero/WizardLM-13b-OpenAssistant-Uncensored': ModelType.IFT,
|
393 |
+
'Monero/WizardLM-30B-Uncensored-Guanaco-SuperCOT-30b': ModelType.IFT,
|
394 |
+
'jzjiao/opt-1.3b-rlhf': ModelType.FT,
|
395 |
+
'HuggingFaceH4/starchat-beta': ModelType.IFT,
|
396 |
+
'KnutJaegersberg/gpt-2-xl-EvolInstruct': ModelType.IFT,
|
397 |
+
'KnutJaegersberg/megatron-GPT-2-345m-EvolInstruct': ModelType.IFT,
|
398 |
+
'KnutJaegersberg/galactica-orca-wizardlm-1.3b': ModelType.IFT,
|
399 |
+
'openchat/openchat_8192': ModelType.IFT,
|
400 |
+
'openchat/openchat_v2': ModelType.IFT,
|
401 |
+
'openchat/openchat_v2_w': ModelType.IFT,
|
402 |
+
'ausboss/llama-13b-supercot': ModelType.IFT,
|
403 |
+
'ausboss/llama-30b-supercot': ModelType.IFT,
|
404 |
+
'Neko-Institute-of-Science/metharme-7b': ModelType.IFT,
|
405 |
+
'Neko-Institute-of-Science/pygmalion-7b': ModelType.FT,
|
406 |
+
'SebastianSchramm/Cerebras-GPT-111M-instruction': ModelType.IFT,
|
407 |
+
'victor123/WizardLM-13B-1.0': ModelType.IFT,
|
408 |
+
'OpenBuddy/openbuddy-openllama-13b-v7-fp16': ModelType.FT,
|
409 |
+
'OpenBuddy/openbuddy-llama2-13b-v8.1-fp16': ModelType.FT,
|
410 |
+
'OpenBuddyEA/openbuddy-llama-30b-v7.1-bf16': ModelType.FT,
|
411 |
+
'baichuan-inc/Baichuan-7B': ModelType.PT,
|
412 |
+
'tiiuae/falcon-40b-instruct': ModelType.IFT,
|
413 |
+
'tiiuae/falcon-40b': ModelType.PT,
|
414 |
+
'tiiuae/falcon-7b': ModelType.PT,
|
415 |
+
'YeungNLP/firefly-llama-13b': ModelType.FT,
|
416 |
+
'YeungNLP/firefly-llama-13b-v1.2': ModelType.FT,
|
417 |
+
'YeungNLP/firefly-llama2-13b': ModelType.FT,
|
418 |
+
'YeungNLP/firefly-ziya-13b': ModelType.FT,
|
419 |
+
'shaohang/Sparse0.5_OPT-1.3': ModelType.FT,
|
420 |
+
'xzuyn/Alpacino-SuperCOT-13B': ModelType.IFT,
|
421 |
+
'xzuyn/MedicWizard-7B': ModelType.FT,
|
422 |
+
'xDAN-AI/xDAN_13b_l2_lora': ModelType.FT,
|
423 |
+
'beomi/KoAlpaca-Polyglot-5.8B': ModelType.FT,
|
424 |
+
'beomi/llama-2-ko-7b': ModelType.IFT,
|
425 |
+
'Salesforce/codegen-6B-multi': ModelType.PT,
|
426 |
+
'Salesforce/codegen-16B-nl': ModelType.PT,
|
427 |
+
'Salesforce/codegen-6B-nl': ModelType.PT,
|
428 |
+
'ai-forever/rugpt3large_based_on_gpt2': ModelType.FT,
|
429 |
+
'gpt2-large': ModelType.PT,
|
430 |
+
'frank098/orca_mini_3b_juniper': ModelType.FT,
|
431 |
+
'frank098/WizardLM_13B_juniper': ModelType.FT,
|
432 |
+
'FPHam/Free_Sydney_13b_HF': ModelType.FT,
|
433 |
+
'huggingface/llama-13b': ModelType.PT,
|
434 |
+
'huggingface/llama-7b': ModelType.PT,
|
435 |
+
'huggingface/llama-65b': ModelType.PT,
|
436 |
+
'huggingface/llama-30b': ModelType.PT,
|
437 |
+
'Henk717/chronoboros-33B': ModelType.IFT,
|
438 |
+
'jondurbin/airoboros-13b-gpt4-1.4': ModelType.IFT,
|
439 |
+
'jondurbin/airoboros-7b': ModelType.IFT,
|
440 |
+
'jondurbin/airoboros-7b-gpt4': ModelType.IFT,
|
441 |
+
'jondurbin/airoboros-7b-gpt4-1.1': ModelType.IFT,
|
442 |
+
'jondurbin/airoboros-7b-gpt4-1.2': ModelType.IFT,
|
443 |
+
'jondurbin/airoboros-7b-gpt4-1.3': ModelType.IFT,
|
444 |
+
'jondurbin/airoboros-7b-gpt4-1.4': ModelType.IFT,
|
445 |
+
'jondurbin/airoboros-l2-7b-gpt4-1.4.1': ModelType.IFT,
|
446 |
+
'jondurbin/airoboros-l2-13b-gpt4-1.4.1': ModelType.IFT,
|
447 |
+
'jondurbin/airoboros-l2-70b-gpt4-1.4.1': ModelType.IFT,
|
448 |
+
'jondurbin/airoboros-13b': ModelType.IFT,
|
449 |
+
'jondurbin/airoboros-33b-gpt4-1.4': ModelType.IFT,
|
450 |
+
'jondurbin/airoboros-33b-gpt4-1.2': ModelType.IFT,
|
451 |
+
'jondurbin/airoboros-65b-gpt4-1.2': ModelType.IFT,
|
452 |
+
'ariellee/SuperPlatty-30B': ModelType.IFT,
|
453 |
+
'danielhanchen/open_llama_3b_600bt_preview': ModelType.FT,
|
454 |
+
'cerebras/Cerebras-GPT-256M': ModelType.PT,
|
455 |
+
'cerebras/Cerebras-GPT-1.3B': ModelType.PT,
|
456 |
+
'cerebras/Cerebras-GPT-13B': ModelType.PT,
|
457 |
+
'cerebras/Cerebras-GPT-2.7B': ModelType.PT,
|
458 |
+
'cerebras/Cerebras-GPT-111M': ModelType.PT,
|
459 |
+
'cerebras/Cerebras-GPT-6.7B': ModelType.PT,
|
460 |
+
'Yhyu13/oasst-rlhf-2-llama-30b-7k-steps-hf': ModelType.RL,
|
461 |
+
'Yhyu13/llama-30B-hf-openassitant': ModelType.FT,
|
462 |
+
'NousResearch/Nous-Hermes-Llama2-13b': ModelType.IFT,
|
463 |
+
'NousResearch/Nous-Hermes-llama-2-7b': ModelType.IFT,
|
464 |
+
'NousResearch/Redmond-Puffin-13B': ModelType.IFT,
|
465 |
+
'NousResearch/Nous-Hermes-13b': ModelType.IFT,
|
466 |
+
'project-baize/baize-v2-7b': ModelType.IFT,
|
467 |
+
'project-baize/baize-v2-13b': ModelType.IFT,
|
468 |
+
'LLMs/WizardLM-13B-V1.0': ModelType.FT,
|
469 |
+
'LLMs/AlpacaGPT4-7B-elina': ModelType.FT,
|
470 |
+
'wenge-research/yayi-7b': ModelType.FT,
|
471 |
+
'wenge-research/yayi-7b-llama2': ModelType.FT,
|
472 |
+
'wenge-research/yayi-13b-llama2': ModelType.FT,
|
473 |
+
'yhyhy3/open_llama_7b_v2_med_instruct': ModelType.IFT,
|
474 |
+
'llama-anon/instruct-13b': ModelType.IFT,
|
475 |
+
'huggingtweets/jerma985': ModelType.FT,
|
476 |
+
'huggingtweets/gladosystem': ModelType.FT,
|
477 |
+
'huggingtweets/bladeecity-jerma985': ModelType.FT,
|
478 |
+
'huggyllama/llama-13b': ModelType.PT,
|
479 |
+
'huggyllama/llama-65b': ModelType.PT,
|
480 |
+
'FabbriSimo01/Facebook_opt_1.3b_Quantized': ModelType.PT,
|
481 |
+
'upstage/Llama-2-70b-instruct': ModelType.IFT,
|
482 |
+
'upstage/Llama-2-70b-instruct-1024': ModelType.IFT,
|
483 |
+
'upstage/llama-65b-instruct': ModelType.IFT,
|
484 |
+
'upstage/llama-30b-instruct-2048': ModelType.IFT,
|
485 |
+
'upstage/llama-30b-instruct': ModelType.IFT,
|
486 |
+
'WizardLM/WizardLM-13B-1.0': ModelType.IFT,
|
487 |
+
'WizardLM/WizardLM-13B-V1.1': ModelType.IFT,
|
488 |
+
'WizardLM/WizardLM-13B-V1.2': ModelType.IFT,
|
489 |
+
'WizardLM/WizardLM-30B-V1.0': ModelType.IFT,
|
490 |
+
'WizardLM/WizardCoder-15B-V1.0': ModelType.IFT,
|
491 |
+
'gpt2': ModelType.PT,
|
492 |
+
'keyfan/vicuna-chinese-replication-v1.1': ModelType.IFT,
|
493 |
+
'nthngdy/pythia-owt2-70m-100k': ModelType.FT,
|
494 |
+
'nthngdy/pythia-owt2-70m-50k': ModelType.FT,
|
495 |
+
'quantumaikr/KoreanLM-hf': ModelType.FT,
|
496 |
+
'quantumaikr/open_llama_7b_hf': ModelType.FT,
|
497 |
+
'quantumaikr/QuantumLM-70B-hf': ModelType.IFT,
|
498 |
+
'MayaPH/FinOPT-Lincoln': ModelType.FT,
|
499 |
+
'MayaPH/FinOPT-Franklin': ModelType.FT,
|
500 |
+
'MayaPH/GodziLLa-30B': ModelType.IFT,
|
501 |
+
'MayaPH/GodziLLa-30B-plus': ModelType.IFT,
|
502 |
+
'MayaPH/FinOPT-Washington': ModelType.FT,
|
503 |
+
'ogimgio/gpt-neo-125m-neurallinguisticpioneers': ModelType.FT,
|
504 |
+
'layoric/llama-2-13b-code-alpaca': ModelType.FT,
|
505 |
+
'CobraMamba/mamba-gpt-3b': ModelType.FT,
|
506 |
+
'CobraMamba/mamba-gpt-3b-v2': ModelType.FT,
|
507 |
+
'CobraMamba/mamba-gpt-3b-v3': ModelType.FT,
|
508 |
+
'timdettmers/guanaco-33b-merged': ModelType.FT,
|
509 |
+
'elinas/chronos-33b': ModelType.IFT,
|
510 |
+
'heegyu/RedTulu-Uncensored-3B-0719': ModelType.IFT,
|
511 |
+
'heegyu/WizardVicuna-Uncensored-3B-0719': ModelType.IFT,
|
512 |
+
'heegyu/WizardVicuna-3B-0719': ModelType.IFT,
|
513 |
+
'meta-llama/Llama-2-7b-chat-hf': ModelType.RL,
|
514 |
+
'meta-llama/Llama-2-7b-hf': ModelType.PT,
|
515 |
+
'meta-llama/Llama-2-13b-chat-hf': ModelType.RL,
|
516 |
+
'meta-llama/Llama-2-13b-hf': ModelType.PT,
|
517 |
+
'meta-llama/Llama-2-70b-chat-hf': ModelType.RL,
|
518 |
+
'meta-llama/Llama-2-70b-hf': ModelType.PT,
|
519 |
+
'xhyi/PT_GPTNEO350_ATG': ModelType.FT,
|
520 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-20b': ModelType.FT,
|
521 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-open-llama-7b-preview-400bt': ModelType.FT,
|
522 |
+
'h2oai/h2ogpt-oig-oasst1-512-6_9b': ModelType.IFT,
|
523 |
+
'h2oai/h2ogpt-oasst1-512-12b': ModelType.IFT,
|
524 |
+
'h2oai/h2ogpt-oig-oasst1-256-6_9b': ModelType.IFT,
|
525 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b-preview-300bt': ModelType.FT,
|
526 |
+
'h2oai/h2ogpt-oasst1-512-20b': ModelType.IFT,
|
527 |
+
'h2oai/h2ogpt-gm-oasst1-en-2048-open-llama-7b-preview-300bt-v2': ModelType.FT,
|
528 |
+
'h2oai/h2ogpt-gm-oasst1-en-1024-12b': ModelType.FT,
|
529 |
+
'h2oai/h2ogpt-gm-oasst1-multilang-1024-20b': ModelType.FT,
|
530 |
+
'bofenghuang/vigogne-13b-instruct': ModelType.IFT,
|
531 |
+
'bofenghuang/vigogne-13b-chat': ModelType.FT,
|
532 |
+
'bofenghuang/vigogne-2-7b-instruct': ModelType.IFT,
|
533 |
+
'bofenghuang/vigogne-7b-instruct': ModelType.IFT,
|
534 |
+
'bofenghuang/vigogne-7b-chat': ModelType.FT,
|
535 |
+
'Vmware/open-llama-7b-v2-open-instruct': ModelType.IFT,
|
536 |
+
'VMware/open-llama-0.7T-7B-open-instruct-v1.1': ModelType.IFT,
|
537 |
+
'ewof/koishi-instruct-3b': ModelType.IFT,
|
538 |
+
'gywy/llama2-13b-chinese-v1': ModelType.FT,
|
539 |
+
'GOAT-AI/GOAT-7B-Community': ModelType.FT,
|
540 |
+
'psyche/kollama2-7b': ModelType.FT,
|
541 |
+
'TheTravellingEngineer/llama2-7b-hf-guanaco': ModelType.FT,
|
542 |
+
'beaugogh/pythia-1.4b-deduped-sharegpt': ModelType.FT,
|
543 |
+
'augtoma/qCammel-70-x': ModelType.IFT,
|
544 |
+
'Lajonbot/Llama-2-7b-chat-hf-instruct-pl-lora_unload': ModelType.IFT,
|
545 |
+
'anhnv125/pygmalion-6b-roleplay': ModelType.FT,
|
546 |
+
'64bits/LexPodLM-13B': ModelType.FT,
|
547 |
}
|
548 |
|
549 |
|
550 |
def get_model_type(leaderboard_data: List[dict]):
|
551 |
for model_data in leaderboard_data:
|
552 |
# Todo @clefourrier once requests are connected with results
|
|
|
553 |
# Stored information
|
554 |
+
request_file = os.path.join("eval-queue", model_data["model_name_for_query"] + "_eval_request_*" + ".json")
|
555 |
+
request_file = glob.glob(request_file)
|
556 |
+
|
557 |
+
if len(request_file) == 0:
|
558 |
+
model_data[AutoEvalColumn.model_type.name] = ""
|
559 |
+
model_data[AutoEvalColumn.model_type_symbol.name] = ""
|
560 |
+
continue
|
561 |
+
|
562 |
+
request_file = request_file[0]
|
563 |
+
|
564 |
+
try:
|
565 |
+
with open(request_file, "r") as f:
|
566 |
+
request = json.load(f)
|
567 |
+
is_delta = request["weight_type"] != "Original"
|
568 |
+
except Exception:
|
569 |
+
is_delta = False
|
570 |
+
|
571 |
+
try:
|
572 |
+
with open(request_file, "r") as f:
|
573 |
+
request = json.load(f)
|
574 |
+
model_type = request["model_type"]
|
575 |
+
model_data[AutoEvalColumn.model_type.name] = model_type
|
576 |
+
model_data[AutoEvalColumn.model_type_symbol.name] = model_type_symbols[model_type] + ("🔺" if is_delta else "")
|
577 |
+
except Exception:
|
578 |
+
model_data[AutoEvalColumn.model_type.name] = "Unknown, add type to request file!"
|
579 |
+
model_data[AutoEvalColumn.model_type_symbol.name] = "?"
|
src/init.py
CHANGED
File without changes
|
src/utils_display.py
CHANGED
@@ -22,7 +22,7 @@ class AutoEvalColumn: # Auto evals column
|
|
22 |
mmlu = ColumnContent("MMLU", "number", True)
|
23 |
truthfulqa = ColumnContent("TruthfulQA", "number", True)
|
24 |
model_type = ColumnContent("Type", "str", False)
|
25 |
-
precision = ColumnContent("Precision", "str", False
|
26 |
license = ColumnContent("Hub License", "str", False)
|
27 |
params = ColumnContent("#Params (B)", "number", False)
|
28 |
likes = ColumnContent("Hub ❤️", "number", False)
|
@@ -43,7 +43,7 @@ class EvalQueueColumn: # Queue column
|
|
43 |
model = ColumnContent("model", "markdown", True)
|
44 |
revision = ColumnContent("revision", "str", True)
|
45 |
private = ColumnContent("private", "bool", True)
|
46 |
-
precision = ColumnContent("precision", "
|
47 |
weight_type = ColumnContent("weight_type", "str", "Original")
|
48 |
status = ColumnContent("status", "str", True)
|
49 |
|
|
|
22 |
mmlu = ColumnContent("MMLU", "number", True)
|
23 |
truthfulqa = ColumnContent("TruthfulQA", "number", True)
|
24 |
model_type = ColumnContent("Type", "str", False)
|
25 |
+
precision = ColumnContent("Precision", "str", False) #, True)
|
26 |
license = ColumnContent("Hub License", "str", False)
|
27 |
params = ColumnContent("#Params (B)", "number", False)
|
28 |
likes = ColumnContent("Hub ❤️", "number", False)
|
|
|
43 |
model = ColumnContent("model", "markdown", True)
|
44 |
revision = ColumnContent("revision", "str", True)
|
45 |
private = ColumnContent("private", "bool", True)
|
46 |
+
precision = ColumnContent("precision", "str", True)
|
47 |
weight_type = ColumnContent("weight_type", "str", "Original")
|
48 |
status = ColumnContent("status", "str", True)
|
49 |
|