andreshere/llama-2-7b-mental-health-counseler

README.md

@@ -0,0 +1,56 @@
+---
+license: llama2
+library_name: peft
+tags:
+- generated_from_trainer
+base_model: meta-llama/Llama-2-7b-chat-hf
+model-index:
+- name: meta-llama/Llama-2-7b-chat-hf
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# meta-llama/Llama-2-7b-chat-hf
+
+This model is a fine-tuned version of [meta-llama/Llama-2-7b-chat-hf](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf) on an unknown dataset.
+
+## 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:
+- learning_rate: 0.0001
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 16
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- num_epochs: 4
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.11.1
+- Transformers 4.41.2
+- Pytorch 2.3.1+cu121
+- Datasets 2.19.2
+- Tokenizers 0.19.1

adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": {
+    "base_model_class": "LlamaForCausalLM",
+    "parent_library": "transformers.models.llama.modeling_llama"
+  },
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-chat-hf",
+  "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": 32,
+  "lora_dropout": 0.1,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c4a176cae3bf7082f2b8750daede32c4ace2e530a84ad6d5172d6723c1bae68
+size 134252336

fine-tuning.llama2.py

@@ -0,0 +1,172 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, Trainer, TrainingArguments
+from datasets import load_dataset, Dataset
+from peft import LoraConfig, get_peft_model
+from transformers import BitsAndBytesConfig, DataCollatorForLanguageModeling
+import evaluate
+import pandas as pd
+import time
+
+# Quantization configuration
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_use_double_quant=True,
+    bnb_4bit_compute_dtype=torch.float16
+)
+
+# Load model and tokenizer
+model_name = "meta-llama/Llama-2-7b-chat-hf"
+tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=True)
+
+# Set the padding token
+tokenizer.pad_token = tokenizer.eos_token
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_name, 
+    quantization_config=bnb_config, 
+    device_map="auto",
+    use_auth_token=True
+)
+
+print(f"Model {model_name} loaded and quantized in 4 bits")
+
+# Load dataset
+dataset = load_dataset("andreshere/counsel_chat", use_auth_token=True)
+print(f"Dataset loaded")
+
+# Configure LoRA
+lora_config = LoraConfig(
+    r=32,
+    lora_alpha=32,
+    target_modules=["q_proj", "k_proj", "v_proj", "o_proj"],
+    lora_dropout=0.1
+)
+
+model = get_peft_model(model, lora_config)
+
+# Adjust dataset without weights
+def preprocess_function(examples):
+    inputs = examples['Context']
+    targets = examples['Response']
+    model_inputs = tokenizer(inputs, max_length=512, truncation=True, padding='max_length')
+    labels = tokenizer(targets, max_length=512, truncation=True, padding='max_length').input_ids
+    
+    model_inputs["labels"] = labels
+    model_inputs["Context"] = inputs
+    model_inputs["Response"] = targets
+    return model_inputs
+
+tokenized_dataset = dataset.map(preprocess_function, batched=True)
+
+train_dataset = tokenized_dataset['train']
+eval_dataset = tokenized_dataset['test']
+
+data_collator = DataCollatorForLanguageModeling(
+    tokenizer=tokenizer,
+    mlm=False,
+)
+
+# Prepare training arguments
+training_args = TrainingArguments(
+    output_dir='./',
+    do_train=True,
+    do_eval=True,
+    eval_strategy='epoch',  # Update to use 'eval_strategy' instead of 'evaluation_strategy'
+    gradient_accumulation_steps=2,
+    auto_find_batch_size=True,
+    weight_decay=0.01,
+    num_train_epochs=4,
+    learning_rate=1e-4,
+    logging_dir='./logs',
+    logging_strategy="steps",
+    logging_steps=10,
+    save_strategy="epoch",
+    save_total_limit=2,
+    save_safetensors=True,
+    eval_steps=10,
+    report_to="tensorboard",
+    hub_strategy="every_save"
+)
+
+# Initialize trainer
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=train_dataset,
+    eval_dataset=eval_dataset,
+    data_collator=data_collator
+)
+
+# Train the model
+trainer.train()
+
+# Save the entire model
+trainer.save_model()
+
+# Push the entire model to the Hugging Face Hub
+trainer.push_to_hub("andreshere/llama-2-7b-mental-health-counseler")
+
+# Load evaluation metrics
+rouge = evaluate.load('rouge')
+bleu = evaluate.load('bleu')
+bertscore = evaluate.load('bertscore')
+meteor = evaluate.load('meteor')
+
+# Function to compute metrics and log compute time
+def compute_metrics_and_log_time(pred):
+    labels_ids = pred.label_ids
+    pred_ids = pred.predictions
+    decoded_preds = tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
+    decoded_labels = tokenizer.batch_decode(labels_ids, skip_special_tokens=True)
+    
+    # Compute metrics
+    rouge_result = rouge.compute(predictions=decoded_preds, references=decoded_labels)
+    bleu_result = bleu.compute(predictions=decoded_preds, references=decoded_labels)
+    bertscore_result = bertscore.compute(predictions=decoded_preds, references=decoded_labels)
+    meteor_result = meteor.compute(predictions=decoded_preds, references=decoded_labels)
+    
+    result = {
+        "rouge": rouge_result,
+        "bleu": bleu_result,
+        "bertscore": bertscore_result,
+        "meteor": meteor_result,
+    }
+    
+    # Log compute time for each response
+    compute_times = []
+    contexts = []
+    original_responses = []
+    model_responses = []
+
+    for i, context in enumerate(pred.features['Context']):
+        start_time = time.time()
+        model_response = decoded_preds[i]
+        end_time = time.time()
+        compute_time = end_time - start_time
+
+        compute_times.append(compute_time)
+        contexts.append(context)
+        original_responses.append(decoded_labels[i])
+        model_responses.append(model_response)
+    
+    # Save the log to a DataFrame
+    log_df = pd.DataFrame({
+        "Context": contexts,
+        "Original Response": original_responses,
+        "Model Response": model_responses,
+        "Compute Time": compute_times
+    })
+    
+    log_df.to_csv("response_log.csv", index=False)
+
+    return result
+
+# Evaluate model
+metrics = trainer.evaluate(eval_dataset=eval_dataset, metric_key_prefix="eval", compute_metrics=compute_metrics_and_log_time)
+print(metrics)
+
+# Export metrics to CSV
+metrics_df = pd.DataFrame(metrics, index=[0])
+metrics_df.to_csv("fine-tuned-metrics.csv", index=False)
+

logs/events.out.tfevents.1717866092.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67d954ad275921629b859935966641f16a0078e335aead3309a02eca1437d621
+size 5228

logs/events.out.tfevents.1717866336.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68506358bcf2fd8e53cfc50212c5fda7c76387d0b1d707e503c2e6ff8178151c
+size 5225

logs/events.out.tfevents.1717866484.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d613f4d97994c6550d1accdc11b1569220f968003989c6287f53e5ceeff5e25d
+size 5225

logs/events.out.tfevents.1717866647.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db49236353ff72105b600b66a2fc42fbcc344995c42685eeb7a610ce292e473e
+size 5225

logs/events.out.tfevents.1717867034.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93e265c3a520cd761e6c4ca063a4ca3eb6f26eb3a92f999dbbf46855d4e5abb0
+size 5225

logs/events.out.tfevents.1717867246.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7de59758d2dd303e0c7cc39057fc0a543cd447901da8d466303fc5c1318f609
+size 17458

logs/events.out.tfevents.1717867605.andrestfm-002

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4b135c9aecab4e959747dfb3a192335021d6bb8a08bc4057c378de7eac53149
+size 48654

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19fd8c9ecc27cc925d0b88edf38f4a94e07e4d3d93eed8d20ce5ccf43119b6f6
+size 5048