Update README.md
Browse files
README.md
CHANGED
@@ -96,7 +96,7 @@ transformers==4.37.2
|
|
96 |
from transformers import AutoModelForSequenceClassification
|
97 |
import torch
|
98 |
|
99 |
-
model_name = "
|
100 |
model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True, torch_dtype=torch.float16).to("cuda")
|
101 |
# You can also use the following code to use flash_attention_2
|
102 |
# model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True,attn_implementation="flash_attention_2", torch_dtype=torch.float16).to("cuda")
|
@@ -121,7 +121,7 @@ from sentence_transformers import CrossEncoder
|
|
121 |
from transformers import LlamaTokenizer
|
122 |
import torch
|
123 |
|
124 |
-
model_name = "
|
125 |
model = CrossEncoder(model_name,max_length=1024,trust_remote_code=True, automodel_args={"torch_dtype": torch.float16})
|
126 |
# You can also use the following code to use flash_attention_2
|
127 |
#model = CrossEncoder(model_name,max_length=1024,trust_remote_code=True, automodel_args={"attn_implementation":"flash_attention_2","torch_dtype": torch.float16})
|
@@ -158,7 +158,7 @@ INSTRUCTION = "Query:"
|
|
158 |
query = f"{INSTRUCTION} {query}"
|
159 |
|
160 |
array = AsyncEngineArray.from_args(
|
161 |
-
[EngineArgs(model_name_or_path = "
|
162 |
)
|
163 |
|
164 |
async def rerank(engine: AsyncEmbeddingEngine):
|
@@ -173,7 +173,7 @@ asyncio.run(rerank(array[0])) # [(RerankReturnType(relevance_score=0.017917344,
|
|
173 |
|
174 |
```python
|
175 |
from FlagEmbedding import FlagReranker
|
176 |
-
model_name = "
|
177 |
model = FlagReranker(model_name, use_fp16=True, query_instruction_for_rerank="Query: ", trust_remote_code=True)
|
178 |
# You can hack the __init__() method of the FlagEmbedding BaseReranker class to use flash_attention_2 for faster inference
|
179 |
# self.model = AutoModelForSequenceClassification.from_pretrained(
|
@@ -231,10 +231,10 @@ We re-rank top-100 documents from `bge-m3` (Dense).
|
|
231 |
|
232 |
## 许可证 License
|
233 |
|
234 |
-
- 本仓库中代码依照 [Apache-2.0 协议](https://github.com/
|
235 |
-
- MiniCPM-Reranker-Light 模型权重的使用则需要遵循 [MiniCPM 模型协议](https://github.com/
|
236 |
- MiniCPM-Reranker-Light 模型权重对学术研究完全开放。如需将模型用于商业用途,请填写[此问卷](https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g)。
|
237 |
|
238 |
-
* The code in this repo is released under the [Apache-2.0](https://github.com/
|
239 |
-
* The usage of MiniCPM-Reranker-Light model weights must strictly follow [MiniCPM Model License.md](https://github.com/
|
240 |
* The models and weights of MiniCPM-Reranker-Light are completely free for academic research. After filling out a ["questionnaire"](https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g) for registration, MiniCPM-Reranker-Light weights are also available for free commercial use.
|
|
|
96 |
from transformers import AutoModelForSequenceClassification
|
97 |
import torch
|
98 |
|
99 |
+
model_name = "openbmb/MiniCPM-Reranker-Light"
|
100 |
model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True, torch_dtype=torch.float16).to("cuda")
|
101 |
# You can also use the following code to use flash_attention_2
|
102 |
# model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True,attn_implementation="flash_attention_2", torch_dtype=torch.float16).to("cuda")
|
|
|
121 |
from transformers import LlamaTokenizer
|
122 |
import torch
|
123 |
|
124 |
+
model_name = "openbmb/MiniCPM-Reranker-Light"
|
125 |
model = CrossEncoder(model_name,max_length=1024,trust_remote_code=True, automodel_args={"torch_dtype": torch.float16})
|
126 |
# You can also use the following code to use flash_attention_2
|
127 |
#model = CrossEncoder(model_name,max_length=1024,trust_remote_code=True, automodel_args={"attn_implementation":"flash_attention_2","torch_dtype": torch.float16})
|
|
|
158 |
query = f"{INSTRUCTION} {query}"
|
159 |
|
160 |
array = AsyncEngineArray.from_args(
|
161 |
+
[EngineArgs(model_name_or_path = "openbmb/MiniCPM-Reranker-Light", engine="torch", dtype="float16", bettertransformer=False, trust_remote_code=True, model_warmup=False)]
|
162 |
)
|
163 |
|
164 |
async def rerank(engine: AsyncEmbeddingEngine):
|
|
|
173 |
|
174 |
```python
|
175 |
from FlagEmbedding import FlagReranker
|
176 |
+
model_name = "openbmb/MiniCPM-Reranker-Light"
|
177 |
model = FlagReranker(model_name, use_fp16=True, query_instruction_for_rerank="Query: ", trust_remote_code=True)
|
178 |
# You can hack the __init__() method of the FlagEmbedding BaseReranker class to use flash_attention_2 for faster inference
|
179 |
# self.model = AutoModelForSequenceClassification.from_pretrained(
|
|
|
231 |
|
232 |
## 许可证 License
|
233 |
|
234 |
+
- 本仓库中代码依照 [Apache-2.0 协议](https://github.com/openbmb/MiniCPM/blob/main/LICENSE)开源。
|
235 |
+
- MiniCPM-Reranker-Light 模型权重的使用则需要遵循 [MiniCPM 模型协议](https://github.com/openbmb/MiniCPM/blob/main/MiniCPM%20Model%20License.md)。
|
236 |
- MiniCPM-Reranker-Light 模型权重对学术研究完全开放。如需将模型用于商业用途,请填写[此问卷](https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g)。
|
237 |
|
238 |
+
* The code in this repo is released under the [Apache-2.0](https://github.com/openbmb/MiniCPM/blob/main/LICENSE) License.
|
239 |
+
* The usage of MiniCPM-Reranker-Light model weights must strictly follow [MiniCPM Model License.md](https://github.com/openbmb/MiniCPM/blob/main/MiniCPM%20Model%20License.md).
|
240 |
* The models and weights of MiniCPM-Reranker-Light are completely free for academic research. After filling out a ["questionnaire"](https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g) for registration, MiniCPM-Reranker-Light weights are also available for free commercial use.
|