yuchenglu commited on
Commit
a8043fd
•
1 Parent(s): b71cf75

Update README: add the details for data collections and fix typos

Browse files
Files changed (1) hide show
  1. README.md +26 -10
README.md CHANGED
@@ -11,24 +11,40 @@ datasets:
11
 
12
  ## Model Description
13
 
14
- LLaMA-2-7B-32K-Chat is an open-source, long-context chat model finetuned from [LLaMA-2-7B-32K](https://huggingface.co/togethercomputer/LLaMA-2-7B-32K), over high-quality instructions and chat data.
15
- We build Llama-2-7B-32K-Chat with less than 200 lines of Python script using [Together API](https://together.ai/blog/api-announcement), and we also make the recipe fully available.
 
16
  We hope that this can enable everyone to finetune their own version of [LLaMA-2-7B-32K](https://huggingface.co/togethercomputer/LLaMA-2-7B-32K) — play with [Together API](https://together.ai/blog/api-announcement) and give us feedback!
 
17
 
 
18
 
19
- ## Data Collection
20
-
21
- LLaMA-2-7B-32K-Chat is fine-tuned over two datasets: (1) 19K single- and multi-round conversations generated by human instructions and
22
- Llama-2-70B-Chat outputs, along with (2) 4K instructions of summarization from the BookSum datasets.
23
- We collected dataset (1) following the distillation paradigm that is used by Alpaca, Vicuna, WizardLM, Orca — producing instructions by querying a powerful LLM.
24
- We release such dataset [here](https://huggingface.co/datasets/togethercomputer/llama-instruct). For more details, please refer to our [Github repo](https://github.com/togethercomputer/LLaMA-2-32K-Chat).
 
 
 
 
25
 
26
 
27
  ## Model Usage
28
 
29
- You can use the [Together API](https://together.ai/blog/api-announcement) to try out LLaMA-2-7B-32K-Chat for inference.
30
  The updated inference stack allows for efficient inference.
31
- The model will also be hosted on [Together Playground](https://api.together.xyz/playground). You can simply play with the model by using prompt formatted by
 
 
 
 
 
 
 
 
 
32
 
33
  ```
34
  [INST] <your instruction here> [\INST].
 
11
 
12
  ## Model Description
13
 
14
+ LLaMA-2-7B-32K-Chat is an open-source, long-context chat model finetuned from [LLaMA-2-7B-32K](https://huggingface.co/togethercomputer/LLaMA-2-7B-32K), over high-quality instruction and chat data.
15
+ We built Llama-2-7B-32K-Chat with less than 200 lines of Python script using [Together API](https://together.ai/blog/api-announcement), and we also make the recipe fully available.
16
+ For more details, please refer to our [Github repo](https://github.com/togethercomputer/LLaMA-2-32K-Chat).
17
  We hope that this can enable everyone to finetune their own version of [LLaMA-2-7B-32K](https://huggingface.co/togethercomputer/LLaMA-2-7B-32K) — play with [Together API](https://together.ai/blog/api-announcement) and give us feedback!
18
+ For more details
19
 
20
+ ## Data Collection Details
21
 
22
+ LLaMA-2-7B-32K-Chat is fine-tuned over a combination of two parts:
23
+ 1. **19K single- and multi-round conversations generated by human instructions and [Llama-2-70B-Chat](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf) outputs**.
24
+ We collected the dataset following the distillation paradigm that is used by Alpaca, Vicuna, WizardLM, Orca — producing instructions by querying a powerful LLM.
25
+ This dataset is also released [here](https://huggingface.co/datasets/togethercomputer/llama-instruct).
26
+ We also share the complete collection recipe [here](https://github.com/togethercomputer/LLaMA-2-32K-Chat).
27
+
28
+ 3. **4K instructions of summarization from the BookSum datasets**.
29
+ BookSum is a unique dataset designed to address the challenges of long-form narrative summarization.
30
+ This dataset features source documents from the literature domain, including novels, plays, and stories, and offers human-written, highly abstractive summaries.
31
+ We here focus on chapter-level data. BookSum poses a unique set of challenges, necessitating that the model comprehensively read through each chapter.
32
 
33
 
34
  ## Model Usage
35
 
36
+ We encourage you to try out this model using the [Together API](https://together.ai/blog/api-announcement).
37
  The updated inference stack allows for efficient inference.
38
+ Alternatively, you can load the model directly from the Hugging Face model hub using
39
+
40
+ ```python
41
+ from transformers import AutoTokenizer, AutoModelForCausalLM
42
+
43
+ tokenizer = AutoTokenizer.from_pretrained("togethercomputer/LLaMA-2-7B-32K")
44
+ model = AutoModelForCausalLM.from_pretrained("togethercomputer/LLaMA-2-7B-32K", trust_remote_code=True, torch_dtype=torch.float16)
45
+ ```
46
+
47
+ The model is also hosted on [Together Playground](https://api.together.xyz/playground). You can simply play with the model by using prompt formatted by
48
 
49
  ```
50
  [INST] <your instruction here> [\INST].