[QEff.finetuning] Adding sample config and ReadMe file

QEfficient/finetune/experimental/configs/README.md

@@ -0,0 +1,175 @@
+---
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause
+---
+***
+
+# Training Configuration with LoRA Finetuning
+
+## Overview
+
+This configuration file defines the setup for fine-tuning a Hugging Face causal language model using **LoRA (Low-Rank Adaptation)** and **PEFT (Parameter-Efficient Fine-Tuning)** techniques. It also includes dataset, training, optimizer, and scheduler settings.
+
+***
+### 1. Model Configuration
+
+Model-related parameters for loading and fine-tuning.
+
+*   **model\_type**: `hf` → Type of model (`hf` for Hugging Face, `custom` for custom models).
+*   **auto\_class\_name**: `AutoModelForCausalLM` → AutoClass used to load the model.
+*   **model\_name**: `HuggingFaceTB/SmolLM-135M` → Pretrained model to fine-tune.
+*   **load\_in\_4bit**: `false` → If `true`, loads model in 4-bit quantization for memory efficiency.
+*   **use\_peft**: `true` → Enables PEFT for parameter-efficient fine-tuning.
+*   **peft\_config**: Defines LoRA parameters when `use_peft` is `true`:
+    *   `lora_r`: Rank for LoRA adapters.
+    *   `lora_alpha`: Scaling factor for LoRA updates.
+    *   `lora_dropout`: Dropout applied to LoRA layers.
+    *   `target_modules`: Modules to apply LoRA (e.g., `q_proj`, `v_proj`).
+    *   `bias`: Bias handling (`none`, `all`, `lora_only`).
+    *   `task_type`: `CAUSAL_LM` → Task type (e.g., `CAUSAL_LM`, `SEQ_2_SEQ_LM`).
+    *   `peft_type`: `LORA` → Fine-tuning method (e.g., `LORA`, `IA3`).
+
+***
+
+
+### 2. Dataset Configuration
+
+This section defines parameters for dataset handling during fine-tuning with Hugging Face models. It covers dataset type, splits, prompt formatting, and DataLoader settings.
+
+*   **tokenizer\_name**: Matches model name.
+*   **dataset\_type**: `seq_completion` → Used for sequence continuation tasks, where the model     predicts the next tokens given an input text (e.g., summarization, text generation).
+*   **dataset\_name**: Dataset name for training.
+*   **train\_split/test\_split**: Defines splits.
+*   **split\_ratio**: For spliting the train/test dataset, only if train split is provided.
+*   **prompt\_func**: Python function to format prompts.
+*   **completion\_template**: `{output}` → string pattern that tells the fine-tuning pipeline which part of the dataset should be treated as the target output (completion) for the model to learn.
+
+### Example Dataset Configs
+
+### **1. Alpaca (yahma/alpaca-cleaned)**
+
+```yaml
+dataset:
+  tokenizer_name: "meta-llama/Llama-3.2-1B"
+  dataset_type: "seq_completion"
+  dataset_name: "yahma/alpaca-cleaned"
+  train_split: "train"
+  test_split: "test"
+  max_seq_length: 512
+  prompt_func: "alpaca_func:create_alpaca_prompt"
+  completion_template: "{output}"
+
+```
+
+***
+
+### **2. Samsum (knkarthick/samsum)**
+
+```yaml
+dataset:
+  tokenizer_name: "meta-llama/Llama-3.2-1B"
+  dataset_type: "seq_completion"
+  dataset_name: "knkarthick/samsum"
+  train_split: "train"
+  test_split: "test"
+  prompt_func: "samsum_func:create_samsum_prompt"
+  completion_template: "{summary}"
+
+```
+
+***
+### **3. gsm8k (openai/gsm8k)**
+
+```yaml
+dataset:
+  tokenizer_name: "meta-llama/Llama-3.2-1B"
+  dataset_type: "seq_completion"
+  dataset_name: "openai/gsm8k"
+  train_split: "train"
+  test_split: "test"
+  prompt_func: "gsm8k_func:create_gsm8k_prompt"
+  completion_template: "{answer}"
+
+```
+
+ ***
+
+***
+### **4. grammar (grammar_dataset)**
+
+```yaml
+dataset:
+  tokenizer_name: "meta-llama/Llama-3.2-1B"
+  dataset_type: "seq_completion"
+  dataset_name: "grammar"
+  train_split: "train"
+  split_ratio: 0.8
+  prompt_func: "gsm8k_func:create_grammar_prompt"
+  completion_template: "{target}"
+```
+
+ *** 
+### Prompt Function Examples
+
+```python
+# Alpaca
+def create_alpaca_prompt(example):
+    return f"### Instruction:\n{example['instruction']}\n\n### Input:\n{example['input']}\n\n Response:\n"
+
+# Samsum
+def create_samsum_prompt(example):
+    return f"Summarize the following conversation:\n\n{example['dialogue']}\n\nSummary:\n"
+
+#gsm8K
+def create_gsm8k_prompt(example):
+    return f"Solve the following math problem step by step:\n\n{example['question']}\n\nAnswer:\n"
+
+#grammar
+def create_grammar_prompt(example):
+    return f"Correct the grammar in the following sentence:\n\n{example['input']}\n\nCorrected:\n"
+
+```
+
+
+***
+
+### 3. Training Configuration
+
+This section defines core parameters for fine-tuning and evaluation.
+
+*   **type**: `sft` → Specifies training type; `sft` means Supervised Fine-Tuning.
+*   **output\_dir**: Directory where model checkpoints and logs are saved.
+*   **do\_eval**: Enables evaluation during training.
+*   **eval\_strategy**: `epoch` → When to run evaluation (e.g., per epoch or steps).
+*   **gradient\_accumulation\_steps**: Accumulate gradients over multiple steps to simulate larger batch size.
+*   **dtype**: `fp16` → Mixed precision for faster training and reduced memory usage.
+*   **gradient\_checkpointing**: Saves memory by recomputing activations during backward pass (slower but memory-efficient).
+*   **torch\_compile**: Wraps your model with torch.compile() (PyTorch 2.0+) to fuse ops, reduce Python overhead, and generate optimized kernels—often yielding speed-ups without code changes.
+*   **Optional distributed configs**: FSDP, DeepSpeed, or DDP for multi-QAIC or large-scale training.
+*    **resume_from_checkpoint**: Path to a checkpoint to resume training from.
+*    **disable_tqdm**: False by default; set to True to disable progress bar (if running in Notebook).
+
+***
+
+### 4. Optimizer & Scheduler
+
+*   **optimizer**: `adamw` – Optimizer for weight-decoupled regularization; options: `adamw`, `adam`, `sgd`.
+    *   **lr**: Initial learning rate (e.g., `5e-5` for fine-tuning).
+    *   **weight\_decay**: Regularization strength (commonly `0.01`).
+
+*   **scheduler**: `cosine` – Learning rate decay strategy; options: `linear`, `cosine`, `cosine_with_restarts`, `polynomial`, `constant`, `constant_with_warmup`, `inverse_sqrt`.
+    *   **warmup\_steps**: Number of steps or ratio (e.g., `100` steps or `0.05` for 5% of total steps).
+    *   Stabilizes early training and improves convergence.
+
+***
+
+### 5. Callbacks
+
+Callbacks allow custom actions during training, such as logging, early stopping, or hardware profiling.
+
+*   **early\_stopping**: Stops training if no improvement in a monitored metric for a defined patience period.
+*   **tensorboard**: Enables logging of metrics and losses to TensorBoard for visualization.
+*   **QAICProfilerCallback**: Profiles QAIC devices over a specified training step range to monitor performance and resource usage.
+*   **QAICOpByOpVerifierCallback**: Verifies QAIC operations step-by-step during a specified training range for correctness and debugging.
+
+***

QEfficient/finetune/experimental/configs/default_config.yaml

@@ -0,0 +1,104 @@
+# -----------------------------------------------------------------------------
+#
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause
+#
+# -----------------------------------------------------------------------------
+
+# Model configuration
+model:
+  model_type: "hf"  # Hugging Face model
+  auto_class_name: "AutoModelForCausalLM"
+  model_name: "HuggingFaceTB/SmolLM-135M"  # Pretrained model name
+  use_peft: true
+  peft_config:
+    lora_r: 8
+    lora_alpha: 16
+    lora_dropout: 0.1
+    target_modules: ["q_proj", "v_proj"]
+    bias: "none"  # Options: none, all, lora_only
+    task_type: "CAUSAL_LM"  # Options: CAUSAL_LM, SEQ_2_SEQ_LM, etc.
+    peft_type: "LORA"  # Options: LORA, IA3, etc.
+
+# Dataset configuration
+dataset:
+  tokenizer_name: "HuggingFaceTB/SmolLM-135M"
+  dataset_type: "seq_completion"
+  # dataset_name: "Arthur-LAGACHERIE/very-smollm-corpus-0.5M"
+  dataset_name: "knkarthick/samsum"
+  train_split: "train"
+  max_seq_length: 512
+  split_ratio: 0.8  # Ratio for train/test split, used when only train_split is provided
+  test_split: "test"
+  group_by_length: True
+  num_workers: 4
+  pin_memory: True
+  persistent_workers: True
+  prefetch_factor: 1
+  drop_last: False
+
+# Training configuration
+training:
+  type: "sft"
+  output_dir: "./training_results"
+  eval_strategy: "epoch"
+  # eval_steps: 100 # If 'eval_strategy' is 'steps' then it will be used.
+  gradient_accumulation_steps: 1
+  num_train_epochs: 1
+  max_steps: -1
+  log_level: "info"
+  log_on_each_node: True
+  logging_strategy: "steps"
+  logging_steps: 10
+  save_strategy: "epoch"
+  # save_steps: 100   # If 'save_strategy' is 'steps' then it will be used.
+  save_total_limit: 5
+  metric_for_best_model: "eval_loss"
+  dtype: "fp16"
+
+  # Uncomment if running in Notebook
+  # disable_tqdm: True
+
+  # Uncomment below fsdp block to enable FSDP training
+  # fsdp: "full_shard"
+  # fsdp_config: "./configs/accelerate/fsdp_config.yaml"
+  # fsdp_config: "./configs/accelerate/fsdp_tp_parallelism_config.yaml"
+
+  # Uncomment below deepspeed block to enable DeepSpeed training
+  # deepspeed_config: "./configs/accelerate/deepSpeed_config.yaml"
+
+  # Uncomment below DDP block to enable DDP training and configure DDP params
+  # ddp_config:
+  #   ddp_backend: "qccl"
+  #   ddp_find_unused_parameters: False
+  #   ddp_bucket_cap_mb: 25
+  #   ddp_broadcast_buffers: null
+  #   ddp_timeout: 1800
+
+  # Uncomment and populate to resume training
+  # resume_from_checkpoint: "./abc"
+  # restore_callback_states_from_checkpoint: True
+
+  gradient_checkpointing: False
+  gradient_checkpointing_kwargs:
+    preserve_rng_state : True
+    use_reenrant: False
+
+  torch_compile: True
+
+# Optimizer configuration
+optimizers:
+  optimizer_name: "adamw"
+  lr: 5e-5
+  weight_decay: 0.01
+
+scheduler:
+  scheduler_name: "cosine"
+  warmup_steps: 100   # warmup_steps or warmup_ratio
+  warmup_ratio: 0.1
+
+callbacks:
+  early_stopping:
+    early_stopping_patience: 3
+    early_stopping_threshold: 0.001
+  tensorboard: