Update README.md

README.md

@@ -9,6 +9,7 @@ tags:
 model-index:
 - name: Qwen2-1.5B-Instruct-Function-Calling-v1
   results: []
+pipeline_tag: text-generation
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -16,21 +17,85 @@ should probably proofread and complete it, then remove this comment. -->
 
 # Qwen2-1.5B-Instruct-Function-Calling-v1
 
-This model is a fine-tuned version of [Qwen/Qwen2-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2-1.5B-Instruct) on an unknown dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.2248
-
-## Model description
-
-More information needed
-
-## Intended uses & limitations
-
-More information needed
-
-## Training and evaluation data
-
-More information needed
+This model is a fine-tuned version of [Qwen/Qwen2-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2-1.5B-Instruct) on [devanshamin/gem-viggo-function-calling](https://huggingface.co/datasets/devanshamin/gem-viggo-function-calling) dataset.
+
+## Basic Usage
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# Load the model and the tokenizer
+model_id = "Qwen2-1.5B-Instruct-Function-Calling-v1"
+model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float32, device_map="auto")
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+def inference(prompt: str) -> str:
+  model_inputs = tokenizer([prompt], return_tensors="pt").to(device)
+  generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=512)
+  generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
+  response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
+  return response
+
+prompt = "What is the meaning of life?"
+messages = [
+  {"role": "system", "content": "You are a helpful assistant."},
+  {"role": "user", "content": prompt}
+]
+prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+response = inference(prompt)
+print(response)
+```
+
+## Tool Usage
+
+### Basic
+
+```python
+
+```
+
+### Advanced
+```python
+import re
+from enum import Enum
+
+from pydantic import BaseModel, Field # pip install pydantic
+from instructor.function_calls import openai_schema # pip install instructor
+
+def get_prompt(tool: str, user_input: str) -> str:
+  system = "You are a helpful assistant with access to the following tools. Use them if required - \n```json\n{}\n```"
+  messages = [
+    {"role": "system", "content": system.format(tool)},
+    {"role": "user", "content": 'Extract the information from the following - \n{}'.format(user_input)}
+  ]
+  prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+  return prompt
+
+# Define functions using pydantic classes
+class PaperCategory(str, Enum):
+  TYPE_1_DIABETES = 'Type 1 Diabetes'
+  TYPE_2_DIABETES = 'Type 2 Diabetes'
+
+class Classification(BaseModel):
+  label: PaperCategory = Field(..., description='Provide the most likely category')
+  reason: str = Field(..., description='Give a detailed explanation with quotes from the abstract explaining why the paper is related to the chosen label.')
+
+function_definition = openai_schema(Classification).openai_schema
+tool = dict(type='function', function=function_definition)
+input_text = "1,25-dihydroxyvitamin D(3) (1,25(OH)(2)D(3)), the biologically active form of vitamin D, is widely recognized as a modulator of the immune system as well as a regulator of mineral metabolism. The objective of this study was to determine the effects of vitamin D status and treatment with 1,25(OH)(2)D(3) on diabetes onset in non-obese diabetic (NOD) mice, a murine model of human type I diabetes. We have found that vitamin D-deficiency increases the incidence of diabetes in female mice from 46% (n=13) to 88% (n=8) and from 0% (n=10) to 44% (n=9) in male mice as of 200 days of age when compared to vitamin D-sufficient animals. Addition of 50 ng of 1,25(OH)(2)D(3)/day to the diet prevented disease onset as of 200 days and caused a significant rise in serum calcium levels, regardless of gender or vitamin D status. Our results indicate that vitamin D status is a determining factor of disease susceptibility and oral administration of 1,25(OH)(2)D(3) prevents diabetes onset in NOD mice through 200 days of age."
+prompt = get_prompt(json.dumps(tool), input_text)
+output = inference(prompt)
+print(output)
+# ```json
+# {"name": "Classification", "arguments": {"label": "Type 1 Diabetes", "reason": "The study investigated the effect of vitamin D status and treatment with 1,25(OH)(2)D(3) on diabetes onset in non-obese diabetic (NOD) mice. It also concluded that vitamin D deficiency leads to an increase in diabetes incidence and that the addition of 1,25(OH)(2)D(3) can prevent diabetes onset in NOD mice."}}
+# ```
+# Extract JSON string using regex
+output = re.search(r'```json\s*(\{.*?\})\s*```', output).group(1)
+output = Classification(**json.loads(_output)['arguments'])
+print(output)
+# Classification(label=<PaperCategory.TYPE_1_DIABETES: 'Type 1 Diabetes'>, reason='The study investigated the effect of vitamin D status and treatment with 1,25(OH)(2)D(3) on diabetes onset in non-obese diabetic (NOD) mice. It also concluded that vitamin D deficiency leads to an increase in diabetes incidence and that the addition of 1,25(OH)(2)D(3) can prevent diabetes onset in NOD mice.')
+```
 
 ## Training procedure
 
@@ -64,8 +129,10 @@ The following hyperparameters were used during training:
 
 ### Framework versions
 
-- PEFT 0.11.1
-- Transformers 4.42.3
-- Pytorch 2.3.1+cu121
-- Datasets 2.20.0
-- Tokenizers 0.19.1
+```text
+peft==0.11.1
+transformers==4.42.3
+torch==2.3.1+cu121
+datasets==2.20.0
+tokenizers==0.19.1
+```