mgalkin commited on
Commit
a17cb10
1 Parent(s): 194a130

updated readme

Browse files
Files changed (1) hide show
  1. README.md +16 -2
README.md CHANGED
@@ -30,11 +30,11 @@ Here on HuggingFace, we provide 3 pre-trained ULTRA checkpoints (all ~169k param
30
 
31
  ## ⚡️ Your Superpowers
32
 
33
- ULTRA performs **link prediction** (KG completion): given a query `(head, relation, ?)`, it ranks all nodes in the graph as potential `tails`.
34
 
35
 
36
  1. Install the dependencies as listed in the Installation instructions on the [GitHub repo](https://github.com/DeepGraphLearning/ULTRA#installation).
37
- 2. Clone this model repo to find the `UltraLinkPrediction` class in `modeling.py` and load the checkpoint (all the necessary model code is in this model repo as well).
38
 
39
  * Run **zero-shot inference** on any graph:
40
 
@@ -50,6 +50,20 @@ test(model, mode="test", dataset=dataset, gpus=None)
50
  # hits@10: 0.666
51
  ```
52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53
  * You can also **fine-tune** ULTRA on each graph, please refer to the [github repo](https://github.com/DeepGraphLearning/ULTRA#run-inference-and-fine-tuning) for more details on training / fine-tuning
54
  * The model code contains 57 different KGs, please refer to the [github repo](https://github.com/DeepGraphLearning/ULTRA#datasets) for more details on what's available.
55
 
 
30
 
31
  ## ⚡️ Your Superpowers
32
 
33
+ ULTRA performs **link prediction** (KG completion aka reasoning): given a query `(head, relation, ?)`, it ranks all nodes in the graph as potential `tails`.
34
 
35
 
36
  1. Install the dependencies as listed in the Installation instructions on the [GitHub repo](https://github.com/DeepGraphLearning/ULTRA#installation).
37
+ 2. Clone this model repo to find the `UltraForKnowledgeGraphReasoning` class in `modeling.py` and load the checkpoint (all the necessary model code is in this model repo as well).
38
 
39
  * Run **zero-shot inference** on any graph:
40
 
 
50
  # hits@10: 0.666
51
  ```
52
 
53
+ Or with `AutoModel`:
54
+
55
+ ```python
56
+ from transformers import AutoModel
57
+ from ultra.datasets import CoDExSmall
58
+ from ultra.eval import test
59
+ model = AutoModel.from_pretrained("mgalkin/ultra_4g", trust_remote_code=True)
60
+ dataset = CoDExSmall(root="./datasets/")
61
+ test(model, mode="test", dataset=dataset, gpus=None)
62
+ # Expected results for ULTRA 4g
63
+ # mrr: 0.464
64
+ # hits@10: 0.666
65
+ ```
66
+
67
  * You can also **fine-tune** ULTRA on each graph, please refer to the [github repo](https://github.com/DeepGraphLearning/ULTRA#run-inference-and-fine-tuning) for more details on training / fine-tuning
68
  * The model code contains 57 different KGs, please refer to the [github repo](https://github.com/DeepGraphLearning/ULTRA#datasets) for more details on what's available.
69