Update README.md
Browse files
README.md
CHANGED
@@ -49,7 +49,7 @@ nDCG and Recall scores of this model(out-of-domain predictions) and other multil
|
|
49 |
alt="drawing"/>
|
50 |
|
51 |
## Usage
|
52 |
-
First install required libraries (Flash Attention supporting GPU is a must for consistency):
|
53 |
|
54 |
```bash
|
55 |
pip install -U einops flash_attn
|
@@ -71,7 +71,7 @@ from pylate import indexes, models, retrieve
|
|
71 |
|
72 |
# Step 1: Load the ColBERT model
|
73 |
model = models.ColBERT(
|
74 |
-
model_name_or_path=
|
75 |
)
|
76 |
try:
|
77 |
model.tokenizer.model_input_names.remove("token_type_ids")
|
|
|
49 |
alt="drawing"/>
|
50 |
|
51 |
## Usage
|
52 |
+
First install required libraries (Flash Attention 2 supporting GPU is a must for consistency otherwise you need to mask query expansion tokens in the output layer manually):
|
53 |
|
54 |
```bash
|
55 |
pip install -U einops flash_attn
|
|
|
71 |
|
72 |
# Step 1: Load the ColBERT model
|
73 |
model = models.ColBERT(
|
74 |
+
model_name_or_path="99eren99/TrColBERT",
|
75 |
)
|
76 |
try:
|
77 |
model.tokenizer.model_input_names.remove("token_type_ids")
|