Tom Aarsen
commited on
Commit
·
2f6ecfd
1
Parent(s):
4ee1aa5
Add "device_map": "auto" to automatically move the model to CUDA if possible
Browse files
README.md
CHANGED
|
@@ -72,7 +72,7 @@ model = SentenceTransformer("Qwen/Qwen3-Embedding-0.6B")
|
|
| 72 |
# together with setting `padding_side` to "left":
|
| 73 |
# model = SentenceTransformer(
|
| 74 |
# "Qwen/Qwen3-Embedding-0.6B",
|
| 75 |
-
# model_kwargs={"attn_implementation": "flash_attention_2"},
|
| 76 |
# tokenizer_kwargs={"padding_side": "left"},
|
| 77 |
# )
|
| 78 |
|
|
|
|
| 72 |
# together with setting `padding_side` to "left":
|
| 73 |
# model = SentenceTransformer(
|
| 74 |
# "Qwen/Qwen3-Embedding-0.6B",
|
| 75 |
+
# model_kwargs={"attn_implementation": "flash_attention_2", "device_map": "auto"},
|
| 76 |
# tokenizer_kwargs={"padding_side": "left"},
|
| 77 |
# )
|
| 78 |
|