set version for v0.9.1

* fix(log): clarify error on dataset loading failed

* fix: add path for easy tracking of broken config

* fix: improve error message based on pr feedback

.github/workflows/tests.yml

@@ -329,18 +329,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: 124
-            cuda_version: 12.4.1
-            python_version: "3.11"
-            pytorch: 2.6.0
-            num_gpus: 1
-            axolotl_extras: llmcompressor
-          - cuda: 124
-            cuda_version: 12.4.1
-            python_version: "3.11"
-            pytorch: 2.4.1
-            num_gpus: 1
-            axolotl_extras:
           - cuda: 124
             cuda_version: 12.4.1
             python_version: "3.11"

docs/custom_integrations.qmd

@@ -49,8 +49,7 @@ sections = [
     ("Knowledge Distillation (KD)", "kd"),
     ("Liger Kernels", "liger"),
     ("Language Model Evaluation Harness (LM Eval)", "lm_eval"),
-    ("Spectrum", "spectrum"),
+    ("Spectrum", "spectrum")
-    ("LLMCompressor", "llm_compressor")
 ]
 
 for section_name, folder_name in sections:

examples/llama-3/sparse-finetuning.yaml

@@ -1,77 +0,0 @@
-base_model: neuralmagic/Sparse-Llama-3.1-8B-2of4
-
-plugins:
-  - axolotl.integrations.llm_compressor.LLMCompressorPlugin
-
-load_in_8bit: false
-load_in_4bit: false
-strict: false
-
-datasets:
-  - path: tatsu-lab/alpaca
-    type: alpaca
-dataset_prepared_path: last_run_prepared
-val_set_size: 0.05
-output_dir: ./outputs/out
-
-sequence_len: 4096
-sample_packing: true
-pad_to_sequence_len: true
-eval_sample_packing: false
-
-wandb_project:
-wandb_entity:
-wandb_watch:
-wandb_name:
-wandb_log_model:
-
-gradient_accumulation_steps: 8
-micro_batch_size: 1
-num_epochs: 1
-optimizer: paged_adamw_8bit
-lr_scheduler: cosine
-learning_rate: 2e-5
-
-train_on_inputs: false
-group_by_length: false
-bf16: auto
-fp16:
-tf32: false
-
-gradient_checkpointing: true
-gradient_checkpointing_kwargs:
-  use_reentrant: false
-early_stopping_patience:
-resume_from_checkpoint:
-logging_steps: 1
-xformers_attention:
-flash_attention: true
-
-warmup_steps: 100
-evals_per_epoch: 2
-eval_table_size:
-saves_per_epoch: 1
-debug:
-deepspeed:
-weight_decay: 0.0
-fsdp:
-fsdp_config:
-special_tokens:
-  pad_token: <|end_of_text|>
-
-llmcompressor:
-  recipe:
-    finetuning_stage:
-      finetuning_modifiers:
-        ConstantPruningModifier:
-          targets: [
-            're:.*q_proj.weight',
-            're:.*k_proj.weight',
-            're:.*v_proj.weight',
-            're:.*o_proj.weight',
-            're:.*gate_proj.weight',
-            're:.*up_proj.weight',
-            're:.*down_proj.weight',
-          ]
-          start: 0
-  save_compressed: true

setup.py

@@ -150,9 +150,6 @@ extras_require = {
     "vllm": [
         "vllm==0.7.2",
     ],
-    "llmcompressor": [
-        "llmcompressor==0.5.1",
-    ],
 }
 
 install_requires, dependency_links, extras_require_build = parse_requirements(

src/axolotl/__init__.py

@@ -4,4 +4,4 @@ import pkgutil
 
 __path__ = pkgutil.extend_path(__path__, __name__)  # Make this a namespace package
 
-__version__ = "0.10.0.dev0"
+__version__ = "0.9.1"

src/axolotl/integrations/llm_compressor/README.md

@@ -1,108 +0,0 @@
-# LLMCompressor Integration
-
-Fine-tune sparsified models in Axolotl using Neural Magic's [LLMCompressor](https://github.com/vllm-project/llm-compressor).
-
-This integration enables fine-tuning of models sparsified using LLMCompressor within the Axolotl training framework. By combining LLMCompressor's model compression capabilities with Axolotl's distributed training pipelines, users can efficiently fine-tune sparse models at scale.
-
-It uses Axolotl’s plugin system to hook into the fine-tuning flows while maintaining sparsity throughout training.
-
----
-
-## Requirements
-
-- Axolotl with `llmcompressor` extras:
-
-  ```bash
-  pip install "axolotl[llmcompressor]"
-  ```
-
-- Requires `llmcompressor >= 0.5.1`
-
-This will install all necessary dependencies to fine-tune sparsified models using the integration.
-
----
-
-## Usage
-
-To enable sparse fine-tuning with this integration, include the plugin in your Axolotl config:
-
-```yaml
-plugins:
-  - axolotl.integrations.llm_compressor.LLMCompressorPlugin
-
-llmcompressor:
-  recipe:
-    finetuning_stage:
-      finetuning_modifiers:
-        ConstantPruningModifier:
-          targets: [
-            're:.*q_proj.weight',
-            're:.*k_proj.weight',
-            're:.*v_proj.weight',
-            're:.*o_proj.weight',
-            're:.*gate_proj.weight',
-            're:.*up_proj.weight',
-            're:.*down_proj.weight',
-          ]
-          start: 0
-  save_compressed: true
-# ... (other training arguments)
-```
-
-This plugin **does not apply pruning or sparsification itself** — it is intended for **fine-tuning models that have already been sparsified**.
-
-Pre-sparsified checkpoints can be:
-- Generated using [LLMCompressor](https://github.com/vllm-project/llm-compressor)
-- Downloaded from [Neural Magic's Hugging Face page](https://huggingface.co/neuralmagic)
-- Any custom LLM with compatible sparsity patterns that you've created yourself
-
-To learn more about writing and customizing LLMCompressor recipes, refer to the official documentation:
-[https://github.com/vllm-project/llm-compressor/blob/main/README.md](https://github.com/vllm-project/llm-compressor/blob/main/README.md)
-
-### Storage Optimization with save_compressed
-
-Setting `save_compressed: true` in your configuration enables saving models in a compressed format, which:
-- Reduces disk space usage by approximately 40%
-- Maintains compatibility with vLLM for accelerated inference
-- Maintains compatibility with llmcompressor for further optimization (example: quantization)
-
-This option is highly recommended when working with sparse models to maximize the benefits of model compression.
-
-### Example Config
-
-See [`examples/llama-3/sparse-finetuning.yaml`](examples/llama-3/sparse-finetuning.yaml) for a complete example.
-
----
-
-## Inference with vLLM
-
-After fine-tuning your sparse model, you can leverage vLLM for efficient inference.
-You can also use LLMCompressor to apply additional quantization to your fine-tuned
-sparse model before inference for even greater performance benefits.:
-
-```python
-from vllm import LLM, SamplingParams
-
-prompts = [
-    "Hello, my name is",
-    "The president of the United States is",
-    "The capital of France is",
-    "The future of AI is",
-]
-sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
-llm = LLM("path/to/your/sparse/model")
-outputs = llm.generate(prompts, sampling_params)
-
-for output in outputs:
-    prompt = output.prompt
-    generated_text = output.outputs[0].text
-    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
-```
-
-For more details on vLLM's capabilities and advanced configuration options, see the [official vLLM documentation](https://docs.vllm.ai/).
-
-## Learn More
-
-For details on available sparsity and quantization schemes, fine-tuning recipes, and usage examples, visit the official LLMCompressor repository:
-
-[https://github.com/vllm-project/llm-compressor](https://github.com/vllm-project/llm-compressor)

src/axolotl/integrations/llm_compressor/__init__.py

@@ -1,5 +0,0 @@
-"""Integration entry point for the LLMCompressor plugin."""
-
-from .plugin import LLMCompressorPlugin
-
-__all__ = ["LLMCompressorPlugin"]

src/axolotl/integrations/llm_compressor/args.py

@@ -1,40 +0,0 @@
-"""
-LLMCompressor and Sparse Finetuning config models.
-"""
-
-from typing import Any
-
-from pydantic import BaseModel, Field
-from typing_extensions import Annotated
-
-
-class CompressionArgs(BaseModel):
-    """Sparse Finetuning config for LLMCompressor."""
-
-    # Typing for recipe is set to Any due to:
-    # https://github.com/vllm-project/llm-compressor/issues/1319
-    recipe: Annotated[
-        Any,
-        Field(
-            description="The recipe containing the compression algorithms and hyperparameters to apply."
-        ),
-    ]
-
-    save_compressed: Annotated[
-        bool,
-        Field(
-            default=False,
-            description="Whether to save the compressed model after training.",
-        ),
-    ]
-
-
-class LLMCompressorArgs(BaseModel):
-    """LLMCompressor configuration BaseModel."""
-
-    llmcompressor: Annotated[
-        CompressionArgs,
-        Field(
-            description="Arguments enabling compression pathways through the LLM Compressor plugins"
-        ),
-    ]

src/axolotl/integrations/llm_compressor/plugin.py

@@ -1,171 +0,0 @@
-"""
-Sparse Finetuning plugin for Axolotl — enables handling of sparse neural networks
-by maintaining masks for zero weights during training.
-"""
-
-import logging
-from functools import wraps
-from typing import Any, Callable, Concatenate, ParamSpec, TypeVar
-
-from llmcompressor import active_session, create_session
-from llmcompressor.core import callbacks as session_callbacks
-from llmcompressor.recipe import Recipe
-from torch.nn import Module
-from transformers.trainer import Trainer
-from transformers.trainer_callback import TrainerCallback, TrainerControl, TrainerState
-from transformers.training_args import TrainingArguments
-
-from axolotl.integrations.base import BasePlugin
-
-P = ParamSpec("P")  # Params for generic function signatures
-R = TypeVar("R")  # Return type for generic function signatures
-
-LOG = logging.getLogger("axolotl.integrations.llm_compressor")
-
-
-class LLMCompressorCallbackHandler(TrainerCallback):
-    """
-    Trainer callback for Sparse Finetuning.
-    Maintains sparsity patterns during training by applying masks after optimization steps,
-    ensuring zero-weight updates are canceled out.
-    """
-
-    def __init__(self, trainer: Trainer, recipe: Any):
-        """
-        Initialize the Sparse Finetuning callback handler.
-
-        Args:
-            trainer (Trainer): Huggingface Trainer instance.
-            recipe (Recipe | dict): Sparse finetuning recipe to apply.
-        """
-        super().__init__()
-        self.trainer = trainer
-        self.recipe = (
-            Recipe.model_validate(recipe) if not isinstance(recipe, Recipe) else recipe
-        )
-        self.original_compute_loss = trainer.compute_loss
-        self.trainer.compute_loss = compute_loss_wrapper(self.trainer.compute_loss)
-        create_session()
-
-    def on_train_begin(
-        self,
-        args: TrainingArguments,
-        state: TrainerState,
-        control: TrainerControl,
-        **kwargs,
-    ) -> None:
-        """
-        Called at the beginning of training. Initializes the compression session.
-
-        Args:
-            args (TrainingArguments): Training arguments.
-            state (TrainerState): Trainer state.
-            control (TrainerControl): Trainer control.
-        """
-        super().on_train_begin(args, state, control, **kwargs)
-        self.trainer.accelerator.wait_for_everyone()
-        active_session().initialize(
-            model=self.trainer.model,
-            optimizer=self.trainer.optimizer,
-            start=state.epoch,
-            recipe=self.recipe,
-        )
-        self.trainer.accelerator.wait_for_everyone()
-
-    def on_step_begin(
-        self,
-        args: TrainingArguments,
-        state: TrainerState,
-        control: TrainerControl,
-        **kwargs,
-    ) -> None:
-        """
-        Called at the beginning of a training step. Triggers batch_start callback.
-        """
-        super().on_step_begin(args, state, control, **kwargs)
-        session_callbacks.batch_start()
-
-    def on_step_end(
-        self,
-        args: TrainingArguments,
-        state: TrainerState,
-        control: TrainerControl,
-        **kwargs,
-    ) -> None:
-        """
-        Called at the end of a training step. Triggers optimizer and batch_end callbacks.
-        """
-        super().on_step_end(args, state, control, **kwargs)
-        session_callbacks.optim_pre_step()
-        session_callbacks.optim_post_step()
-        session_callbacks.batch_end()
-
-    def on_train_end(
-        self,
-        args: TrainingArguments,
-        state: TrainerState,
-        control: TrainerControl,
-        **kwargs,
-    ) -> None:
-        """
-        Called at the end of training. Finalizes the compression session.
-        """
-        super().on_train_end(args, state, control, **kwargs)
-        active_session().finalize()
-        self.trainer.compute_loss_func = self.original_compute_loss
-
-
-class LLMCompressorPlugin(BasePlugin):
-    """
-    Sparse Finetuning plugin for Axolotl integration.
-    """
-
-    def get_input_args(self) -> str:
-        """
-        Returns the path to the plugin's argument definition.
-
-        Returns:
-            str: Dotted path to the LLMCompressorArgs class.
-        """
-        return "axolotl.integrations.llm_compressor.args.LLMCompressorArgs"
-
-    def add_callbacks_post_trainer(self, cfg: Any, trainer: Trainer) -> list:
-        """
-        Adds Sparse Finetuning callback to the Trainer instance.
-
-        Args:
-            cfg (Any): Configuration object containing the sparse recipe.
-            trainer (Trainer): Huggingface Trainer instance.
-
-        Returns:
-            list: List containing the configured callback instances.
-        """
-        LOG.info("Adding Sparse Finetuning callback to the trainer")
-        callback = LLMCompressorCallbackHandler(
-            trainer=trainer,
-            recipe=cfg.llmcompressor.recipe,
-        )
-        return [callback]
-
-
-def compute_loss_wrapper(
-    compute_loss_func: Callable[Concatenate[Module, P], R],
-) -> Callable[Concatenate[Module, P], R]:
-    """
-    Wraps the loss computation function to trigger the loss_calculated callback.
-
-    Args:
-        compute_loss_func (Callable): Original loss computation function.
-
-    Returns:
-        Callable: Wrapped function that also invokes the loss_calculated callback.
-    """
-
-    @wraps(compute_loss_func)
-    def compute_and_notify(model: Module, *args: P.args, **kwargs: P.kwargs) -> R:
-        loss = compute_loss_func(model, *args, **kwargs)
-        if active_session().lifecycle.initialized_ and model.training:
-            session_callbacks.loss_calculated(loss=loss)
-        return loss
-
-    return compute_and_notify

src/axolotl/integrations/llm_compressor/utils.py

@@ -1,40 +0,0 @@
-"""Utilities for llmcompressor integration with axolotl."""
-
-from typing import Union
-
-from llmcompressor.transformers.sparsification.compressed_tensors_utils import (
-    modify_save_pretrained,
-)
-from transformers import PreTrainedModel, Trainer
-
-
-def save_compressed_model(
-    model: PreTrainedModel,
-    output_dir: Union[str, bytes],
-    trainer: Trainer,
-    safe_serialization: bool = False,
-    save_compressed: bool = False,
-) -> None:
-    """
-    Synchronize processes, apply compression hooks, and save the model.
-
-    Args:
-        model (PreTrainedModel): The model to be saved.
-        output_dir (str or bytes): Path where the model files will be written.
-        trainer (Trainer): Hugging Face Trainer for process synchronization.
-        safe_serialization (bool): Use safe serialization if True.
-        save_compressed (bool): Write compressed tensors if True.
-    """
-    trainer.accelerator.wait_for_everyone()
-
-    # Only the main process writes the files
-    if not trainer.accelerator.is_main_process:
-        return
-
-    modify_save_pretrained(model)
-    model.save_pretrained(
-        output_dir,
-        safe_serialization=safe_serialization,
-        save_compressed=save_compressed,
-        skip_sparsity_compression_stats=not save_compressed,
-    )

src/axolotl/train.py

@@ -294,23 +294,8 @@ def save_trained_model(
             trainer.model.save_pretrained(
                 cfg.output_dir, safe_serialization=safe_serialization
             )
-
         model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
 
-    if hasattr(cfg, "llmcompressor") and cfg.llmcompressor:
-        # TODO: add integration support so this can be implemented completely within the plugin
-        from axolotl.integrations.llm_compressor.utils import (
-            save_compressed_model,
-        )
-
-        save_compressed_model(
-            model=model,
-            output_dir=cfg.output_dir,
-            trainer=trainer,
-            safe_serialization=safe_serialization,
-            save_compressed=cfg.llmcompressor.save_compressed,
-        )
-
 
 def create_model_card(cfg: DictDefault, trainer: Trainer):
     """

src/axolotl/utils/gradient_checkpointing/__init__.py

@@ -5,11 +5,8 @@ from functools import partial
 
 from packaging import version
 
-from axolotl.utils.gradient_checkpointing.offload_cpu import (
+from axolotl.utils.gradient_checkpointing.unsloth import (
-    CPU_Offloaded_Gradient_Checkpointer,
+    Unsloth_Offloaded_Gradient_Checkpointer,
-)
-from axolotl.utils.gradient_checkpointing.offload_disk import (
-    DiskOffloadedGradientCheckpointer,
 )
 
 transformers_version = version.parse(importlib.metadata.version("transformers"))
@@ -29,31 +26,12 @@ def hf_grad_checkpoint_offload_wrapper(
     decoder_layer, *args, use_reentrant=None
 ):  # pylint: disable=unused-argument
     if uses_gc_layers(decoder_layer):
-        return CPU_Offloaded_Gradient_Checkpointer.apply(
+        return Unsloth_Offloaded_Gradient_Checkpointer.apply(
             decoder_layer,
             *args,
         )
 
-    return CPU_Offloaded_Gradient_Checkpointer.apply(
+    return Unsloth_Offloaded_Gradient_Checkpointer.apply(
-        (
-            decoder_layer.func.__self__
-            if isinstance(decoder_layer, partial)
-            else decoder_layer.__self__
-        ),
-        *args,
-    )
-
-
-def hf_grad_checkpoint_disk_offload_wrapper(
-    decoder_layer, *args, use_reentrant=None
-):  # pylint: disable=unused-argument
-    if uses_gc_layers(decoder_layer):
-        return DiskOffloadedGradientCheckpointer.apply(
-            decoder_layer,
-            *args,
-        )
-
-    return DiskOffloadedGradientCheckpointer.apply(
         (
             decoder_layer.func.__self__
             if isinstance(decoder_layer, partial)

src/axolotl/utils/gradient_checkpointing/offload_disk.py

@@ -1,93 +0,0 @@
-"""Disk offloaded checkpointing"""
-
-import os
-import tempfile
-import uuid
-
-import torch
-
-torch_cuda_amp_custom_fwd = torch.amp.custom_fwd(device_type="cuda")
-torch_cuda_amp_custom_bwd = torch.amp.custom_bwd(device_type="cuda")
-
-
-class DiskOffloadedGradientCheckpointer(torch.autograd.Function):
-    """
-    Saves both VRAM and RAM by offloading activations to disk.
-    Greater hit to performance than RAM offloading, but useful for extremely memory-constrained environments.
-    """
-
-    # Create a temporary directory for storing tensors
-    _temp_dir = tempfile.mkdtemp(prefix="disk_checkpoint_")
-
-    @staticmethod
-    def _get_temp_file_path():
-        """Generate a unique file path for tensor storage"""
-        return os.path.join(
-            DiskOffloadedGradientCheckpointer._temp_dir, f"{uuid.uuid4()}.pt"
-        )
-
-    @staticmethod
-    @torch_cuda_amp_custom_fwd
-    def forward(ctx, forward_function, hidden_states, *args):
-        # Generate a unique file path for this tensor
-        file_path = DiskOffloadedGradientCheckpointer._get_temp_file_path()
-
-        # Save tensor to disk in a non-blocking way (detached from compute)
-        # First move to CPU, then save
-        cpu_hidden_states = hidden_states.detach().cpu()
-        torch.save(cpu_hidden_states, file_path)
-
-        # Free CPU memory
-        del cpu_hidden_states
-
-        # Run forward pass
-        with torch.no_grad():
-            output = forward_function(hidden_states, *args)
-
-        # Store the path instead of the tensor
-        ctx.save_for_backward(torch.tensor([0]))  # Dummy tensor
-        ctx.file_path = file_path
-        ctx.forward_function = forward_function
-        ctx.args = args
-        return output
-
-    @staticmethod
-    @torch_cuda_amp_custom_bwd
-    def backward(ctx, dY):  # pylint: disable=invalid-name
-        # Load the hidden states from disk
-        hidden_states = torch.load(ctx.file_path, weights_only=True)
-
-        # Move to CUDA and prepare for gradient computation
-        hidden_states = hidden_states.to("cuda", non_blocking=True).detach()
-        hidden_states.requires_grad = True
-
-        # Clean up the temporary file
-        try:
-            os.remove(ctx.file_path)
-        except FileNotFoundError:
-            pass  # Ignore errors in file deletion
-
-        # Compute gradients
-        with torch.enable_grad():
-            output = ctx.forward_function(hidden_states, *ctx.args)
-        # pylint: disable=duplicate-code
-        torch.autograd.backward(output, dY)
-
-        return (
-            None,
-            hidden_states.grad,
-        ) + (
-            None,
-        ) * len(ctx.args)
-
-    @staticmethod
-    def cleanup():
-        """Clean up the temporary directory when done"""
-        import shutil
-
-        try:
-            shutil.rmtree(
-                DiskOffloadedGradientCheckpointer._temp_dir
-            )  # pylint: disable=protected-access
-        except FileNotFoundError:
-            pass

src/axolotl/utils/gradient_checkpointing/unsloth.py

@@ -1,4 +1,4 @@
-"""CPU offloaded checkpointing"""
+"""Unsloth checkpointing"""
 
 # Copyright 2023-present Daniel Han-Chen & the Unsloth team. All rights reserved.
 #
@@ -26,7 +26,7 @@ else:
     torch_cuda_amp_custom_bwd = torch.amp.custom_bwd(device_type="cuda")
 
 
-class CPU_Offloaded_Gradient_Checkpointer(  # pylint: disable=invalid-name
+class Unsloth_Offloaded_Gradient_Checkpointer(  # pylint: disable=invalid-name
     torch.autograd.Function
 ):
     """

src/axolotl/utils/models.py

@@ -70,10 +70,7 @@ from axolotl.utils.distributed import (
     is_local_main_process,
     is_main_process,
 )
-from axolotl.utils.gradient_checkpointing import (
+from axolotl.utils.gradient_checkpointing import hf_grad_checkpoint_offload_wrapper
-    hf_grad_checkpoint_disk_offload_wrapper,
-    hf_grad_checkpoint_offload_wrapper,
-)
 from axolotl.utils.lora_embeddings import get_linear_embedding_layers
 from axolotl.utils.model_shard_quant import load_sharded_model, load_sharded_model_quant
 
@@ -144,22 +141,6 @@ def check_model_config(cfg: DictDefault, model_config: PretrainedConfig):
         hasattr(model_config, "quantization_config")
         and model_config.quantization_config
     )
-
-    # Detect compressed-tensors config
-    is_compressed_tensors_config = (
-        quant_config_exists
-        and model_config.quantization_config.get("quant_method") == "compressed-tensors"
-    )
-
-    if is_compressed_tensors_config:
-        if model_config.quantization_config.get("config_groups"):
-            LOG.warning(
-                "Found `config_groups` in a compressed-tensors config. "
-                "QAT integration with llmcompressor is not tested."
-            )
-        # Skip further quant checks for compressed-tensors
-        return
-
     quant_config_method_is_gptq = (
         quant_config_exists
         and "quant_method" in model_config.quantization_config
@@ -622,10 +603,6 @@ class ModelLoader:
 
         if self.cfg.gradient_checkpointing in ["unsloth", "offload"]:
             transformers.modeling_utils.checkpoint = hf_grad_checkpoint_offload_wrapper
-        if self.cfg.gradient_checkpointing == "offload_disk":
-            transformers.modeling_utils.checkpoint = (
-                hf_grad_checkpoint_disk_offload_wrapper
-            )
 
         if self.cfg.flash_attention:
             self.patch_attention()

src/axolotl/utils/samplers/multipack.py

@@ -78,11 +78,15 @@ def pack_group(
     Returns:
         List of bins, where each bin contains indices of sequences assigned to it
     """
+    # Get sorting indices and sort lengths in descending order
+    indices = np.argsort(sequence_lengths)[::-1]
+    sorted_lengths = sequence_lengths[indices]
+
     bins_remaining_space: list = []  # Tracks remaining capacity in each bin
     bins_assigned_sequences: list = []  # Tracks sequence indices assigned to each bin
 
-    for seq_id, size in enumerate(sequence_lengths):
+    for seq_id, size in enumerate(sorted_lengths):
-        global_idx = seq_id + group_offset
+        global_idx = indices[seq_id] + group_offset
 
         # Try to place sequence in existing bins
         add_new_bin = True

src/axolotl/utils/schemas/config.py

@@ -178,9 +178,9 @@ class AxolotlInputConfig(
 
     # torch_dtype: torch.dtype | None
 
-    gradient_checkpointing: (
+    gradient_checkpointing: Literal["unsloth", "offload"] | bool | None = Field(
-        Literal["unsloth", "offload", "offload_disk"] | bool | None
+        default=False
-    ) = Field(default=False)
+    )
     gradient_checkpointing_kwargs: dict[str, Any] | None = None
 
     unfrozen_parameters: list[str] | None = None

tests/e2e/integrations/test_kd.py

@@ -90,7 +90,7 @@ class TestKnowledgeDistillation:
         train(cfg=cfg, dataset_meta=dataset_meta)
         assert (Path(temp_dir) / "model.safetensors").exists()
         check_tensorboard(
-            temp_dir + "/runs", "train/loss", 1.2, "Train Loss (%s) is too high"
+            temp_dir + "/runs", "train/loss", 1.0, "Train Loss is too high"
         )
 
     @pytest.mark.parametrize(
@@ -121,5 +121,5 @@ class TestKnowledgeDistillation:
         train(cfg=cfg, dataset_meta=dataset_meta)
         assert (Path(temp_dir) / "adapter_model.safetensors").exists()
         check_tensorboard(
-            temp_dir + "/runs", "train/loss", 1.2, "Train Loss (%s) is too high"
+            temp_dir + "/runs", "train/loss", 1.0, "Train Loss is too high"
         )

tests/e2e/integrations/test_llm_compressor.py

@@ -1,111 +0,0 @@
-"""
-E2E smoke tests for LLMCompressorPlugin integration
-"""
-
-from pathlib import Path
-
-import pytest
-
-from axolotl.cli.args import TrainerCliArgs
-from axolotl.common.datasets import load_datasets
-from axolotl.train import train
-from axolotl.utils.config import normalize_config, prepare_plugins, validate_config
-from axolotl.utils.dict import DictDefault
-
-from tests.e2e.utils import (
-    check_model_output_exists,
-    require_llmcompressor,
-    require_torch_2_4_1,
-)
-
-MODELS = [
-    "nm-testing/llama2.c-stories42M-pruned2.4-compressed",
-    "nm-testing/llama2.c-stories42M-gsm8k-sparse-only-compressed",
-]
-
-
-@pytest.mark.parametrize(
-    "base_model", MODELS, ids=["no-checkpoint-recipe", "with-checkpoint-recipe"]
-)
-@pytest.mark.parametrize(
-    "save_compressed", [True, False], ids=["save_compressed", "save_uncompressed"]
-)
-class TestLLMCompressorIntegration:
-    """
-    e2e tests for axolotl.integrations.llm_compressor.LLMCompressorPlugin
-    """
-
-    @require_llmcompressor
-    @require_torch_2_4_1
-    def test_llmcompressor_plugin(
-        self, temp_dir, base_model: str, save_compressed: bool
-    ):
-        from llmcompressor import active_session
-
-        # core cfg
-        cfg = DictDefault(
-            {
-                "base_model": base_model,
-                "plugins": ["axolotl.integrations.llm_compressor.LLMCompressorPlugin"],
-                "sequence_len": 1024,
-                "val_set_size": 0.05,
-                "special_tokens": {"pad_token": "<|endoftext|>"},
-                "datasets": [{"path": "mhenrichsen/alpaca_2k_test", "type": "alpaca"}],
-                "num_epochs": 1,
-                "micro_batch_size": 2,
-                "gradient_accumulation_steps": 2,
-                "output_dir": temp_dir,
-                "learning_rate": 1e-5,
-                "optimizer": "adamw_torch_fused",
-                "lr_scheduler": "cosine",
-                "save_safetensors": True,
-                "bf16": "auto",
-                "max_steps": 5,
-                "llmcompressor": {
-                    "recipe": {
-                        "finetuning_stage": {
-                            "finetuning_modifiers": {
-                                "ConstantPruningModifier": {
-                                    "targets": [
-                                        "re:.*q_proj.weight",
-                                        "re:.*k_proj.weight",
-                                        "re:.*v_proj.weight",
-                                        "re:.*o_proj.weight",
-                                        "re:.*gate_proj.weight",
-                                        "re:.*up_proj.weight",
-                                        "re:.*down_proj.weight",
-                                    ],
-                                    "start": 0,
-                                },
-                            },
-                        },
-                    },
-                    "save_compressed": save_compressed,
-                },
-            }
-        )
-
-        prepare_plugins(cfg)
-        cfg = validate_config(cfg)
-        normalize_config(cfg)
-        cli_args = TrainerCliArgs()
-        dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
-
-        try:
-            train(cfg=cfg, dataset_meta=dataset_meta)
-            check_model_output_exists(temp_dir, cfg)
-            _check_llmcompressor_model_outputs(temp_dir, save_compressed)
-        finally:
-            active_session().reset()
-
-
-def _check_llmcompressor_model_outputs(temp_dir, save_compressed):
-    if save_compressed:
-        assert (Path(temp_dir) / "recipe.yaml").exists()
-
-        from compressed_tensors import ModelCompressor
-        from compressed_tensors.config import Sparse24BitMaskConfig
-
-        compressor = ModelCompressor.from_pretrained(temp_dir)
-        assert compressor is not None
-        assert isinstance(compressor.sparsity_config, Sparse24BitMaskConfig)

tests/e2e/utils.py

@@ -105,25 +105,7 @@ def require_vllm(test_case):
             return False
 
     return unittest.skipUnless(
-        is_vllm_installed(), "test requires vllm to be installed"
+        is_vllm_installed(), "test requires a vllm to be installed"
-    )(test_case)
-
-
-def require_llmcompressor(test_case):
-    """
-    Decorator marking a test that requires a llmcompressor to be installed
-    """
-
-    def is_llmcompressor_installed():
-        try:
-            import llmcompressor  # pylint: disable=unused-import  # noqa: F401
-
-            return True
-        except ImportError:
-            return False
-
-    return unittest.skipUnless(
-        is_llmcompressor_installed(), "test requires llmcompressor to be installed"
     )(test_case)
 
 

tests/test_packed_batch_sampler.py

@@ -106,4 +106,3 @@ class TestBatchedSamplerPacking:
 
         original_idxs = set(range(len(train_dataset)))
         assert original_idxs == set(batch_idxs)
-        assert len(batch_idxs) == len(set(batch_idxs))