Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.psd filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+__pycache__/
+.idea/
+
+

README.md

@@ -1,3 +1,78 @@
----
-license: mit
----
+---
+license: mit
+language:
+- en
+base_model:
+- meta-llama/Meta-Llama-3-8B-Instruct
+pipeline_tag: text-generation
+tags:
+- Food
+- nutrient prediction 
+- healthy diet
+- Food NEL
+- Food NER
+---
+# 'FoodyLLM: A FAIR-aligned specialized large language model for food and nutrition analysis' 
+
+## The model is based on Meta-Llama-3-8B-Instruct, which was fine-tuned (using LoRA) for food and nutrition analysis. 
+
+More specifically, it can conduct the following tasks:
+- Assessing recipe nutritional profiles
+- Classifying recipes by traffic light nutrition labels (see https://www.food.gov.uk/safety-hygiene/check-the-label for details on the labeling)
+- Extract food named entities from text (Food NER)
+- Link the food entities to three distinct ontologies, Hansard taxonomy, FoodOn and SNOMED-CT (Food NEL)
+
+## How to use it: ##
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+import torch
+
+if __name__ == '__main__':
+    base_model = "meta-llama/Meta-Llama-3-8B-Instruct"
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+    bnb_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_quant_type="nf4",
+        bnb_4bit_compute_dtype=torch.float16,
+        bnb_4bit_use_double_quant=True,
+    )
+
+    model = AutoModelForCausalLM.from_pretrained(
+        base_model,
+        quantization_config=bnb_config,
+        device_map={"": 0},
+        attn_implementation="eager"
+    )
+
+    tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
+    tokenizer.pad_token = '<|pad|>'
+    tokenizer.pad_token_id = 128255
+
+    #Load LORA weights
+    model.load_adapter("Matej/FoodyLLM")
+    model.config.use_cache = True
+    model.eval()
+
+    system_prompt = ""
+    user_prompt = "Calculate the nutrient values per 100 grams in a recipe with these ingredients: 10 ounce candies, marshmallows, 1/4 cup butter, without salt, 6 cup cocoa, dry powder, unsweetened"
+
+    messages = [
+        {
+            "role": "user",
+            "content": f"{system_prompt} {user_prompt}".strip()
+        }
+    ]
+
+    prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+
+    #Here we have a batch of one
+    tokenizer_input = [prompt]
+
+    inputs = tokenizer(tokenizer_input, return_tensors="pt", padding=True, truncation=True, max_length=1024).to(device)
+    generated_ids = model.generate(**inputs, max_new_tokens=1024, do_sample=True)
+    answers = tokenizer.batch_decode(generated_ids[:, inputs['input_ids'].shape[1]:])
+    answers = [x.split('<|eot_id|>')[0].strip() for x in answers]
+    print(answers)
+```

adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "meta-llama/Meta-Llama-3-8B-Instruct",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "down_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "gate_proj",
+    "k_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cee014ffbf0e4595d8773716cdc3693d5e3ed05033d64b52687d09cf4e93b4af
+size 167832240

apply.py

@@ -0,0 +1,50 @@
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+import torch
+
+if __name__ == '__main__':
+    base_model = "meta-llama/Meta-Llama-3-8B-Instruct"
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+    bnb_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_quant_type="nf4",
+        bnb_4bit_compute_dtype=torch.float16,
+        bnb_4bit_use_double_quant=True,
+    )
+
+    model = AutoModelForCausalLM.from_pretrained(
+        base_model,
+        quantization_config=bnb_config,
+        device_map={"": 0},
+        attn_implementation="eager"
+    )
+
+    tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
+    tokenizer.pad_token = '<|pad|>'
+    tokenizer.pad_token_id = 128255
+
+    #Load LORA weights
+    model.load_adapter("Matej/FoodyLLM")
+    model.config.use_cache = True
+    model.eval()
+
+    system_prompt = ""
+    user_prompt = "Calculate the nutrient values per 100 grams in a recipe with these ingredients: 10 ounce candies, marshmallows, 1/4 cup butter, without salt, 6 cup cocoa, dry powder, unsweetened"
+
+    messages = [
+        {
+            "role": "user",
+            "content": f"{system_prompt} {user_prompt}".strip()
+        }
+    ]
+
+    prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+
+    #Here we have a batch of one
+    tokenizer_input = [prompt]
+
+    inputs = tokenizer(tokenizer_input, return_tensors="pt", padding=True, truncation=True, max_length=1024).to(device)
+    generated_ids = model.generate(**inputs, max_new_tokens=1024, do_sample=True)
+    answers = tokenizer.batch_decode(generated_ids[:, inputs['input_ids'].shape[1]:])
+    answers = [x.split('<|eot_id|>')[0].strip() for x in answers]
+    print(answers)

requirements.txt

@@ -0,0 +1,3 @@
+transformers
+torch
+peft