dhanilka commited on
Commit
083a9c9
1 Parent(s): ba99897

Upload 2 files

Browse files
Files changed (2) hide show
  1. README.md +79 -3
  2. tokenizer.json +0 -0
README.md CHANGED
@@ -6,6 +6,82 @@ tags:
6
  inference:
7
  parameters:
8
  temperature: 0.7
9
- language:
10
- - en
11
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  inference:
7
  parameters:
8
  temperature: 0.7
9
+ ---
10
+
11
+ # Model Card for Mistral-7B-Instruct-v0.1
12
+
13
+ The Mistral-7B-Instruct-v0.1 Large Language Model (LLM) is a instruct fine-tuned version of the [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) generative text model using a variety of publicly available conversation datasets.
14
+
15
+ For full details of this model please read our [paper](https://arxiv.org/abs/2310.06825) and [release blog post](https://mistral.ai/news/announcing-mistral-7b/).
16
+
17
+ ## Instruction format
18
+
19
+ In order to leverage instruction fine-tuning, your prompt should be surrounded by `[INST]` and `[/INST]` tokens. The very first instruction should begin with a begin of sentence id. The next instructions should not. The assistant generation will be ended by the end-of-sentence token id.
20
+
21
+ E.g.
22
+ ```
23
+ text = "<s>[INST] What is your favourite condiment? [/INST]"
24
+ "Well, I'm quite partial to a good squeeze of fresh lemon juice. It adds just the right amount of zesty flavour to whatever I'm cooking up in the kitchen!</s> "
25
+ "[INST] Do you have mayonnaise recipes? [/INST]"
26
+ ```
27
+
28
+ This format is available as a [chat template](https://huggingface.co/docs/transformers/main/chat_templating) via the `apply_chat_template()` method:
29
+
30
+ ```python
31
+ from transformers import AutoModelForCausalLM, AutoTokenizer
32
+
33
+ device = "cuda" # the device to load the model onto
34
+
35
+ model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.1")
36
+ tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.1")
37
+
38
+ messages = [
39
+ {"role": "user", "content": "What is your favourite condiment?"},
40
+ {"role": "assistant", "content": "Well, I'm quite partial to a good squeeze of fresh lemon juice. It adds just the right amount of zesty flavour to whatever I'm cooking up in the kitchen!"},
41
+ {"role": "user", "content": "Do you have mayonnaise recipes?"}
42
+ ]
43
+
44
+ encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")
45
+
46
+ model_inputs = encodeds.to(device)
47
+ model.to(device)
48
+
49
+ generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
50
+ decoded = tokenizer.batch_decode(generated_ids)
51
+ print(decoded[0])
52
+ ```
53
+
54
+ ## Model Architecture
55
+ This instruction model is based on Mistral-7B-v0.1, a transformer model with the following architecture choices:
56
+ - Grouped-Query Attention
57
+ - Sliding-Window Attention
58
+ - Byte-fallback BPE tokenizer
59
+
60
+ ## Troubleshooting
61
+ - If you see the following error:
62
+ ```
63
+ Traceback (most recent call last):
64
+ File "", line 1, in
65
+ File "/transformers/models/auto/auto_factory.py", line 482, in from_pretrained
66
+ config, kwargs = AutoConfig.from_pretrained(
67
+ File "/transformers/models/auto/configuration_auto.py", line 1022, in from_pretrained
68
+ config_class = CONFIG_MAPPING[config_dict["model_type"]]
69
+ File "/transformers/models/auto/configuration_auto.py", line 723, in getitem
70
+ raise KeyError(key)
71
+ KeyError: 'mistral'
72
+ ```
73
+
74
+ Installing transformers from source should solve the issue
75
+ pip install git+https://github.com/huggingface/transformers
76
+
77
+ This should not be required after transformers-v4.33.4.
78
+
79
+ ## Limitations
80
+
81
+ The Mistral 7B Instruct model is a quick demonstration that the base model can be easily fine-tuned to achieve compelling performance.
82
+ It does not have any moderation mechanisms. We're looking forward to engaging with the community on ways to
83
+ make the model finely respect guardrails, allowing for deployment in environments requiring moderated outputs.
84
+
85
+ ## The Mistral AI Team
86
+
87
+ Albert Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, William El Sayed.
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff