You can deploy this model and start earning money today!

princegupta19998/
gemma-7b-chatgpt-prompts

Name: gemma-7b-chatgpt-prompts
Availability: InStock

safetensors

Views

No views yet

Read file: Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb Read file: Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb Read file: Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb Read file: gemma-7b-chatgpt-prompts/README.md Read file: README.md Ran tool Ran tool Read file: gemma-7b-chatgpt-prompts/adapter_config.json Read file: gemma-7b-chatgpt-prompts/special_tokens_map.json Read file: gemma-7b-chatgpt-prompts/tokenizer_config.json Here is a summary and analysis of the notebook Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb and the model output directory, to create a clear and informative README.md for your project.

README.md for Gemma-7B ChatGPT Prompts Fine-Tuning with Unsloth

Project Overview

This project demonstrates how to fine-tune Google's Gemma-7B language model using the Unsloth library for efficient LoRA (Low-Rank Adaptation) training. The workflow is implemented in the notebook Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb and uses the fka/awesome-chatgpt-prompts dataset to teach the model to respond to a variety of instructions and prompts in a ChatGPT-like format.

Features

Efficient LoRA fine-tuning using Unsloth for speed and memory savings.
Instruction-tuning on a curated set of ChatGPT-style prompts.
Colab-ready: Designed for use on GPU-enabled environments (e.g., Google Colab).
Model and tokenizer export for easy deployment and sharing.

Directory Structure

.
├── Unsloth_finetuning_gemma_3_4b_pt_lora.ipynb   # Main notebook for fine-tuning
└── gemma-7b-chatgpt-prompts/                    # Output directory for model and tokenizer
    ├── adapter_config.json
    ├── adapter_model.safetensors
    ├── special_tokens_map.json
    ├── tokenizer_config.json
    ├── tokenizer.json
    ├── tokenizer.model
    └── README.md  # (This file)

Quickstart

1. Install Dependencies

!pip install unsloth trl peft accelerate bitsandbytes transformers datasets fsspec

2. Check GPU

python

import torch
print(torch.cuda.is_available())
print(torch.cuda.get_device_name(0))

3. Load and Prepare Dataset

python

from datasets import load_dataset

dataset = load_dataset("fka/awesome-chatgpt-prompts", split="train")
dataset = dataset.map(lambda x: {
    "text": f"### Instruction: {x['act']}\n### Prompt: {x['prompt']}\n### Response:"
})

4. Load and Patch Model

python

from unsloth import FastLanguageModel
import transformers
from transformers.cache_utils import StaticCache, HybridCache
import transformers.models.gemma3.modeling_gemma3

transformers.models.gemma3.modeling_gemma3.StaticCache = StaticCache
transformers.models.gemma3.modeling_gemma3.HybridCache = HybridCache

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="google/gemma-7b",
    max_seq_length=2048,
    dtype=None,
    load_in_4bit=True,
    device_map="auto",
)

5. Add LoRA Adapters

python

model = FastLanguageModel.get_peft_model(
    model,
    r=64,
    target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
    lora_alpha=128,
    lora_dropout=0,
    bias="none",
    use_gradient_checkpointing="unsloth",
    random_state=3407,
    use_rslora=False,
    loftq_config=None,
)

6. Fine-Tune

python

from trl import SFTTrainer
from transformers import TrainingArguments

trainer = SFTTrainer(
    model=model,
    tokenizer=tokenizer,
    train_dataset=dataset,
    dataset_text_field="text",
    max_seq_length=2048,
    dataset_num_proc=2,
    args=TrainingArguments(
        per_device_train_batch_size=2,
        gradient_accumulation_steps=4,
        warmup_steps=10,
        num_train_epochs=3,
        learning_rate=2e-4,
        fp16=not torch.cuda.is_bf16_supported(),
        bf16=torch.cuda.is_bf16_supported(),
        logging_steps=25,
        optim="adamw_8bit",
        weight_decay=0.01,
        lr_scheduler_type="linear",
        seed=3407,
        output_dir="outputs",
        save_strategy="epoch",
        save_total_limit=2,
        dataloader_pin_memory=False,
        report_to="none"
    ),
)
trainer.train()

7. Save Model and Tokenizer

python

model.save_pretrained("gemma-7b-chatgpt-prompts")
tokenizer.save_pretrained("gemma-7b-chatgpt-prompts")

Model Details

Base Model: google/gemma-7b (4-bit quantized via Unsloth)
LoRA Config: r=64, alpha=128, dropout=0, target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Dataset: fka/awesome-chatgpt-prompts
Training: 3 epochs, batch size 2, gradient accumulation 4, AdamW 8-bit optimizer

Tokenizer & Special Tokens

Tokenizer files: tokenizer.json, tokenizer.model, tokenizer_config.json, special_tokens_map.json
Special tokens: <pad>, <eos>, <bos>, <unk>, <start_of_turn>, <end_of_turn>

Inference Example

python

inputs = tokenizer("### Instruction: Give me a recipe for chocolate cake.\n### Prompt: \n### Response:", return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=256, use_cache=True, do_sample=True, top_k=50, top_p=0.95)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Export & Download

The notebook includes code to save and download the model directory for local or cloud storage.
For Google Colab, you can use:
python
1from google.colab import files 2files.download("gemma-7b-chatgpt-prompts")

princegupta19998/gemma-7b-chatgpt-prompts

README.md for Gemma-7B ChatGPT Prompts Fine-Tuning with Unsloth

Project Overview

Features

Directory Structure

Quickstart

1. Install Dependencies

2. Check GPU

3. Load and Prepare Dataset

4. Load and Patch Model

5. Add LoRA Adapters

6. Fine-Tune

7. Save Model and Tokenizer

Model Details

Tokenizer & Special Tokens

Inference Example

Export & Download

References

princegupta19998/
gemma-7b-chatgpt-prompts