update

README.md

@@ -1,12 +1,13 @@
 ---
-title: Dnabert2 Demo
-emoji: 🦀
-colorFrom: blue
-colorTo: gray
+title: DNABERT-2 demo
+emoji: 🐨
+colorFrom: yellow
+colorTo: green
 sdk: gradio
 sdk_version: 4.38.1
 app_file: app.py
 pinned: false
+license: bigscience-openrail-m
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,13 @@
+
+from transformers import AutoTokenizer, AutoModel, pipeline
+from transformers.models.bert.configuration_bert import BertConfig
+import gradio as gr
+
+config = BertConfig.from_pretrained("czl/dnabert2")
+tokenizer = AutoTokenizer.from_pretrained("czl/dnabert2", trust_remote_code=True)
+model = AutoModel.from_pretrained("czl/dnabert2", trust_remote_code=True, config=config)
+
+pipe = pipeline("feature-extraction", model=model, tokenizer=tokenizer)
+
+demo = gr.Interface.from_pipeline(pipe)
+demo.launch()

config.json

@@ -0,0 +1,33 @@
+{
+    "_name_or_path": "czl/dnabert2",
+    "alibi_starting_size": 512,
+    "architectures": [
+        "BertForMaskedLM"
+    ],
+    "attention_probs_dropout_prob": 0,
+    "auto_map": {
+        "AutoConfig": "configuration_bert.BertConfig",
+        "AutoModel": "bert_layers.BertModel",
+        "AutoModelForMaskedLM": "bert_layers.BertForMaskedLM",
+        "AutoModelForSequenceClassification": "bert_layers.BertForSequenceClassification"
+    },
+    "classifier_dropout": null,
+    "gradient_checkpointing": false,
+    "hidden_act": "gelu",
+    "hidden_dropout_prob": 0.1,
+    "hidden_size": 768,
+    "initializer_range": 0.02,
+    "intermediate_size": 3072,
+    "layer_norm_eps": 1e-12,
+    "max_position_embeddings": 512,
+    "model_type": "bert",
+    "num_attention_heads": 12,
+    "num_hidden_layers": 12,
+    "position_embedding_type": "absolute",
+    "torch_dtype": "float32",
+    "transformers_version": "4.28.0",
+    "type_vocab_size": 2,
+    "use_cache": true,
+    "vocab_size": 4096,
+    "flash_attn": false
+}

configuration_bert.py

@@ -0,0 +1,26 @@
+# Copyright 2022 MosaicML Examples authors
+# SPDX-License-Identifier: Apache-2.0
+
+from transformers import BertConfig as TransformersBertConfig
+
+
+class BertConfig(TransformersBertConfig):
+
+    def __init__(
+        self,
+        alibi_starting_size: int = 512,
+        attention_probs_dropout_prob: float = 0.0,
+        **kwargs,
+    ):
+        """Configuration class for MosaicBert.
+
+        Args:
+            alibi_starting_size (int): Use `alibi_starting_size` to determine how large of an alibi tensor to
+                create when initializing the model. You should be able to ignore this parameter in most cases.
+                Defaults to 512.
+            attention_probs_dropout_prob (float): By default, turn off attention dropout in Mosaic BERT
+                (otherwise, Flash Attention will be off by default). Defaults to 0.0.
+        """
+        super().__init__(
+            attention_probs_dropout_prob=attention_probs_dropout_prob, **kwargs)
+        self.alibi_starting_size = alibi_starting_size

requirements.txt

@@ -0,0 +1,11 @@
+transformers[torch]
+torch
+torchvision
+torchaudio
+einops
+peft
+omegaconf
+evaluate
+accelerate
+gradio
+spaces