Update README.md
Browse files
README.md
CHANGED
@@ -1,33 +1,65 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
base_model: openchat/openchat_3.5
|
|
|
|
|
|
|
|
|
4 |
tags:
|
5 |
- generated_from_trainer
|
|
|
6 |
model-index:
|
7 |
- name: smol-7b
|
8 |
results: []
|
9 |
---
|
10 |
|
11 |
-
|
12 |
-
should probably proofread and complete it, then remove this comment. -->
|
13 |
|
14 |
-
|
15 |
|
16 |
-
|
17 |
-
It achieves the following results on the evaluation set:
|
18 |
-
- Loss: 2.0409
|
19 |
|
20 |
-
|
21 |
|
22 |
-
|
23 |
|
24 |
-
|
25 |
|
26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
|
28 |
-
##
|
29 |
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
|
32 |
## Training procedure
|
33 |
|
@@ -58,3 +90,16 @@ The following hyperparameters were used during training:
|
|
58 |
- Pytorch 2.1.1+cu121
|
59 |
- Datasets 2.14.6
|
60 |
- Tokenizers 0.14.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
base_model: openchat/openchat_3.5
|
4 |
+
datasets:
|
5 |
+
- HuggingFaceH4/no_robots
|
6 |
+
language:
|
7 |
+
- en
|
8 |
tags:
|
9 |
- generated_from_trainer
|
10 |
+
pipeline_tag: text-generation
|
11 |
model-index:
|
12 |
- name: smol-7b
|
13 |
results: []
|
14 |
---
|
15 |
|
16 |
+
# Smol 7B
|
|
|
17 |
|
18 |
+
This model is a fine-tuned version of [openchat/openchat_3.5](https://huggingface.co/openchat/openchat_3.5) on the open source dataset [HuggingFaceH4/no_robots](https://huggingface.co/datasets/HuggingFaceH4/no_robots) using the recipes published in [The Alignment Handbook](https://github.com/huggingface/alignment-handbook).
|
19 |
|
20 |
+
## Model date
|
|
|
|
|
21 |
|
22 |
+
rishiraj/smol-7b was trained between 1st and 3rd December, 2023.
|
23 |
|
24 |
+
## Evaluation
|
25 |
|
26 |
+
It achieves the following results on the [Open_LLM_Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard). At the time of release, smol-7b is the highest ranked 7B chat model on the [MMLU Benchmark](https://paperswithcode.com/sota/multi-task-language-understanding-on-mmlu).
|
27 |
|
28 |
+
| Model | Average | ARC | HellaSwag | MMLU | TruthfulQA | Winogrande | GSM8K |
|
29 |
+
| ---------------------------- | ------- | ----- | --------- | ----- | ---------- | ---------- | ----- |
|
30 |
+
| rishiraj/smol-7b | 67.11 | 63.74 | 84.77 | 65 | 46.17 | 80.66 | 62.32 |
|
31 |
+
| argilla/notus-7b-v1 | 63.49 | 64.59 | 84.83 | 63.04 | 54.35 | 79.56 | 34.57 |
|
32 |
+
| Intel/neural-chat-7b-v3-1 | 61.59 | 66.21 | 83.64 | 62.37 | 59.65 | 78.14 | 19.56 |
|
33 |
+
| HuggingFaceH4/zephyr-7b-beta | 61.59 | 62.46 | 84.35 | 60.7 | 57.83 | 77.11 | 27.07 |
|
34 |
+
| Qwen/Qwen-7B | 59.19 | 51.37 | 78.47 | 59.84 | 47.79 | 72.69 | 44.96 |
|
35 |
+
| microsoft/Orca-2-7b | 54.55 | 54.1 | 76.19 | 56.37 | 52.45 | 73.48 | 14.71 |
|
36 |
+
| 01-ai/Yi-6B | 54.08 | 55.55 | 76.57 | 64.11 | 41.96 | 74.19 | 12.13 |
|
37 |
|
38 |
+
## Inference procedure
|
39 |
|
40 |
+
Here's how you can run the model using the pipeline() function from 🤗 Transformers:
|
41 |
+
|
42 |
+
```
|
43 |
+
import torch
|
44 |
+
from transformers import pipeline
|
45 |
+
|
46 |
+
pipe = pipeline("text-generation", model="rishiraj/smol-7b", torch_dtype=torch.bfloat16, device_map="auto")
|
47 |
+
|
48 |
+
# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
|
49 |
+
messages = [
|
50 |
+
{
|
51 |
+
"role": "system",
|
52 |
+
"content": "You are a friendly chatbot who always responds in the style of a pirate"
|
53 |
+
},
|
54 |
+
{
|
55 |
+
"role": "user",
|
56 |
+
"content": "How many helicopters can a human eat in one sitting?"
|
57 |
+
}
|
58 |
+
]
|
59 |
+
prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
60 |
+
outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
61 |
+
print(outputs[0]["generated_text"])
|
62 |
+
```
|
63 |
|
64 |
## Training procedure
|
65 |
|
|
|
90 |
- Pytorch 2.1.1+cu121
|
91 |
- Datasets 2.14.6
|
92 |
- Tokenizers 0.14.1
|
93 |
+
|
94 |
+
## Citation Information
|
95 |
+
|
96 |
+
```
|
97 |
+
@misc{rishiraj2023smol,
|
98 |
+
author = {Rishiraj Acharya},
|
99 |
+
title = {Smol 7B},
|
100 |
+
year = {2023},
|
101 |
+
publisher = {Hugging Face},
|
102 |
+
journal = {Hugging Face repository},
|
103 |
+
howpublished = {\url{https://huggingface.co/rishiraj/smol-7b}}
|
104 |
+
}
|
105 |
+
```
|