update readme

README.md

@@ -23,23 +23,9 @@ It achieves the following results on the evaluation set:
 - Question: {'precision': 0.8211009174311926, 'recall': 0.8403755868544601, 'f1': 0.8306264501160092, 'number': 1065}
 - Overall Precision: 0.7599
 - Overall Recall: 0.8083
-- Overall F1: 0.7834
+- Overall F1: 0.7866
 - Overall Accuracy: 0.8106
 
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
-
-## Training procedure
-
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
@@ -52,30 +38,144 @@ The following hyperparameters were used during training:
 - num_epochs: 15
 - mixed_precision_training: Native AMP
 
-### Training results
-
-| Training Loss | Epoch | Step | Validation Loss | Answer | Header | Question | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy |
-|:-------------:|:-----:|:----:|:---------------:|:--------------------------------------------------------------------------------------------------------:|:----------------------------------------------------------------------------------------------------------:|:---------------------------------------------------------------------------------------------------------:|:-----------------:|:--------------:|:----------:|:----------------:|
-| 0.1724 | 1.0 | 10 | 0.7657 | {'precision': 0.7097826086956521, 'recall': 0.8071693448702101, 'f1': 0.7553499132446501, 'number': 809} | {'precision': 0.3893129770992366, 'recall': 0.42857142857142855, 'f1': 0.40800000000000003, 'number': 119} | {'precision': 0.7941176470588235, 'recall': 0.8366197183098592, 'f1': 0.8148148148148148, 'number': 1065} | 0.7340 | 0.8003 | 0.7657 | 0.8134 |
-| 0.1451 | 2.0 | 20 | 0.8099 | {'precision': 0.7136659436008677, 'recall': 0.8133498145859085, 'f1': 0.7602541883304449, 'number': 809} | {'precision': 0.4215686274509804, 'recall': 0.36134453781512604, 'f1': 0.3891402714932127, 'number': 119} | {'precision': 0.809437386569873, 'recall': 0.8375586854460094, 'f1': 0.823257960313798, 'number': 1065} | 0.7493 | 0.7993 | 0.7735 | 0.8125 |
-| 0.1179 | 3.0 | 30 | 0.8622 | {'precision': 0.7099892588614393, 'recall': 0.8170580964153276, 'f1': 0.7597701149425288, 'number': 809} | {'precision': 0.4074074074074074, 'recall': 0.46218487394957986, 'f1': 0.4330708661417323, 'number': 119} | {'precision': 0.8123300090661831, 'recall': 0.8413145539906103, 'f1': 0.8265682656826567, 'number': 1065} | 0.7432 | 0.8088 | 0.7746 | 0.8074 |
-| 0.0988 | 4.0 | 40 | 0.8587 | {'precision': 0.7141327623126338, 'recall': 0.8244746600741656, 'f1': 0.7653471026965003, 'number': 809} | {'precision': 0.4166666666666667, 'recall': 0.5042016806722689, 'f1': 0.4562737642585551, 'number': 119} | {'precision': 0.8370998116760828, 'recall': 0.8347417840375587, 'f1': 0.8359191349318289, 'number': 1065} | 0.7551 | 0.8108 | 0.7820 | 0.8157 |
-| 0.0848 | 5.0 | 50 | 0.8933 | {'precision': 0.7255813953488373, 'recall': 0.7713226205191595, 'f1': 0.7477531455961653, 'number': 809} | {'precision': 0.4024390243902439, 'recall': 0.5546218487394958, 'f1': 0.46643109540636046, 'number': 119} | {'precision': 0.8201834862385321, 'recall': 0.8394366197183099, 'f1': 0.8296983758700696, 'number': 1065} | 0.7493 | 0.7948 | 0.7714 | 0.8056 |
-| 0.073 | 6.0 | 60 | 0.9009 | {'precision': 0.7344444444444445, 'recall': 0.8170580964153276, 'f1': 0.7735517846693973, 'number': 809} | {'precision': 0.41721854304635764, 'recall': 0.5294117647058824, 'f1': 0.4666666666666667, 'number': 119} | {'precision': 0.8107370336669699, 'recall': 0.8366197183098592, 'f1': 0.8234750462107209, 'number': 1065} | 0.7512 | 0.8103 | 0.7796 | 0.8123 |
-| 0.0655 | 7.0 | 70 | 0.9117 | {'precision': 0.7367231638418079, 'recall': 0.8059332509270705, 'f1': 0.769775678866588, 'number': 809} | {'precision': 0.4357142857142857, 'recall': 0.5126050420168067, 'f1': 0.47104247104247104, 'number': 119} | {'precision': 0.8170955882352942, 'recall': 0.8347417840375587, 'f1': 0.8258244310264746, 'number': 1065} | 0.7582 | 0.8038 | 0.7803 | 0.8088 |
-| 0.0599 | 8.0 | 80 | 0.9414 | {'precision': 0.7298474945533769, 'recall': 0.8281829419035847, 'f1': 0.7759119861030689, 'number': 809} | {'precision': 0.41496598639455784, 'recall': 0.5126050420168067, 'f1': 0.4586466165413534, 'number': 119} | {'precision': 0.8100810081008101, 'recall': 0.8450704225352113, 'f1': 0.8272058823529411, 'number': 1065} | 0.7495 | 0.8184 | 0.7824 | 0.8089 |
-| 0.0551 | 9.0 | 90 | 0.9548 | {'precision': 0.746031746031746, 'recall': 0.8133498145859085, 'f1': 0.7782377291543465, 'number': 809} | {'precision': 0.42953020134228187, 'recall': 0.5378151260504201, 'f1': 0.47761194029850745, 'number': 119} | {'precision': 0.823963133640553, 'recall': 0.8394366197183099, 'f1': 0.8316279069767442, 'number': 1065} | 0.7637 | 0.8108 | 0.7866 | 0.8111 |
-| 0.0483 | 10.0 | 100 | 0.9684 | {'precision': 0.7390326209223848, 'recall': 0.8121137206427689, 'f1': 0.773851590106007, 'number': 809} | {'precision': 0.42, 'recall': 0.5294117647058824, 'f1': 0.46840148698884754, 'number': 119} | {'precision': 0.8232044198895028, 'recall': 0.8394366197183099, 'f1': 0.8312412831241283, 'number': 1065} | 0.7595 | 0.8098 | 0.7839 | 0.8091 |
-| 0.0424 | 11.0 | 110 | 0.9858 | {'precision': 0.7392290249433107, 'recall': 0.8059332509270705, 'f1': 0.7711413364872857, 'number': 809} | {'precision': 0.4258064516129032, 'recall': 0.5546218487394958, 'f1': 0.48175182481751827, 'number': 119} | {'precision': 0.8252788104089219, 'recall': 0.8338028169014085, 'f1': 0.8295189163942083, 'number': 1065} | 0.7601 | 0.8058 | 0.7823 | 0.8094 |
-| 0.0402 | 12.0 | 120 | 0.9920 | {'precision': 0.7315436241610739, 'recall': 0.8084054388133498, 'f1': 0.7680563711098063, 'number': 809} | {'precision': 0.4460431654676259, 'recall': 0.5210084033613446, 'f1': 0.48062015503875966, 'number': 119} | {'precision': 0.8205128205128205, 'recall': 0.8413145539906103, 'f1': 0.8307834955957348, 'number': 1065} | 0.7586 | 0.8088 | 0.7829 | 0.8111 |
-| 0.0392 | 13.0 | 130 | 1.0027 | {'precision': 0.7463193657984145, 'recall': 0.8145859085290482, 'f1': 0.7789598108747045, 'number': 809} | {'precision': 0.4397163120567376, 'recall': 0.5210084033613446, 'f1': 0.47692307692307695, 'number': 119} | {'precision': 0.8216911764705882, 'recall': 0.8394366197183099, 'f1': 0.8304691128657686, 'number': 1065} | 0.7647 | 0.8103 | 0.7868 | 0.8104 |
-| 0.0361 | 14.0 | 140 | 1.0027 | {'precision': 0.7421171171171171, 'recall': 0.8145859085290482, 'f1': 0.7766647024160284, 'number': 809} | {'precision': 0.43884892086330934, 'recall': 0.5126050420168067, 'f1': 0.4728682170542636, 'number': 119} | {'precision': 0.8205128205128205, 'recall': 0.8413145539906103, 'f1': 0.8307834955957348, 'number': 1065} | 0.7626 | 0.8108 | 0.7860 | 0.8115 |
-| 0.0349 | 15.0 | 150 | 1.0045 | {'precision': 0.7348314606741573, 'recall': 0.8084054388133498, 'f1': 0.7698646262507357, 'number': 809} | {'precision': 0.44285714285714284, 'recall': 0.5210084033613446, 'f1': 0.47876447876447875, 'number': 119} | {'precision': 0.8211009174311926, 'recall': 0.8403755868544601, 'f1': 0.8306264501160092, 'number': 1065} | 0.7599 | 0.8083 | 0.7834 | 0.8106 |
-
-
-### Framework versions
-
-- Transformers 4.21.2
-- Pytorch 1.11.0+cu113
-- Datasets 2.5.1
-- Tokenizers 0.12.1
+## Deploy Model with Inference Endpoints
+
+Before we can get started, make sure you meet all of the following requirements:
+
+1. An Organization/User with an active plan and *WRITE* access to the model repository.
+2. Can access the UI: [https://ui.endpoints.huggingface.co](https://ui.endpoints.huggingface.co/endpoints)
+
+
+
+### 1. Deploy LayoutLM and Send requests
+
+In this tutorial, you will learn how to deploy a [LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm) to [Hugging Face Inference Endpoints](https://huggingface.co/inference-endpoints) and how you can integrate it via an API into your products. 
+
+This tutorial is not covering how you create the custom handler for inference. If you want to learn how to create a custom Handler for Inference Endpoints, you can either checkout the [documentation](https://huggingface.co/docs/inference-endpoints/guides/custom_handler) or go through [“Custom Inference with Hugging Face Inference Endpoints”](https://www.philschmid.de/custom-inference-handler) 
+
+We are going to deploy [philschmid/layoutlm-funsd](https://huggingface.co/philschmid/layoutlm-funsd) which implements the following `handler.py` 
+
+```python
+from typing import Dict, List, Any
+from transformers import LayoutLMForTokenClassification, LayoutLMv2Processor
+import torch
+from subprocess import run
+
+# install tesseract-ocr and pytesseract
+run("apt install -y tesseract-ocr", shell=True, check=True)
+run("pip install pytesseract", shell=True, check=True)
+
+# helper function to unnormalize bboxes for drawing onto the image
+def unnormalize_box(bbox, width, height):
+ return [
+ width * (bbox[0] / 1000),
+ height * (bbox[1] / 1000),
+ width * (bbox[2] / 1000),
+ height * (bbox[3] / 1000),
+ ]
+
+# set device
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+class EndpointHandler:
+ def __init__(self, path=""):
+ # load model and processor from path
+ self.model = LayoutLMForTokenClassification.from_pretrained(path).to(device)
+ self.processor = LayoutLMv2Processor.from_pretrained(path)
+
+ def __call__(self, data: Dict[str, bytes]) -> Dict[str, List[Any]]:
+ """
+ Args:
+ data (:obj:):
+ includes the deserialized image file as PIL.Image
+ """
+ # process input
+ image = data.pop("inputs", data)
+
+ # process image
+ encoding = self.processor(image, return_tensors="pt")
+
+ # run prediction
+ with torch.inference_mode():
+ outputs = self.model(
+ input_ids=encoding.input_ids.to(device),
+ bbox=encoding.bbox.to(device),
+ attention_mask=encoding.attention_mask.to(device),
+ token_type_ids=encoding.token_type_ids.to(device),
+ )
+ predictions = outputs.logits.softmax(-1)
+
+ # post process output
+ result = []
+ for item, inp_ids, bbox in zip(
+ predictions.squeeze(0).cpu(), encoding.input_ids.squeeze(0).cpu(), encoding.bbox.squeeze(0).cpu()
+ ):
+ label = self.model.config.id2label[int(item.argmax().cpu())]
+ if label == "O":
+ continue
+ score = item.max().item()
+ text = self.processor.tokenizer.decode(inp_ids)
+ bbox = unnormalize_box(bbox.tolist(), image.width, image.height)
+ result.append({"label": label, "score": score, "text": text, "bbox": bbox})
+ return {"predictions": result}
+```
+
+### 2. Send HTTP request using Python
+
+Hugging Face Inference endpoints can directly work with binary data, this means that we can directly send our image from our document to the endpoint. We are going to use `requests` to send our requests. (make your you have it installed `pip install requests`)
+
+```python
+import json
+import requests as r
+import mimetypes
+
+ENDPOINT_URL="" # url of your endpoint
+HF_TOKEN="" # organization token where you deployed your endpoint
+
+def predict(path_to_image:str=None):
+ with open(path_to_image, "rb") as i:
+ b = i.read()
+ headers= {
+ "Authorization": f"Bearer {HF_TOKEN}",
+ "Content-Type": mimetypes.guess_type(path_to_image)[0]
+ }
+ response = r.post(ENDPOINT_URL, headers=headers, data=b)
+ return response.json()
+
+prediction = predict(path_to_image="path_to_your_image.png")
+
+print(prediction)
+# {'predictions': [{'label': 'I-ANSWER', 'score': 0.4823932945728302, 'text': '[CLS]', 'bbox': [0.0, 0.0, 0.0, 0.0]}, {'label': 'B-HEADER', 'score': 0.992474377155304, 'text': 'your', 'bbox': [1712.529, 181.203, 1859.949, 228.88799999999998]},
+```
+
+
+### 3. Draw result on image
+
+To get a better understanding of what the model predicted you can also draw the predictions on the provided image. 
+
+```python
+from PIL import Image, ImageDraw, ImageFont
+
+# draw results on image
+def draw_result(path_to_image,result):
+ image = Image.open(path_to_image)
+ label2color = {
+ "B-HEADER": "blue",
+ "B-QUESTION": "red",
+ "B-ANSWER": "green",
+ "I-HEADER": "blue",
+ "I-QUESTION": "red",
+ "I-ANSWER": "green",
+ }
+
+ # draw predictions over the image
+ draw = ImageDraw.Draw(image)
+ font = ImageFont.load_default()
+ for res in result:
+ draw.rectangle(res["bbox"], outline="black")
+ draw.rectangle(res["bbox"], outline=label2color[res["label"]])
+ draw.text((res["bbox"][0] + 10, res["bbox"][1] - 10), text=res["label"], fill=label2color[res["label"]], font=font)
+ return image
+
+draw_result("path_to_your_image.png", prediction["predictions"])
+```