Update README.md
Browse files
README.md
CHANGED
@@ -11,13 +11,17 @@ tags:
|
|
11 |
Arctic is a Dense-MoE Hybrid transformer architecture pre-trained from scratch by the Snowflake AI
|
12 |
Research Team. We are releasing model checkpoints for both the base and instruct-tuned versions of
|
13 |
Arctic under an Apache-2.0 license. This means you can use them freely in your own research,
|
14 |
-
prototypes, and products. Please see our blog
|
|
|
15 |
for more information on Arctic and links to other relevant resources such as our series of cookbooks
|
16 |
covering topics around training your own custom MoE models, how to produce high-quality training data,
|
17 |
and much more.
|
18 |
|
19 |
-
* [Arctic-Base](
|
20 |
-
* [Acrtic-Instruct](
|
|
|
|
|
|
|
21 |
|
22 |
**Model developers** Snowflake
|
23 |
|
@@ -43,49 +47,19 @@ model implementation. Until this support is released please follow these instruc
|
|
43 |
required dependencies for using Arctic:
|
44 |
|
45 |
```python
|
46 |
-
pip install git+https://github.com/Snowflake-Labs/transformers.git
|
47 |
```
|
48 |
|
49 |
Arctic leverages several features from [DeepSpeed](https://github.com/microsoft/DeepSpeed), you will need to
|
50 |
install the latest version of DeepSpeed to get all of these required features:
|
51 |
|
52 |
```python
|
53 |
-
pip install "deepspeed>=0.
|
54 |
```
|
55 |
|
56 |
### Inference
|
57 |
|
58 |
-
|
59 |
-
can also use `transformers` to load
|
60 |
-
the model for text generation. Due to the model size we recommend using a single 8xH100 instance from your
|
61 |
-
favorite cloud provider such as: AWS [p5.48xlarge](https://aws.amazon.com/ec2/instance-types/p5/),
|
62 |
-
Azure [ND96isr_H100_v5](https://learn.microsoft.com/en-us/azure/virtual-machines/nd-h100-v5-series), etc.
|
63 |
-
|
64 |
-
In addition, if you would like to access Acrtic via API we have colloborated with several inference API
|
65 |
-
providers to host Acrtic such as AWS, Microsoft Azure, NVIDIA Foundry, Lamini, Perplexity, Replicate and Together.
|
66 |
-
|
67 |
-
```python
|
68 |
-
import torch
|
69 |
-
from transformers import AutoTokenizer, AutoModelForCausalLM
|
70 |
-
|
71 |
-
tokenizer = AutoTokenizer.from_pretrained("snowflake/arctic")
|
72 |
-
model = AutoModelForCausalLM.from_pretrained("snowflake/arctic", device_map="auto", torch_dtype=torch.bfloat16)
|
73 |
-
|
74 |
-
input_text = "Hello my name is "
|
75 |
-
input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")
|
76 |
-
|
77 |
-
outputs = model.generate(**input_ids, max_new_tokens=20)
|
78 |
-
print(tokenizer.decode(outputs[0]))
|
79 |
-
```
|
80 |
-
|
81 |
-
### Fine-Tuning
|
82 |
-
|
83 |
-
TODO: add link and extra details about fine-tuning scripts
|
84 |
-
|
85 |
-
## Metrics
|
86 |
-
|
87 |
-
TODO: add summary of metrics here, we don't necessarily need to compare to others but we can if we want
|
88 |
-
|
89 |
-
## Training Data
|
90 |
|
91 |
-
|
|
|
|
11 |
Arctic is a Dense-MoE Hybrid transformer architecture pre-trained from scratch by the Snowflake AI
|
12 |
Research Team. We are releasing model checkpoints for both the base and instruct-tuned versions of
|
13 |
Arctic under an Apache-2.0 license. This means you can use them freely in your own research,
|
14 |
+
prototypes, and products. Please see our blog
|
15 |
+
[Snowflake Arctic: The Best LLM for Enterprise AI — Efficiently Intelligent, Truly Open](https://www.snowflake.com/blog/arctic-open-and-efficient-foundation-language-models-snowflake)
|
16 |
for more information on Arctic and links to other relevant resources such as our series of cookbooks
|
17 |
covering topics around training your own custom MoE models, how to produce high-quality training data,
|
18 |
and much more.
|
19 |
|
20 |
+
* [Arctic-Base](https://huggingface.co/Snowflake/snowflake-arctic-base/)
|
21 |
+
* [Acrtic-Instruct](https://huggingface.co/Snowflake/snowflake-arctic-instruct/)
|
22 |
+
|
23 |
+
For the latest details about Snowflake Arctic including tutorials, etc. please refer to our github repo:
|
24 |
+
* https://github.com/Snowflake-Labs/snowflake-arctic
|
25 |
|
26 |
**Model developers** Snowflake
|
27 |
|
|
|
47 |
required dependencies for using Arctic:
|
48 |
|
49 |
```python
|
50 |
+
pip install git+https://github.com/Snowflake-Labs/transformers.git@arctic
|
51 |
```
|
52 |
|
53 |
Arctic leverages several features from [DeepSpeed](https://github.com/microsoft/DeepSpeed), you will need to
|
54 |
install the latest version of DeepSpeed to get all of these required features:
|
55 |
|
56 |
```python
|
57 |
+
pip install "deepspeed>=0.14.2"
|
58 |
```
|
59 |
|
60 |
### Inference
|
61 |
|
62 |
+
The Arctic github page has several resources around running inference:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
+
* Example with pure-HF: https://github.com/Snowflake-Labs/snowflake-arctic/blob/main/inference
|
65 |
+
* Tutorial using vLLM: https://github.com/Snowflake-Labs/snowflake-arctic/tree/main/inference/vllm
|