speed up flash-attn inference

* Feat(doc): Improve sharegpt doc

* Fix typo

.github/FUNDING.yml

@@ -0,0 +1,13 @@
+# These are supported funding model platforms
+
+github: OpenAccess-AI-Collective # Replace with up to 4 GitHub Sponsors-enabled usernames e.g., [user1, user2]
+patreon: # Replace with a single Patreon username
+open_collective: # Replace with a single Open Collective username
+ko_fi: # Replace with a single Ko-fi username
+tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
+community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
+liberapay: # Replace with a single Liberapay username
+issuehunt: # Replace with a single IssueHunt username
+otechie: # Replace with a single Otechie username
+lfx_crowdfunding: # Replace with a single LFX Crowdfunding project-name e.g., cloud-foundry
+custom: # Replace with up to 4 custom sponsorship URLs e.g., ['link1', 'link2']

README.md

@@ -136,7 +136,7 @@ Have dataset(s) in one of the following format (JSONL recommended):
   ```json
   {"instruction": "...", "input": "...", "output": "..."}
   ```
-- `sharegpt:chat`: conversations
+- `sharegpt:chat`: conversations where `from` is `human`/`gpt`
   ```json
   {"conversations": [{"from": "...", "value": "..."}]}
   ```
@@ -225,6 +225,10 @@ Have dataset(s) in one of the following format (JSONL recommended):
   ```json
   {"conversations": [{"role": "...", "value": "..."}]}
   ```
+- `sharegpt_simple.load_guanaco`: conversations where `from` is `prompter`/`assistant` instead of default sharegpt
+  ```json
+  {"conversations": [{"from": "...", "value": "..."}]}
+  ```
 - `sharegpt_jokes`: creates a chat where bot is asked to tell a joke, then explain why the joke is funny
   ```json
   {"conversations": [{"title": "...", "text": "...", "explanation": "..."}]}
@@ -375,7 +379,14 @@ dataset_shard_idx:
 sequence_len: 2048
 # max sequence length to concatenate training samples together up to
 # inspired by StackLLaMA. see https://huggingface.co/blog/stackllama#supervised-fine-tuning
+# FutureWarning: This will soon be DEPRECATED
 max_packed_sequence_len: 1024
+# use efficient multi-packing with block diagonal attention and per sequence position_ids. Recommend set to 'true'
+sample_packing:
+# you can set these packing optimizations AFTER starting a training at least once.
+# The trainer will provide recommended values for these values.
+sample_packing_eff_est:
+total_num_tokens:
 
 # if you want to use 'lora' or 'qlora' or leave blank to train all parameters in original model
 adapter: lora
@@ -401,11 +412,12 @@ lora_out_dir:
 lora_fan_in_fan_out: false
 
 # wandb configuration if you're using it
-wandb_mode:
-wandb_project:
+wandb_mode: # "offline" to save run metadata locally and not sync to the server, "disabled" to turn off wandb
+wandb_project: # your wandb project name
+wandb_entity: # a wandb Team name if using a Team
 wandb_watch:
-wandb_run_id:
-wandb_log_model: # 'checkpoint'
+wandb_run_id: # set the name of your wandb run
+wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only at the end of training
 
 # where to save the finished model to
 output_dir: ./completed-model
@@ -420,13 +432,16 @@ learning_rate: 0.00003
 logging_steps:
 save_steps:
 eval_steps:
+save_total_limit:
 
 # save model as safetensors (require safetensors package)
 save_safetensors:
 
 # whether to mask out or include the human's prompt from the training labels
 train_on_inputs: false
-# don't use this, leads to wonky training (according to someone on the internet)
+# group similarly sized data to minimize padding
+# may be slower to start, as it must download and sort the entire dataset
+# note that training loss may have an oscillating pattern with this enabled
 group_by_length: false
 
 # Whether to use gradient checkpointing https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
@@ -472,6 +487,10 @@ landmark_attention:
 # xpos RoPE see https://github.com/kaiokendev/cutoff-len-is-context-len/blob/main/util/xpos_rope_llama_monkey_patch.py
 # llama only
 xpos_rope:
+# RoPE Scaling https://github.com/huggingface/transformers/pull/24653
+rope_scaling:
+  type: # linear | dynamic
+  factor: # float
 
 # resume from a specific checkpoint dir
 resume_from_checkpoint:
@@ -503,6 +522,9 @@ torchdistx_path:
 # Set padding for data collator to 'longest'
 collator_pad_to_longest:
 
+# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
+pretraining_dataset:
+
 # Debug mode
 debug:
 
@@ -522,7 +544,14 @@ Run
 accelerate launch scripts/finetune.py configs/your_config.yml
 ```
 
-#### Multi-GPU Config
+#### Multi-GPU
+
+You can optionally pre-tokenize dataset with the following before finetuning:
+```bash
+CUDA_VISIBLE_DEVICES="" accelerate ... --prepare_ds_only
+```
+
+##### Config
 
 - llama FSDP
 ```yaml
@@ -537,6 +566,18 @@ fsdp_config:
 
 - llama Deepspeed: append `ACCELERATE_USE_DEEPSPEED=true` in front of finetune command
 
+##### Weights & Biases Logging
+
+- wandb options
+```yaml
+wandb_mode:
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_run_id:
+wandb_log_model:
+```
+
 ### Inference
 
 Pass the appropriate flag to the train command:

deepspeed/zero3.json

@@ -37,18 +37,18 @@
       "lr": "auto",
       "betas": [
         0.9,
-        0.999
+        0.95
       ],
       "eps": 1e-8,
       "weight_decay": "auto"
     }
   },
   "scheduler": {
-    "type": "OneCycle",
+    "type": "WarmupLR",
     "params": {
-      "cycle_min_lr": 0.00001,
-      "cycle_max_lr": 0.00003,
-      "cycle_first_step_size": 120
+      "warmup_min_lr": "auto",
+      "warmup_max_lr": "auto",
+      "warmup_num_steps": "auto"
     }
   },
   "train_batch_size": "auto",

examples/cerebras/qlora.yml

@@ -23,6 +23,7 @@ lora_target_modules:
 lora_target_linear:
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:
@@ -35,7 +36,7 @@ torchdistx_path:
 lr_scheduler: cosine
 learning_rate: 0.0002
 train_on_inputs: false
-group_by_length: true
+group_by_length: false
 bf16: true
 fp16: false
 tf32: true

examples/falcon/config-7b-lora.yml

@@ -24,6 +24,7 @@ lora_target_modules:
 lora_target_linear: true
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/falcon/config-7b-qlora.yml

@@ -38,6 +38,7 @@ lora_target_linear: true
 lora_fan_in_fan_out:
 
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/falcon/config-7b.yml

@@ -24,6 +24,7 @@ lora_target_modules:
 lora_target_linear: true
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/gptj/qlora.yml

@@ -20,6 +20,7 @@ lora_target_modules:
 lora_target_linear: true
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:
@@ -32,7 +33,7 @@ torchdistx_path:
 lr_scheduler: cosine
 learning_rate: 0.0001
 train_on_inputs: false
-group_by_length: true
+group_by_length: false
 bf16: true
 fp16: false
 tf32: true

examples/gptq-lora-7b/config.yml

@@ -22,6 +22,7 @@ lora_target_modules:
   - v_proj
 lora_fan_in_fan_out: false
 wandb_project: llama-7b-lora-int4
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/jeopardy-bot/config.yml

@@ -18,6 +18,7 @@ lora_dropout:
 lora_target_modules:
 lora_fan_in_fan_out: false
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/llama-2/lora.yml

@@ -26,6 +26,7 @@ lora_target_linear: true
 lora_fan_in_fan_out:
 
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:
@@ -38,7 +39,7 @@ lr_scheduler: cosine
 learning_rate: 0.0002
 
 train_on_inputs: false
-group_by_length: true
+group_by_length: false
 bf16: true
 fp16: false
 tf32: false

examples/llama-2/qlora.yml

@@ -27,6 +27,7 @@ lora_target_linear: true
 lora_fan_in_fan_out:
 
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:
@@ -39,7 +40,7 @@ lr_scheduler: cosine
 learning_rate: 0.0002
 
 train_on_inputs: false
-group_by_length: true
+group_by_length: false
 bf16: true
 fp16: false
 tf32: false

examples/mpt-7b/config.yml

@@ -20,6 +20,7 @@ lora_target_modules:
   - v_proj
 lora_fan_in_fan_out: false
 wandb_project: mpt-alpaca-7b
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/openllama-3b/config.yml

@@ -22,6 +22,7 @@ lora_target_modules:
 lora_target_linear:
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/openllama-3b/lora.yml

@@ -28,6 +28,7 @@ lora_target_modules:
   - o_proj
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/openllama-3b/qlora.yml

@@ -22,6 +22,7 @@ lora_target_modules:
 lora_target_linear: true
 lora_fan_in_fan_out:
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:
@@ -34,7 +35,7 @@ torchdistx_path:
 lr_scheduler: cosine
 learning_rate: 0.0002
 train_on_inputs: false
-group_by_length: true
+group_by_length: false
 bf16: true
 fp16: false
 tf32: true

examples/pythia-12b/config.yml

@@ -23,6 +23,7 @@ lora_target_modules:
 lora_target_linear: true
 lora_fan_in_fan_out: true  # pythia/GPTNeoX lora specific
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/pythia/lora.yml

@@ -17,6 +17,7 @@ lora_target_modules:
 lora_target_linear:
 lora_fan_in_fan_out: true  # pythia/GPTNeoX lora specific
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/redpajama/config-3b.yml

@@ -21,6 +21,7 @@ lora_target_modules:
   - v_proj
 lora_fan_in_fan_out: false
 wandb_project: redpajama-alpaca-3b
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/replit-3b/config-lora.yml

@@ -20,6 +20,7 @@ lora_target_modules:
   - mlp_down
 lora_fan_in_fan_out:
 wandb_project: lora-replit
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

examples/xgen-7b/xgen-7b-8k-qlora.yml

@@ -37,6 +37,7 @@ lora_target_linear: true
 lora_fan_in_fan_out:
 
 wandb_project:
+wandb_entity:
 wandb_watch:
 wandb_run_id:
 wandb_log_model:

requirements.txt

@@ -1,6 +1,6 @@
 peft @ git+https://github.com/huggingface/peft.git
 transformers @ git+https://github.com/huggingface/transformers.git
-bitsandbytes>=0.39.0
+bitsandbytes>=0.41.1
 accelerate @ git+https://github.com/huggingface/accelerate@2a289f6108e77a77a4efffb3f6316bc98538413b
 addict
 fire
@@ -13,9 +13,12 @@ einops
 xformers
 optimum
 hf_transfer
+numba
+numpy==1.24.4
 # qlora things
 bert-score==0.3.13
 evaluate==0.4.0
 rouge-score==0.1.2
 scipy
 scikit-learn==1.2.2
+pynvml

scripts/finetune.py

@@ -18,12 +18,17 @@ from optimum.bettertransformer import BetterTransformer
 from transformers import GenerationConfig, TextStreamer
 
 from axolotl.logging_config import configure_logging
+from axolotl.utils.config import normalize_config, validate_config
 from axolotl.utils.data import load_prepare_datasets, load_pretraining_dataset
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.distributed import barrier, is_main_process
 from axolotl.utils.models import load_model, load_tokenizer
 from axolotl.utils.tokenization import check_dataset_labels
-from axolotl.utils.trainer import setup_trainer
-from axolotl.utils.validation import validate_config
+from axolotl.utils.trainer import (
+    calculate_total_num_steps,
+    process_datasets_for_packing,
+    setup_trainer,
+)
 from axolotl.utils.wandb import setup_wandb_env_vars
 
 project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
@@ -38,27 +43,6 @@ DEFAULT_DATASET_PREPARED_PATH = "last_run_prepared"
 os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
 
 
-def choose_device(cfg):
-    def get_device():
-        try:
-            if torch.cuda.is_available():
-                return f"cuda:{cfg.local_rank}"
-
-            if torch.backends.mps.is_available():
-                return "mps"
-
-            raise SystemError("No CUDA/mps device found")
-        except Exception:  # pylint: disable=broad-exception-caught
-            return "cpu"
-
-    cfg.device = get_device()
-    if cfg.device_map != "auto":
-        if cfg.device.startswith("cuda"):
-            cfg.device_map = {"": cfg.local_rank}
-        else:
-            cfg.device_map = {"": cfg.device}
-
-
 def get_multi_line_input() -> Optional[str]:
     print("Give me an instruction (Ctrl + D to finish): ")
     instruction = ""
@@ -188,36 +172,13 @@ def train(
 
     validate_config(cfg)
 
-    # setup some derived config / hyperparams
-    cfg.gradient_accumulation_steps = cfg.gradient_accumulation_steps or (
-        cfg.batch_size // cfg.micro_batch_size
-    )
-    cfg.batch_size = (
-        cfg.batch_size or cfg.micro_batch_size * cfg.gradient_accumulation_steps
-    )
-    cfg.world_size = int(os.environ.get("WORLD_SIZE", 1))
-    cfg.local_rank = int(os.environ.get("LOCAL_RANK", 0))
-    choose_device(cfg)
-    cfg.ddp = cfg.ddp if cfg.ddp is not None else cfg.world_size != 1
-    if cfg.ddp:
-        cfg.device_map = {"": int(os.environ.get("LOCAL_RANK", 0))}
-        cfg.batch_size = cfg.batch_size * cfg.world_size
+    normalize_config(cfg)
 
     setup_wandb_env_vars(cfg)
-    if cfg.device == "mps":
-        cfg.load_in_8bit = False
-        cfg.tf32 = False
-        if cfg.bf16:
-            cfg.fp16 = True
-        cfg.bf16 = False
-
-    if cfg.tf32:
-        torch.backends.cuda.matmul.allow_tf32 = True
 
     # load the tokenizer first
-    tokenizer_config = cfg.tokenizer_config or cfg.base_model_config
-    LOG.info(f"loading tokenizer... {tokenizer_config}")
-    tokenizer = load_tokenizer(tokenizer_config, cfg.tokenizer_type, cfg)
+    LOG.info(f"loading tokenizer... {cfg.tokenizer_config or cfg.base_model_config}")
+    tokenizer = load_tokenizer(cfg)
 
     if (
         check_not_in(["shard", "merge_lora"], kwargs) and not cfg.inference
@@ -231,12 +192,25 @@ def train(
                 cfg.pretraining_dataset,
                 tokenizer,
                 max_tokens=cfg.sequence_len,
-                seed=cfg.seed,
+                seed=cfg.seed or 42,
             )
             # https://discuss.huggingface.co/t/how-to-use-huggingface-trainer-streaming-datasets-without-wrapping-it-with-torchdatas-iterablewrapper/25230
             train_dataset = train_dataset.with_format("torch")
             eval_dataset = None
 
+        if is_main_process():
+            # process on rank 0 first so it gets cached so other ranks load from cache
+            train_dataset, eval_dataset = process_datasets_for_packing(
+                cfg, train_dataset, eval_dataset
+            )
+        barrier()
+        if not is_main_process():
+            train_dataset, eval_dataset = process_datasets_for_packing(
+                cfg, train_dataset, eval_dataset
+            )
+        barrier()
+        total_num_steps = calculate_total_num_steps(cfg, train_dataset, tokenizer)
+
     if cfg.debug or "debug" in kwargs:
         LOG.info("check_dataset_labels...")
         check_dataset_labels(
@@ -251,15 +225,10 @@ def train(
         return
 
     # Load the model and tokenizer
-    LOG.info("loading model and peft_config...")
-    model, peft_config = load_model(
-        cfg.base_model,
-        cfg.base_model_config,
-        cfg.model_type,
-        tokenizer,
-        cfg,
-        adapter=cfg.adapter,
-    )
+    LOG.info("loading model and (optionally) peft_config...")
+    model, peft_config = load_model(cfg, tokenizer)
+
+    safe_serialization = cfg.save_safetensors is True
 
     if "merge_lora" in kwargs and cfg.adapter is not None:
         LOG.info("running merge of LoRA with base model")
@@ -268,7 +237,11 @@ def train(
 
         if cfg.local_rank == 0:
             LOG.info("saving merged model")
-            model.save_pretrained(str(Path(cfg.output_dir) / "merged"))
+            model.save_pretrained(
+                str(Path(cfg.output_dir) / "merged"),
+                safe_serialization=safe_serialization,
+            )
+            tokenizer.save_pretrained(str(Path(cfg.output_dir) / "merged"))
         return
 
     if cfg.inference:
@@ -283,10 +256,12 @@ def train(
         return
 
     if "shard" in kwargs:
-        model.save_pretrained(cfg.output_dir)
+        model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
         return
 
-    trainer = setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer)
+    trainer = setup_trainer(
+        cfg, train_dataset, eval_dataset, model, tokenizer, total_num_steps
+    )
 
     model.config.use_cache = False
 
@@ -305,7 +280,7 @@ def train(
         def terminate_handler(_, __, model):
             if cfg.flash_optimum:
                 model = BetterTransformer.reverse(model)
-            model.save_pretrained(cfg.output_dir)
+            model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
             sys.exit(0)
 
         signal.signal(
@@ -332,6 +307,7 @@ def train(
 
     if not Path(cfg.output_dir).is_dir():
         os.makedirs(cfg.output_dir, exist_ok=True)
+    tokenizer.save_pretrained(cfg.output_dir)
     if cfg.flash_optimum:
         with torch.backends.cuda.sdp_kernel(
             enable_flash=True, enable_math=True, enable_mem_efficient=True
@@ -345,13 +321,11 @@ def train(
     # TODO do we need this fix? https://huggingface.co/docs/accelerate/usage_guides/fsdp#saving-and-loading
     # only save on rank 0, otherwise it corrupts output on multi-GPU when multiple processes attempt to write the same file
     if cfg.fsdp:
-        model.save_pretrained(cfg.output_dir)
+        trainer.save_model(cfg.output_dir)
     elif cfg.local_rank == 0:
         if cfg.flash_optimum:
             model = BetterTransformer.reverse(model)
-        model.save_pretrained(cfg.output_dir)
-
-    # trainer.save_model(cfg.output_dir)  # TODO this may be needed for deepspeed to work? need to review another time
+        model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
 
 
 if __name__ == "__main__":

src/axolotl/datasets.py

@@ -5,7 +5,7 @@ import os
 from typing import List
 
 import torch
-from datasets import IterableDataset
+from datasets import Dataset, IterableDataset
 
 from .prompt_tokenizers import PromptTokenizingStrategy
 
@@ -18,9 +18,9 @@ from .prompt_tokenizers import PromptTokenizingStrategy
 LOG = logging.getLogger("axolotl")
 
 
-class TokenizedPromptDataset(IterableDataset):
+class TokenizedPromptDataset(Dataset):
     """
-    Iterable dataset that returns tokenized prompts from a stream of text files.
+    Dataset that returns tokenized prompts from a stream of text files.
         Args:
             prompt_tokenizer (PromptTokenizingStrategy): The prompt tokenizing method for proccessing the data.
             dataset (dataset.Dataset): Dataset with text files.
@@ -30,19 +30,18 @@ class TokenizedPromptDataset(IterableDataset):
         self,
         prompt_tokenizer: PromptTokenizingStrategy,
         dataset: IterableDataset,
+        **kwargs,
     ):
         self.prompt_tokenizer = prompt_tokenizer
-        self.dataset = dataset
+        super().__init__(self.process(dataset).data, **kwargs)
 
-    def __iter__(self):
-        features = self.dataset.features.keys()
-        num_proc = os.cpu_count()
-        return iter(
-            self.dataset.map(
-                self.prompt_tokenizer.tokenize_prompt,
-                num_proc=num_proc,
-                remove_columns=features,
-            )
+    def process(self, dataset):
+        features = dataset.features.keys()
+        num_proc = min(64, os.cpu_count())
+        return dataset.map(
+            self.prompt_tokenizer.tokenize_prompt,
+            num_proc=num_proc,
+            remove_columns=features,
         )
 
 
@@ -77,14 +76,21 @@ class ConstantLengthDataset(IterableDataset):
             self.tokens_dtype = torch.int64
 
     def __iter__(self):
-        buffer = {"input_ids": [], "attention_mask": [], "labels": []}
+        buffer = {
+            "input_ids": [],
+            "attention_mask": [],
+            "labels": [],
+            "position_ids": [],
+        }
         buffer_len = 0
         for dataset in self.datasets:
+            idx = 0
             iterator = iter(dataset)
             more_examples = True
             while more_examples:
                 try:
                     example = next(iterator)
+                    idx += 1
                 except StopIteration:
                     more_examples = False
                     example = None
@@ -106,6 +112,9 @@ class ConstantLengthDataset(IterableDataset):
                         attention_mask = torch.cat(buffer["attention_mask"], dim=-1)[
                             : self.seq_length
                         ]
+                        position_ids = torch.cat(buffer["position_ids"], dim=-1)[
+                            : self.seq_length
+                        ]
                         labels = torch.cat(buffer["labels"], dim=-1)[: self.seq_length]
                         if labels.size() == input_ids.size() and (
                             attention_mask.size() == input_ids.size()
@@ -114,6 +123,7 @@ class ConstantLengthDataset(IterableDataset):
                                 "input_ids": input_ids,
                                 "labels": labels,
                                 "attention_mask": attention_mask,
+                                "position_ids": position_ids,
                             }
                         else:
                             LOG.warning(
@@ -123,8 +133,10 @@ class ConstantLengthDataset(IterableDataset):
                         "input_ids": [],
                         "attention_mask": [],
                         "labels": [],
+                        "position_ids": [],
                     }
                     buffer_len = 0
+                    idx = 1
 
                 if example:
                     # FIXME
@@ -133,11 +145,6 @@ class ConstantLengthDataset(IterableDataset):
                         input_ids = example["input_ids"]
                         attention_mask = example["attention_mask"]
                         labels = example["labels"]
-                        if (
-                            buffer["input_ids"]
-                            and input_ids[0] == self.tokenizer.bos_token_id
-                        ):
-                            attention_mask[0] = 0
 
                         if add_concat_token:
                             input_ids.append(self.concat_token_id)
@@ -148,13 +155,17 @@ class ConstantLengthDataset(IterableDataset):
                             input_ids, dtype=self.tokens_dtype
                         )
                         attention_mask_with_concat = torch.tensor(
-                            attention_mask, dtype=self.tokens_dtype
+                            [idx * m for m in attention_mask], dtype=torch.int16
                         )
                         labels_with_concat = torch.tensor(
                             labels, dtype=self.tokens_dtype
                         )
+                        position_ids = torch.arange(
+                            len(input_ids), dtype=self.tokens_dtype
+                        )
 
                         buffer["input_ids"].append(input_ids_with_concat)
                         buffer["attention_mask"].append(attention_mask_with_concat)
                         buffer["labels"].append(labels_with_concat)
+                        buffer["position_ids"].append(position_ids)
                         buffer_len += len(input_ids)

src/axolotl/monkeypatch/llama_attn_hijack_flash.py

@@ -2,122 +2,38 @@
 
 # copied from https://github.com/lm-sys/FastChat/blob/main/fastchat/train/llama_flash_attn_monkey_patch.py
 
+import warnings
 from typing import Optional, Tuple
 
 import torch
+import torch.nn.functional as F
 import transformers
 from einops import rearrange
 from flash_attn.bert_padding import pad_input, unpad_input
-from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func
-from transformers.models.llama.modeling_llama import apply_rotary_pos_emb
+from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, repeat_kv
 
+from axolotl.monkeypatch.utils import get_cu_seqlens_from_pos_ids
 
-def forward(
-    self,
-    hidden_states: torch.Tensor,
-    attention_mask: Optional[torch.Tensor] = None,
-    position_ids: Optional[torch.Tensor] = None,
-    past_key_value: Optional[Tuple[torch.Tensor]] = None,
-    output_attentions: bool = False,
-    use_cache: bool = False,
-) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
-    """Input shape: Batch x Time x Channel
-
-    attention_mask: [bsz, q_len]
-    """
-    # pylint: disable=duplicate-code
-    bsz, q_len, _ = hidden_states.size()
-
-    query_states = (
-        self.q_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
+try:
+    from flash_attn.flash_attn_interface import (  # pylint: disable=ungrouped-imports
+        flash_attn_kvpacked_func,
+        flash_attn_varlen_kvpacked_func,
+        flash_attn_varlen_qkvpacked_func,
     )
-    key_states = (
-        self.k_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
+except ImportError:
+    from flash_attn.flash_attn_interface import (
+        flash_attn_unpadded_kvpacked_func as flash_attn_varlen_kvpacked_func,
     )
-    value_states = (
-        self.v_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
+    from flash_attn.flash_attn_interface import (
+        flash_attn_unpadded_qkvpacked_func as flash_attn_varlen_qkvpacked_func,
     )
-    # [bsz, q_len, nh, hd]
-    # [bsz, nh, q_len, hd]
 
-    kv_seq_len = key_states.shape[-2]
-    assert past_key_value is None, "past_key_value is not supported"
 
-    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
-    query_states, key_states = apply_rotary_pos_emb(
-        query_states, key_states, cos, sin, position_ids
-    )
-    # [bsz, nh, t, hd]
-    assert not output_attentions, "output_attentions is not supported"
-    assert not use_cache, "use_cache is not supported"
-
-    # Flash attention codes from
-    # https://github.com/HazyResearch/flash-attention/blob/main/flash_attn/flash_attention.py
-
-    # transform the data into the format required by flash attention
-    qkv = torch.stack(
-        [query_states, key_states, value_states], dim=2
-    )  # [bsz, nh, 3, q_len, hd]
-    qkv = qkv.transpose(1, 3)  # [bsz, q_len, 3, nh, hd]
-    # We have disabled _prepare_decoder_attention_mask in LlamaModel
-    # the attention_mask should be the same as the key_padding_mask
-    key_padding_mask = attention_mask
-
-    if key_padding_mask is None:
-        qkv = rearrange(qkv, "b s ... -> (b s) ...")
-        max_s = q_len
-        cu_q_lens = torch.arange(
-            0,
-            (bsz + 1) * q_len,
-            step=q_len,
-            dtype=torch.int32,
-            device=qkv.device,
-        )
-        output = flash_attn_varlen_qkvpacked_func(
-            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
-        )
-        output = rearrange(output, "(b s) ... -> b s ...", b=bsz)
-    else:
-        nheads = qkv.shape[-2]
-
-        # pylint: disable=invalid-name
-        x = rearrange(qkv, "b s three h d -> b s (three h d)")
-        x_unpad, indices, cu_q_lens, max_s = unpad_input(x, key_padding_mask)
-        x_unpad = rearrange(
-            x_unpad,
-            "nnz (three h d) -> nnz three h d",
-            three=3,
-            h=nheads,
-        )
-        output_unpad = flash_attn_varlen_qkvpacked_func(
-            x_unpad,
-            cu_q_lens,
-            max_s,
-            0.0,
-            softmax_scale=None,
-            causal=True,
-        )
-        output = rearrange(
-            pad_input(
-                rearrange(output_unpad, "nnz h d -> nnz (h d)"),
-                indices,
-                bsz,
-                q_len,
-            ),
-            "b s (h d) -> b s h d",
-            h=nheads,
-        )
-    return (
-        self.o_proj(rearrange(output, "b s h d -> b s (h d)")),
-        None,
-        None,
+def replace_llama_attn_with_flash_attn():
+    transformers.models.llama.modeling_llama.LlamaModel._prepare_decoder_attention_mask = (  # pylint: disable=protected-access
+        _prepare_decoder_attention_mask
     )
+    transformers.models.llama.modeling_llama.LlamaAttention.forward = flashattn_forward
 
 
 # Disable the transformation of the attention mask in LlamaModel as the flash attention
@@ -133,8 +49,310 @@ def _prepare_decoder_attention_mask(
     return attention_mask
 
 
-def replace_llama_attn_with_flash_attn():
-    transformers.models.llama.modeling_llama.LlamaModel._prepare_decoder_attention_mask = (  # pylint: disable=protected-access
-        _prepare_decoder_attention_mask
+def flashattn_forward(
+    self,
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.Tensor] = None,
+    past_key_value: Optional[Tuple[torch.Tensor]] = None,
+    output_attentions: bool = False,
+    use_cache: bool = False,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    """Input shape: Batch x Time x Channel
+
+    attention_mask: [bsz, q_len]
+    """
+    # pylint: disable=duplicate-code
+    bsz, q_len, _ = hidden_states.size()
+
+    if not hasattr(self, "pretraining_tp"):
+        self.pretraining_tp = 1
+
+    if self.pretraining_tp > 1:
+        key_value_slicing = (
+            self.num_key_value_heads * self.head_dim
+        ) // self.pretraining_tp
+        query_slices = self.q_proj.weight.split(
+            (self.num_heads * self.head_dim) // self.pretraining_tp, dim=0
+        )
+        key_slices = self.k_proj.weight.split(key_value_slicing, dim=0)
+        value_slices = self.v_proj.weight.split(key_value_slicing, dim=0)
+
+        query_states = [
+            F.linear(hidden_states, query_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        query_states = torch.cat(query_states, dim=-1)
+
+        key_states = [
+            F.linear(hidden_states, key_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        key_states = torch.cat(key_states, dim=-1)
+
+        value_states = [
+            F.linear(hidden_states, value_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        value_states = torch.cat(value_states, dim=-1)
+
+    else:
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+    query_states = query_states.view(
+        bsz, q_len, self.num_heads, self.head_dim
+    ).transpose(1, 2)
+    key_states = key_states.view(
+        bsz, q_len, self.num_key_value_heads, self.head_dim
+    ).transpose(1, 2)
+    value_states = value_states.view(
+        bsz, q_len, self.num_key_value_heads, self.head_dim
+    ).transpose(1, 2)
+    # [bsz, q_len, nh, hd]
+    # [bsz, nh, q_len, hd]
+
+    kv_seq_len = key_states.shape[-2]
+    if past_key_value is not None:
+        kv_seq_len += past_key_value[0].shape[-2]
+
+    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+    query_states, key_states = apply_rotary_pos_emb(
+        query_states, key_states, cos, sin, position_ids
+    )
+    # [bsz, nh, t, hd]
+
+    if past_key_value is not None:
+        # reuse k, v, self_attention
+        key_states = torch.cat([past_key_value[0], key_states], dim=2)
+        value_states = torch.cat([past_key_value[1], value_states], dim=2)
+
+    past_key_value = (key_states, value_states) if use_cache else None
+
+    # repeat k/v heads if n_kv_heads < n_heads
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    if output_attentions:
+        warnings.warn(
+            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
+        )
+
+    #
+    # flash-attn v2 start
+    #
+
+    if self.training:
+        # during training q,k,v always have same seqlen
+        assert key_states.shape == query_states.shape
+        is_causal = True
+    else:
+        # turn off FA causal mask after first inference autoregressive iteration
+        # only on first autoregressive step q,k,v have same seqlen
+        is_causal = past_key_value is not None
+
+    if self.training and attention_mask.shape[0] == 1:
+        # special handling using sample packing
+        qkv = torch.stack(
+            [query_states, key_states, value_states], dim=2
+        )  # [bsz, nh, 3, q_len, hd]
+        qkv = qkv.transpose(1, 3)  # [bsz, q_len, 3, nh, hd]
+        qkv = rearrange(qkv, "b s ... -> (b s) ...")
+        cu_q_lens, max_s = get_cu_seqlens_from_pos_ids(position_ids)
+        cu_q_lens = cu_q_lens.squeeze()
+
+        output = flash_attn_varlen_qkvpacked_func(
+            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=is_causal
+        )
+        output = rearrange(output, "(b s) ... -> b s ...", b=bsz)
+    elif query_states.shape == key_states.shape:
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        qkv_unpad, cu_seqlens_q, max_seqlen_q, _, output_pad_fn = generate_qkv(
+            query_states,
+            key_states,
+            value_states,
+            qkvpacked=True,
+            # We have disabled _prepare_decoder_attention_mask in LlamaModel
+            # the attention_mask should be the same as the key_padding_mask
+            key_padding_mask=attention_mask,
+            query_padding_mask=attention_mask[:, -query_states.size(1) :]
+            if attention_mask is not None
+            else None,
+        )
+        output_unpad = flash_attn_varlen_qkvpacked_func(
+            qkv_unpad,
+            cu_seqlens_q,
+            max_seqlen_q,
+            0.0,
+            softmax_scale=None,
+            causal=is_causal,
+        )
+        output = output_pad_fn(output_unpad)
+    else:
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        if attention_mask is None or attention_mask.all().item():
+            output = flash_attn_kvpacked_func(
+                query_states,
+                torch.stack([key_states, value_states], 2),
+                causal=is_causal,
+            )
+        else:
+            (  # pylint: disable=unbalanced-tuple-unpacking
+                q_unpad,
+                kv_unpad,
+                cu_seqlens_q,
+                cu_seqlens_k,
+                max_seqlen_q,
+                max_seqlen_k,
+                _,
+                _,
+                output_pad_fn,
+            ) = generate_qkv(
+                query_states,
+                key_states,
+                value_states,
+                kvpacked=True,
+                key_padding_mask=attention_mask,
+                query_padding_mask=attention_mask[:, -query_states.size(1) :]
+                if attention_mask is not None
+                else None,
+            )
+            output_unpad = flash_attn_varlen_kvpacked_func(
+                q_unpad,
+                kv_unpad,
+                cu_seqlens_q,
+                cu_seqlens_k,
+                max_seqlen_q,
+                max_seqlen_k,
+                0.0,
+                softmax_scale=None,
+                causal=is_causal,
+            )
+            output = output_pad_fn(output_unpad)
+
+    attn_output = output
+    if attn_output.size() != (bsz, q_len, self.num_heads, self.head_dim):
+        raise ValueError(
+            f"`attn_output` should be of size {(bsz, q_len, self.num_heads, self.head_dim)}, but is"
+            f" {attn_output.size()}"
+        )
+    attn_output = rearrange(attn_output, "b s h d -> b s (h d)")
+
+    #
+    # flash-attn v2 end
+    #
+
+    if self.pretraining_tp > 1:
+        attn_output = attn_output.split(self.hidden_size // self.pretraining_tp, dim=2)
+        o_proj_slices = self.o_proj.weight.split(
+            self.hidden_size // self.pretraining_tp, dim=1
+        )
+        attn_output = sum(
+            F.linear(attn_output[i], o_proj_slices[i])
+            for i in range(self.pretraining_tp)
+        )
+    else:
+        attn_output = self.o_proj(attn_output)
+
+    return attn_output, None, past_key_value
+
+
+# based on https://github.com/Dao-AILab/flash-attention/blob/364a5b/tests/test_flash_attn.py#L38
+def generate_qkv(
+    q,
+    k,
+    v,
+    query_padding_mask=None,
+    key_padding_mask=None,
+    kvpacked=False,
+    qkvpacked=False,
+):  # pylint: disable=invalid-name,unnecessary-lambda-assignment
+    """
+    Arguments:
+        q: (batch_size, seqlen_q, nheads, d)
+        k: (batch_size, seqlen_k, nheads_k, d)
+        v: (batch_size, seqlen_k, nheads_k, d)
+        query_padding_mask: (batch_size, seqlen), bool
+        key_padding_mask: (batch_size, seqlen), bool
+    """
+    assert not (kvpacked and qkvpacked)
+    batch_size, seqlen_q, nheads, d = q.shape
+    _, seqlen_k, nheads_k, _ = k.shape
+    assert k.shape == (batch_size, seqlen_k, nheads_k, d)
+    assert v.shape == (batch_size, seqlen_k, nheads_k, d)
+
+    if query_padding_mask is not None:
+        q_unpad, indices_q, cu_seqlens_q, max_seqlen_q = unpad_input(
+            q, query_padding_mask
+        )
+
+        output_pad_fn = lambda output_unpad: pad_input(  # noqa: E731
+            output_unpad, indices_q, batch_size, seqlen_q
+        )
+
+    else:
+        q_unpad = rearrange(q, "b s h d -> (b s) h d")
+        cu_seqlens_q = torch.arange(
+            0,
+            (batch_size + 1) * seqlen_q,
+            step=seqlen_q,
+            dtype=torch.int32,
+            device=q_unpad.device,
+        )
+        max_seqlen_q = seqlen_q
+
+        output_pad_fn = lambda output_unpad: rearrange(  # noqa: E731
+            output_unpad, "(b s) h d -> b s h d", b=batch_size
+        )
+
+    if key_padding_mask is not None:
+        k_unpad, _, cu_seqlens_k, max_seqlen_k = unpad_input(k, key_padding_mask)
+        v_unpad, _, _, _ = unpad_input(v, key_padding_mask)
+    else:
+        k_unpad = rearrange(k, "b s h d -> (b s) h d")
+        v_unpad = rearrange(v, "b s h d -> (b s) h d")
+        cu_seqlens_k = torch.arange(
+            0,
+            (batch_size + 1) * seqlen_k,
+            step=seqlen_k,
+            dtype=torch.int32,
+            device=k_unpad.device,
+        )
+        max_seqlen_k = seqlen_k
+
+    if qkvpacked:
+        assert nheads == nheads_k
+        qkv_unpad = torch.stack([q_unpad, k_unpad, v_unpad], dim=1)
+        qkv = torch.stack([q, k, v], dim=2)
+        return (qkv_unpad, cu_seqlens_q, max_seqlen_q, qkv, output_pad_fn)
+
+    if kvpacked:
+        kv_unpad = torch.stack([k_unpad, v_unpad], dim=1)
+        kv = torch.stack([k, v], dim=2)
+        return (
+            q_unpad,
+            kv_unpad,
+            cu_seqlens_q,
+            cu_seqlens_k,
+            max_seqlen_q,
+            max_seqlen_k,
+            q,
+            kv,
+            output_pad_fn,
+        )
+
+    return (
+        q_unpad,
+        k_unpad,
+        v_unpad,
+        cu_seqlens_q,
+        cu_seqlens_k,
+        max_seqlen_q,
+        max_seqlen_k,
+        q,
+        k,
+        v,
+        output_pad_fn,
     )
-    transformers.models.llama.modeling_llama.LlamaAttention.forward = forward

src/axolotl/monkeypatch/llama_attn_hijack_sdp.py

@@ -0,0 +1,140 @@
+"""
+Patched LlamaAttention to use torch.nn.functional.scaled_dot_product_attention
+"""
+
+import warnings
+from typing import Optional, Tuple
+
+import torch
+import torch.nn.functional as F
+import transformers.models.llama.modeling_llama
+from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, repeat_kv
+
+
+def hijack_llama_sdp_attention():
+    transformers.models.llama.modeling_llama.LlamaAttention.forward = (
+        sdp_attention_forward
+    )
+
+
+def sdp_attention_forward(
+    self,
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.LongTensor] = None,
+    past_key_value: Optional[Tuple[torch.Tensor]] = None,
+    output_attentions: bool = False,
+    use_cache: bool = False,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    # pylint: disable=duplicate-code
+    bsz, q_len, _ = hidden_states.size()
+
+    if not hasattr(self, "pretraining_tp"):
+        self.pretraining_tp = 1
+
+    if self.pretraining_tp > 1:
+        key_value_slicing = (
+            self.num_key_value_heads * self.head_dim
+        ) // self.pretraining_tp
+        query_slices = self.q_proj.weight.split(
+            (self.num_heads * self.head_dim) // self.pretraining_tp, dim=0
+        )
+        key_slices = self.k_proj.weight.split(key_value_slicing, dim=0)
+        value_slices = self.v_proj.weight.split(key_value_slicing, dim=0)
+
+        query_states = [
+            F.linear(hidden_states, query_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        query_states = torch.cat(query_states, dim=-1)
+
+        key_states = [
+            F.linear(hidden_states, key_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        key_states = torch.cat(key_states, dim=-1)
+
+        value_states = [
+            F.linear(hidden_states, value_slices[i]) for i in range(self.pretraining_tp)
+        ]
+        value_states = torch.cat(value_states, dim=-1)
+
+    else:
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+    query_states = query_states.view(
+        bsz, q_len, self.num_heads, self.head_dim
+    ).transpose(1, 2)
+    key_states = key_states.view(
+        bsz, q_len, self.num_key_value_heads, self.head_dim
+    ).transpose(1, 2)
+    value_states = value_states.view(
+        bsz, q_len, self.num_key_value_heads, self.head_dim
+    ).transpose(1, 2)
+    # [bsz, q_len, nh, hd]
+    # [bsz, nh, q_len, hd]
+
+    kv_seq_len = key_states.shape[-2]
+    if past_key_value is not None:
+        kv_seq_len += past_key_value[0].shape[-2]
+
+    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+    query_states, key_states = apply_rotary_pos_emb(
+        query_states, key_states, cos, sin, position_ids
+    )
+    # [bsz, nh, t, hd]
+
+    if past_key_value is not None:
+        # reuse k, v, self_attention
+        key_states = torch.cat([past_key_value[0], key_states], dim=2)
+        value_states = torch.cat([past_key_value[1], value_states], dim=2)
+
+    past_key_value = (key_states, value_states) if use_cache else None
+
+    # repeat k/v heads if n_kv_heads < n_heads
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    if output_attentions:
+        warnings.warn(
+            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
+        )
+
+    #
+    # sdp-attn start
+    #
+
+    with torch.backends.cuda.sdp_kernel():
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask,
+            is_causal=False,
+        )
+
+    if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+        raise ValueError(
+            f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+            f" {attn_output.size()}"
+        )
+    attn_output = attn_output.transpose(1, 2)
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+    #
+    # sdp-attn end
+    #
+
+    if self.pretraining_tp > 1:
+        attn_output = attn_output.split(self.hidden_size // self.pretraining_tp, dim=2)
+        o_proj_slices = self.o_proj.weight.split(
+            self.hidden_size // self.pretraining_tp, dim=1
+        )
+        attn_output = sum(
+            F.linear(attn_output[i], o_proj_slices[i])
+            for i in range(self.pretraining_tp)
+        )
+    else:
+        attn_output = self.o_proj(attn_output)
+
+    return attn_output, None, past_key_value

src/axolotl/monkeypatch/llama_attn_hijack_xformers.py

@@ -3,13 +3,13 @@ Directly copied the code from https://raw.githubusercontent.com/oobabooga/text-g
 """
 
 import logging
-import math
+import warnings
 from typing import Optional, Tuple
 
 import torch
 import torch.nn.functional as F
 import transformers.models.llama.modeling_llama
-from torch import nn
+from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, repeat_kv
 
 try:
     import xformers.ops
@@ -21,12 +21,6 @@ def hijack_llama_attention():
     transformers.models.llama.modeling_llama.LlamaAttention.forward = xformers_forward
 
 
-def hijack_llama_sdp_attention():
-    transformers.models.llama.modeling_llama.LlamaAttention.forward = (
-        sdp_attention_forward
-    )
-
-
 def xformers_forward(
     self,
     hidden_states: torch.Tensor,
@@ -81,15 +75,15 @@ def xformers_forward(
     value_states = value_states.view(
         bsz, q_len, self.num_key_value_heads, self.head_dim
     ).transpose(1, 2)
+    # [bsz, q_len, nh, hd]
+    # [bsz, nh, q_len, hd]
 
     kv_seq_len = key_states.shape[-2]
     if past_key_value is not None:
         kv_seq_len += past_key_value[0].shape[-2]
+
     cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
-    (
-        query_states,
-        key_states,
-    ) = transformers.models.llama.modeling_llama.apply_rotary_pos_emb(
+    query_states, key_states = apply_rotary_pos_emb(
         query_states, key_states, cos, sin, position_ids
     )
     # [bsz, nh, t, hd]
@@ -102,73 +96,50 @@ def xformers_forward(
     past_key_value = (key_states, value_states) if use_cache else None
 
     # repeat k/v heads if n_kv_heads < n_heads
-    key_states = transformers.models.llama.modeling_llama.repeat_kv(
-        key_states, self.num_key_value_groups
-    )
-    value_states = transformers.models.llama.modeling_llama.repeat_kv(
-        value_states, self.num_key_value_groups
-    )
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
 
-    # We only apply xformers optimizations if we don't need to output the whole attention matrix
-    if not output_attentions:
-        query_states = query_states.transpose(1, 2)
-        key_states = key_states.transpose(1, 2)
-        value_states = value_states.transpose(1, 2)
+    if output_attentions:
+        warnings.warn(
+            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
+        )
 
-        # This is a nasty hack. We know attention_mask in transformers is either LowerTriangular or all Zeros.
-        # We therefore check if one element in the upper triangular portion is zero. If it is, then the mask is all zeros.
-        if attention_mask is None or attention_mask[0, 0, 0, 1] == 0:
-            # input and output should be of form (bsz, q_len, num_heads, head_dim)
-            attn_output = xformers.ops.memory_efficient_attention(
-                query_states, key_states, value_states, attn_bias=None
-            )
-        else:
-            # input and output should be of form (bsz, q_len, num_heads, head_dim)
-            attn_output = xformers.ops.memory_efficient_attention(
-                query_states,
-                key_states,
-                value_states,
-                attn_bias=xformers.ops.LowerTriangularMask(),
-            )
-        attn_weights = None
+    #
+    # xformers-attn start
+    #
+
+    query_states = query_states.transpose(1, 2)
+    key_states = key_states.transpose(1, 2)
+    value_states = value_states.transpose(1, 2)
+
+    # This is a nasty hack. We know attention_mask in transformers is either LowerTriangular or all Zeros.
+    # We therefore check if one element in the upper triangular portion is zero. If it is, then the mask is all zeros.
+    if attention_mask is None or attention_mask[0, 0, 0, 1] == 0:
+        # input and output should be of form (bsz, q_len, num_heads, head_dim)
+        attn_output = xformers.ops.memory_efficient_attention(
+            query_states, key_states, value_states, attn_bias=None
+        )
     else:
-        attn_weights = torch.matmul(
-            query_states, key_states.transpose(2, 3)
-        ) / math.sqrt(self.head_dim)
-
-        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
-            raise ValueError(
-                f"Attention weights should be of size {(bsz * self.num_heads, q_len, kv_seq_len)}, but is"
-                f" {attn_weights.size()}"
-            )
-
-        if attention_mask is not None:
-            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
-                raise ValueError(
-                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
-                )
-            attn_weights = attn_weights + attention_mask
-            attn_weights = torch.max(
-                attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min)
-            )
-
-        # upcast attention to fp32
-        attn_weights = nn.functional.softmax(
-            attn_weights, dim=-1, dtype=torch.float32
-        ).to(query_states.dtype)
-        attn_output = torch.matmul(attn_weights, value_states)
-
-        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
-            raise ValueError(
-                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
-                f" {attn_output.size()}"
-            )
-
-        attn_output = attn_output.transpose(1, 2).contiguous()
-        # end x-formers vs. not x-formers if-else block
+        # input and output should be of form (bsz, q_len, num_heads, head_dim)
+        attn_output = xformers.ops.memory_efficient_attention(
+            query_states,
+            key_states,
+            value_states,
+            # attn_bias=attention_mask,
+            attn_bias=xformers.ops.LowerTriangularMask(),
+        )
 
+    if attn_output.size() != (bsz, q_len, self.num_heads, self.head_dim):
+        raise ValueError(
+            f"`attn_output` should be of size {(bsz, q_len, self.num_heads, self.head_dim)}, but is"
+            f" {attn_output.size()}"
+        )
     attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
 
+    #
+    # xformers-attn end
+    #
+
     if self.pretraining_tp > 1:
         attn_output = attn_output.split(self.hidden_size // self.pretraining_tp, dim=2)
         o_proj_slices = self.o_proj.weight.split(
@@ -181,103 +152,4 @@ def xformers_forward(
     else:
         attn_output = self.o_proj(attn_output)
 
-    return attn_output, attn_weights, past_key_value
-
-
-def sdp_attention_forward(
-    self,
-    hidden_states: torch.Tensor,
-    attention_mask: Optional[torch.Tensor] = None,
-    position_ids: Optional[torch.LongTensor] = None,
-    past_key_value: Optional[Tuple[torch.Tensor]] = None,
-    output_attentions: bool = False,
-    use_cache: bool = False,
-) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
-    # pylint: disable=duplicate-code
-    bsz, q_len, _ = hidden_states.size()
-
-    query_states = (
-        self.q_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
-    )
-    key_states = (
-        self.k_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
-    )
-    value_states = (
-        self.v_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim)
-        .transpose(1, 2)
-    )
-
-    kv_seq_len = key_states.shape[-2]
-    if past_key_value is not None:
-        kv_seq_len += past_key_value[0].shape[-2]
-    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
-    (
-        query_states,
-        key_states,
-    ) = transformers.models.llama.modeling_llama.apply_rotary_pos_emb(
-        query_states, key_states, cos, sin, position_ids
-    )
-    # [bsz, nh, t, hd]
-
-    if past_key_value is not None:
-        # reuse k, v, self_attention
-        key_states = torch.cat([past_key_value[0], key_states], dim=2)
-        value_states = torch.cat([past_key_value[1], value_states], dim=2)
-
-    past_key_value = (key_states, value_states) if use_cache else None
-
-    # We only apply sdp attention if we don't need to output the whole attention matrix
-    if not output_attentions:
-        with torch.backends.cuda.sdp_kernel():
-            attn_output = torch.nn.functional.scaled_dot_product_attention(
-                query_states,
-                key_states,
-                value_states,
-                attn_mask=attention_mask,
-                is_causal=False,
-            )
-            attn_weights = None
-    else:
-        attn_weights = torch.matmul(
-            query_states, key_states.transpose(2, 3)
-        ) / math.sqrt(self.head_dim)
-
-        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
-            raise ValueError(
-                f"Attention weights should be of size {(bsz * self.num_heads, q_len, kv_seq_len)}, but is"
-                f" {attn_weights.size()}"
-            )
-
-        if attention_mask is not None:
-            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
-                raise ValueError(
-                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
-                )
-            attn_weights = attn_weights + attention_mask
-            attn_weights = torch.max(
-                attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min)
-            )
-
-        # upcast attention to fp32
-        attn_weights = nn.functional.softmax(
-            attn_weights, dim=-1, dtype=torch.float32
-        ).to(query_states.dtype)
-        attn_output = torch.matmul(attn_weights, value_states)
-
-        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
-            raise ValueError(
-                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
-                f" {attn_output.size()}"
-            )
-
-    attn_output = attn_output.transpose(1, 2)
-    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
-
-    attn_output = self.o_proj(attn_output)
-
-    return attn_output, attn_weights, past_key_value
+    return attn_output, None, past_key_value

src/axolotl/monkeypatch/llama_expand_mask.py

@@ -0,0 +1,52 @@
+"""
+expands the binary attention mask per 3.2.2 of https://arxiv.org/pdf/2107.02027.pdf
+"""
+from typing import Optional
+
+import torch
+
+
+def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
+    """
+    Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
+    This expansion handles packed sequences so that sequences share the same attention mask integer value
+    when they attend to each other within that sequence.
+    This expansion transforms the mask to lower triangular form to prevent future peeking.
+    """
+    bsz, src_len = mask.size()
+    tgt_len = tgt_len if tgt_len is not None else src_len
+
+    mask = mask.unsqueeze(1).unsqueeze(2)
+    mask = mask.expand(bsz, 1, tgt_len, src_len)
+
+    # Create a binary mask from the original mask where zeros remain zeros and all other values are set to one
+    binary_mask = torch.where(
+        mask != 0,
+        torch.tensor(1).to(dtype),
+        torch.tensor(0).to(dtype),
+    )
+
+    # Create a block-diagonal mask.
+    # we multiply by the binary mask so that 0's in the original mask are correctly excluded
+    zero_one_mask = torch.eq(mask, mask.transpose(-1, -2)).int() * binary_mask
+
+    # Now let's create a lower triangular mask of ones that will zero out the upper triangular part
+    lower_triangular_ones = torch.tril(torch.ones((tgt_len, src_len), dtype=dtype)).to(
+        mask.device
+    )
+
+    # Use the lower triangular mask to zero out the upper triangular part of the zero_one_mask
+    masked_zero_one_mask = zero_one_mask * lower_triangular_ones
+    inverted_mask = 1.0 - masked_zero_one_mask
+
+    return inverted_mask.masked_fill(
+        inverted_mask.to(torch.bool), torch.finfo(dtype).min
+    )
+
+
+def hijack_expand_mask():
+    import transformers
+
+    transformers.models.llama.modeling_llama._expand_mask = (  # pylint: disable=protected-access
+        _expand_mask
+    )

src/axolotl/monkeypatch/utils.py

@@ -0,0 +1,103 @@
+"""
+Shared utils for the monkeypatches
+"""
+import torch
+
+
+def get_cu_seqlens(attn_mask):
+    """generate a cumulative sequence length mask for flash attention using attn mask"""
+    if len(attn_mask.shape) == 1:
+        attn_mask = attn_mask.unsqueeze(0)
+
+    device = attn_mask.device
+    results = []
+    max_seq_lens = []
+
+    for row in attn_mask:
+        # Exclude zeros to avoid adding their positions to the mask
+        t_non_zeros = row[row != 0]
+        # Find where the sequence number changes (including the first position)
+        seq_change = torch.cat(
+            [
+                torch.tensor([1], dtype=torch.int32, device=device),
+                t_non_zeros[1:] != t_non_zeros[:-1],
+            ]
+        )
+        # Get the indices where the sequence changes
+        change_indices = torch.cat(
+            [
+                (seq_change == 1).nonzero(as_tuple=True)[0],
+                torch.tensor([len(t_non_zeros)], dtype=torch.int32, device=device),
+            ]
+        )
+        # Calculate the sequence lengths
+        seq_lengths = change_indices[1:] - change_indices[:-1]
+        # Calculate the length of the final sequence or padding
+        final_seq_length = len(row) - change_indices[-1]
+        # Append the length of the final sequence or padding to seq_lengths
+        if final_seq_length.item():
+            seq_lengths = torch.cat(
+                [
+                    seq_lengths,
+                    torch.tensor(
+                        [final_seq_length.item()], dtype=torch.int32, device=device
+                    ),
+                ]
+            )
+        # Calculate the cumulative sequence lengths
+        cu_seqlens = torch.cat(
+            [torch.tensor([0], dtype=torch.int32, device=device), seq_lengths.cumsum(0)]
+        )
+        max_seq_len = (cu_seqlens[1:] - cu_seqlens[:-1]).max()
+        results.append(cu_seqlens)
+        max_seq_lens.append(max_seq_len)
+
+    return torch.stack(results).to(dtype=torch.int32), torch.stack(max_seq_lens)
+
+
+def get_cu_seqlens_from_pos_ids(position_ids):
+    """generate a cumulative sequence length mask for flash attention using pos ids"""
+    if len(position_ids.shape) == 1:
+        position_ids = position_ids.unsqueeze(0)
+
+    device = position_ids.device
+    results = []
+    max_seq_lens = []
+
+    for row in position_ids:
+        # Count the number of consecutive zeros from the right side
+        padding_length = (row == 0).int().flip(dims=[0]).cumprod(dim=0).sum().item()
+
+        # Adjust the row to exclude padding
+        adjusted_row = row[:-padding_length] if padding_length else row.clone()
+
+        # Find where the position resets to 0 (indicating a new sequence)
+        seq_starts = torch.cat(
+            [
+                torch.tensor([True], dtype=torch.bool, device=device),
+                adjusted_row[1:] == 0,
+            ]
+        )
+        # Get the indices where the sequence starts
+        start_indices = torch.cat(
+            [
+                (seq_starts).nonzero(as_tuple=True)[0],
+                torch.tensor([len(adjusted_row)], dtype=torch.int32, device=device),
+            ]
+        )
+        # Calculate the sequence lengths
+        seq_lengths = start_indices[1:] - start_indices[:-1]
+        # Calculate the cumulative sequence lengths
+        cu_seqlens = torch.cat(
+            [torch.tensor([0], dtype=torch.int32, device=device), seq_lengths.cumsum(0)]
+        )
+        # Append the padding length to the cumulative sequence lengths
+        if padding_length:
+            cu_seqlens = torch.cat(
+                [cu_seqlens, torch.tensor([len(row)], dtype=torch.int32, device=device)]
+            )
+        max_seq_len = (cu_seqlens[1:] - cu_seqlens[:-1]).max()
+        results.append(cu_seqlens)
+        max_seq_lens.append(max_seq_len)
+
+    return torch.stack(results).to(dtype=torch.int32), torch.stack(max_seq_lens)

src/axolotl/prompt_strategies/alpaca_w_system.py

@@ -66,7 +66,11 @@ class SystemDataPrompter(AlpacaPrompter):
     ) -> Generator[str, None, None]:
         # returns the full prompt from instruction and optional input
         # if a label (=response, =output) is provided, it's also appended.
-        formatted_sys_prompt = f"### System:\n{system}\n\n" if system else ""
+        formatted_sys_prompt = (
+            self.system_format.format(system=system)
+            if system and self.system_format
+            else ""
+        )
         if input:
             res = formatted_sys_prompt + self.turn_format.format(
                 instruction=instruction, input=input
@@ -86,12 +90,20 @@ class OpenOrcaSystemDataPrompter(SystemDataPrompter):
     """
 
     def match_prompt_style(self):
+        # pylint: disable=duplicate-code
         if self.prompt_style == PromptStyle.INSTRUCT.value:
             self.turn_format = "### User:\n{instruction}\n\n### Additional Context:\n{input}\n\n### Assistant:\n"
             self.turn_no_input_format = "### User:\n{instruction}\n\n### Assistant:\n"
         if self.prompt_style == PromptStyle.CHAT.value:
             self.turn_format = "USER: {instruction}\n{input}\nASSISTANT:"
             self.turn_no_input_format = "USER: {instruction}\nASSISTANT:"
+            self.system_format = "SYSTEM: {system}\n"
+        if self.prompt_style == PromptStyle.CHATML.value:
+            self.turn_format = "<|im_start|>user\n{instruction}\n{input}<|im_end|>\n<|im_start|>assistant\n"
+            self.turn_no_input_format = (
+                "<|im_start|>user\n{instruction}<|im_end|>\n<|im_start|>assistant\n"
+            )
+            self.system_format = "<|im_start|>system\n{system}<|im_end|>\n"
 
 
 class OpenOrcaPromptTokenizingStrategy(InstructionWSystemPromptTokenizingStrategy):
@@ -137,3 +149,12 @@ def load_open_orca(tokenizer, cfg):
         cfg.train_on_inputs,
         cfg.sequence_len,
     )
+
+
+def load_open_orca_chatml(tokenizer, cfg):
+    return OpenOrcaPromptTokenizingStrategy(
+        OpenOrcaSystemDataPrompter(PromptStyle.CHATML.value),
+        tokenizer,
+        cfg.train_on_inputs,
+        cfg.sequence_len,
+    )

src/axolotl/prompt_strategies/llama2_chat.py

@@ -0,0 +1,205 @@
+"""
+Prompt Strategy for finetuning Llama2 chat models
+see also https://github.com/facebookresearch/llama/blob/6c7fe276574e78057f917549435a2554000a876d/llama/generation.py#L213 for ma reference implementation.
+
+This implementation is based on the Vicuna PR and the fastchat repo, see also:
+https://github.com/lm-sys/FastChat/blob/cdd7730686cb1bf9ae2b768ee171bdf7d1ff04f3/fastchat/conversation.py#L847
+
+Use dataset type: "llama2_chat" in conig.yml to use this prompt style.
+
+E.g. in the config.yml:
+```
+datasets:
+  - path: llama_finetune_train.jsonl
+    type: llama2_chat
+```
+
+The dataset itself should look like this:
+```
+{'conversations':[{"from": "human", "value": "Who are you?"}, {"from": "gpt", "value": "I am Vicuna"},...]}
+```
+in a jsonl file. The first message should be from the human, the second from gpt.
+For a custom system message, the first "from" can be "system" (followed by alternating "human" and "gpt" turns).
+
+Important: Don't use "special_tokens:" in your config.yml if you are not sure what you are doing!
+"""
+
+import logging
+from dataclasses import dataclass, field
+from typing import Generator, List, Sequence
+
+from axolotl.prompt_tokenizers import PromptTokenizingStrategy
+from axolotl.prompters import IGNORE_TOKEN_ID, SHAREGPT_ASSERTION_FAILED_ROLE
+
+
+@dataclass
+class Llama2ChatConversation:
+    """A class that manages prompt templates and keeps all conversation history.
+    copied from https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py"""
+
+    name: str = "llama2"
+    # The system prompt
+    system: str = (
+        "[INST] <<SYS>>\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. "
+        "Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. "
+        "Please ensure that your responses are socially unbiased and positive in nature.\n\n"
+        "If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. "
+        "If you don't know the answer to a question, please don't share false information.\n<</SYS>>\n\n"
+    )
+    roles: Sequence[str] = ("[INST]", "[/INST]")
+    messages: List[List[str]] = field(default_factory=list)
+    offset: int = 0
+    sep = " "
+    sep2 = " </s><s>"
+    stop_token_ids = [2]
+
+    def get_prompt(self) -> str:
+        """Get the prompt for generation."""
+        seps = [self.sep, self.sep2]
+        ret = ""
+        for i, (role, message) in enumerate(self.messages):
+            if (i == len(self.messages) - 1) and (role == self.roles[0]):
+                # last message is from user (due to length),
+                #  return prompt without it for training
+                return ret
+            if i == 0:
+                ret += self.system + message.strip()
+            else:
+                ret += role + " " + message.strip() + seps[i % 2]
+        return ret
+
+    def append_message(self, role: str, message: str):
+        """Append a new message."""
+        self.messages.append([role, message])
+
+
+class LLama2ChatTokenizingStrategy(PromptTokenizingStrategy):
+    """
+    Tokenizing strategy for ShareGPT prompts.
+    adapted from https://github.com/lm-sys/FastChat/blob/main/fastchat/train/train.py
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.sequence_len = 4096
+        self.tokenizer.add_special_tokens({"pad_token": "<pad>"})
+        # https://huggingface.co/meta-llama/Llama-2-7b-chat-hf/blob/main/added_tokens.json
+
+    def tokenize_prompt(self, prompt):
+        conv = next(self.prompter.build_prompt(prompt))
+        conversation_str = conv.get_prompt()
+
+        # Tokenize conversations
+        input_ids = self.tokenizer(
+            conversation_str,
+            return_tensors="pt",
+            padding="max_length",
+            max_length=self.sequence_len,
+            truncation=True,
+        ).input_ids[0]
+        target = input_ids.clone()
+
+        # Mask targets. Only compute loss on the assistant outputs.
+        sep = conv.roles[1]
+
+        total_len = int(target.ne(self.tokenizer.pad_token_id).sum())
+
+        turns = conversation_str.split(conv.sep2)
+        cur_len = 1
+        target[:cur_len] = IGNORE_TOKEN_ID
+        for turn in turns:
+            if turn == "":
+                break
+            turn_len = len(self.tokenizer(turn).input_ids)
+
+            parts = turn.split(sep)
+            if len(parts) != 2:
+                break
+            parts[0] += sep
+            # "-1" is hardcoded for the LLaMA tokenizer to make the offset correct.
+            instruction_len = len(self.tokenizer(parts[0]).input_ids) - 1
+
+            # Ignore the user instructions
+            target[cur_len - 1 : cur_len + instruction_len] = IGNORE_TOKEN_ID
+            cur_len += turn_len + 2  # due to length of role token
+
+        target[cur_len:] = IGNORE_TOKEN_ID
+
+        if cur_len < self.sequence_len:
+            if cur_len != total_len:
+                target[:] = IGNORE_TOKEN_ID
+                logging.warning(
+                    f"WARNING: tokenization mismatch: {cur_len} vs. {total_len}."
+                    f" (ignored)"
+                )
+
+        attention_mask = input_ids.ne(self.tokenizer.pad_token_id).tolist()
+        input_ids = input_ids.tolist()
+        target = target.tolist()
+        # this is a fix for the tokenizer which tokenizes [ differently with eos tokens and
+        # follows the original llama implementation
+        for i in range(2, total_len - 2):
+            if input_ids[i] == 29961:
+                input_ids[i] = 518
+            if target[i] == 29961:
+                target[i] = 518
+        return {
+            "input_ids": input_ids,
+            "labels": target,
+            "attention_mask": attention_mask,
+        }
+
+
+class Llama2ChatPrompter:  # pylint: disable=too-few-public-methods
+    """
+    A prompter that generates prompts for Llama2 models.
+    """
+
+    system_prompt = (
+        "[INST] <<SYS>>\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. "
+        "Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. "
+        "Please ensure that your responses are socially unbiased and positive in nature.\n\n"
+        "If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. "
+        "If you don't know the answer to a question, please don't share false information.\n<</SYS>>\n\n"
+    )
+
+    def build_prompt(self, source) -> Generator[Llama2ChatConversation, None, None]:
+        # see https://github.com/lm-sys/FastChat/blob/da0641e567cf93756b0978ab5a6b092e96f06240/fastchat/train/train.py#L78
+        source = source["conversations"]  # fix data structure for datasets
+
+        # if system prompt provided, use it
+        if source[0]["from"] == "system":
+            system = f"[INST] <<SYS>>\n{source[0]['value']}\n<</SYS>>\n\n"
+            source = source[1:]
+        else:
+            system = self.system_prompt
+
+        conv = Llama2ChatConversation(system=system)
+
+        if len(source) < 2:
+            # If there isn't a back and forth conversation, ignore it
+            # also happens on the data splitting leaving empty conversations
+            raise IndexError
+
+        roles = {"human": conv.roles[0], "gpt": conv.roles[1]}
+
+        if roles[source[0]["from"]] != conv.roles[0]:
+            # Skip the first one if it is not from human
+            source = source[1:]
+
+        conv.messages = []  # pylint: disable=R0801
+        for j, sentence in enumerate(source):
+            role = roles[sentence["from"]]
+            assert role == conv.roles[j % 2], SHAREGPT_ASSERTION_FAILED_ROLE
+            if sentence["value"]:
+                conv.append_message(role, sentence["value"])
+        yield conv
+
+
+def load(tokenizer, cfg) -> LLama2ChatTokenizingStrategy:
+    return LLama2ChatTokenizingStrategy(
+        Llama2ChatPrompter(),
+        tokenizer,
+        cfg.train_on_inputs,
+        cfg.sequence_len,
+    )

src/axolotl/prompt_strategies/orcamini.py

@@ -0,0 +1,46 @@
+"""
+Prompt Strategy for finetuning Orca Mini (v2) models
+see also https://huggingface.co/psmathur/orca_mini_v2_7b for more information
+
+Use dataset type: orcamini in conig.yml to use this prompt style.
+
+Compared to the alpaca_w_system.open_orca dataset type,
+this one specifies the system prompt with "### System:".
+
+Not suited/tested for multiple-turn conversations without further adjustments.
+"""
+from typing import Generator, Union
+
+from axolotl.prompt_strategies.alpaca_w_system import OpenOrcaPromptTokenizingStrategy
+from axolotl.prompters import AlpacaPrompter
+
+
+class OrcaMiniPrompter(AlpacaPrompter):
+    """Adjusted Prompter for Orca Mini (v2) datasets"""
+
+    def match_prompt_style(self):
+        self.turn_no_input_format = (
+            "### System:\n{system}\n\n### User:\n{instruction}\n\n### Response:\n"
+        )
+
+    def build_prompt_w_system(
+        self,
+        system: str,
+        instruction: str,
+        output: Union[None, str] = None,
+    ) -> Generator[str, None, None]:
+        # returns the full prompt from instruction and optional input
+        # if a label (=response, =output) is provided, it's also appended.
+        res = self.turn_no_input_format.format(system=system, instruction=instruction)
+        if output:
+            res = f"{res}{output}"
+        yield res
+
+
+def load(tokenizer, cfg):
+    return OpenOrcaPromptTokenizingStrategy(
+        OrcaMiniPrompter(),
+        tokenizer,
+        cfg.train_on_inputs,
+        cfg.sequence_len,
+    )

src/axolotl/prompters.py

@@ -16,6 +16,7 @@ class PromptStyle(Enum):
 
     INSTRUCT = "instruct"
     CHAT = "chat"
+    CHATML = "chatml"
 
 
 class AlpacaPrompter:
@@ -25,6 +26,7 @@ class AlpacaPrompter:
 
     system_prompt = "Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n"
     system_no_input_prompt = "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
+    system_format: str
     turn_format: str
     turn_no_input_format: str
     prompt_style: Optional[PromptStyle] = None
@@ -34,14 +36,23 @@ class AlpacaPrompter:
         self.match_prompt_style()
 
     def match_prompt_style(self):
+        # pylint: disable=duplicate-code
         if self.prompt_style == PromptStyle.INSTRUCT.value:
             self.turn_format = "### Instruction:\n{instruction}\n\n### Input:\n{input}\n\n### Response:\n"
             self.turn_no_input_format = (
                 "### Instruction:\n{instruction}\n\n### Response:\n"
             )
+            self.system_format = "### System:\n{system}\n\n"
         if self.prompt_style == PromptStyle.CHAT.value:
             self.turn_format = "USER: {instruction}\n{input}\nASSISTANT:"
             self.turn_no_input_format = "USER: {instruction}\nASSISTANT:"
+            self.system_format = "SYSTEM: {system}\n"
+        if self.prompt_style == PromptStyle.CHATML.value:
+            self.turn_format = "<|im_start|>user\n{instruction}\n{input}<|im_end|>\n<|im_start|>assistant\n"
+            self.turn_no_input_format = (
+                "<|im_start|>user\n{instruction}<|im_end|>\n<|im_start|>assistant\n"
+            )
+            self.system_format = "<|im_start|>system\n{system}<|im_end|>\n"
 
     def build_prompt(
         self,
@@ -260,6 +271,11 @@ class Conversation:
         self.messages.append([role, message])
 
 
+SHAREGPT_ASSERTION_FAILED_ROLE = (
+    "Role did not alternate between turns (gpt and human). Please check your data."
+)
+
+
 class ShareGPTPrompter:  # pylint: disable=too-few-public-methods
     """
     A prompter that generates prompts for the ShareGPT
@@ -316,7 +332,7 @@ class ShareGPTPrompter:  # pylint: disable=too-few-public-methods
         conv.messages = []
         for j, sentence in enumerate(source):
             role = roles[sentence["from"]]
-            assert role == conv.roles[j % 2]
+            assert role == conv.roles[j % 2], SHAREGPT_ASSERTION_FAILED_ROLE
             conv.append_message(role, sentence["value"])
 
         for part in conv.get_prompt():

src/axolotl/utils/bench.py

@@ -0,0 +1,40 @@
+"""Benchmarking and measurement utilities"""
+
+import pynvml
+import torch
+
+
+def gpu_memory_usage(device=0):
+    return torch.cuda.memory_allocated(device) / 1024.0**3
+
+
+def gpu_memory_usage_all(device=0):
+    usage = torch.cuda.memory_allocated(device) / 1024.0**3
+    reserved = torch.cuda.memory_reserved(device) / 1024.0**3
+    smi = gpu_memory_usage_smi(device)
+    return usage, reserved - usage, max(0, smi - reserved)
+
+
+def gpu_memory_usage_smi(device=0):
+    if isinstance(device, torch.device):
+        device = device.index
+    if isinstance(device, str) and device.startswith("cuda:"):
+        device = int(device[5:])
+
+    pynvml.nvmlInit()
+    handle = pynvml.nvmlDeviceGetHandleByIndex(device)
+    info = pynvml.nvmlDeviceGetMemoryInfo(handle)
+    return info.used / 1024.0**3
+
+
+def log_gpu_memory_usage(log, msg, device):
+    usage, cache, misc = gpu_memory_usage_all(device)
+    extras = []
+    if cache > 0:
+        extras.append(f"+{cache:.03f}GB cache")
+    if misc > 0:
+        extras.append(f"+{misc:.03f}GB misc")
+    log.info(
+        f"GPU memory usage {msg}: {usage:.03f}GB ({', '.join(extras)})", stacklevel=2
+    )
+    return usage, cache, misc

src/axolotl/utils/callbacks.py

@@ -1,5 +1,6 @@
 """Callbacks for Trainer class"""
 
+import logging
 import os
 
 from optimum.bettertransformer import BetterTransformer
@@ -11,6 +12,10 @@ from transformers import (
 )
 from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR, IntervalStrategy
 
+from axolotl.utils.bench import log_gpu_memory_usage
+
+LOG = logging.getLogger("axolotl.callbacks")
+
 
 class SavePeftModelCallback(TrainerCallback):  # pylint: disable=too-few-public-methods
     """Callback to save the PEFT adapter"""
@@ -67,3 +72,25 @@ class SaveBetterTransformerModelCallback(
             # the trainer will raise an exception since it can't save a BetterTransformer wrapped model
             control.should_save = False
         return control
+
+
+class GPUStatsCallback(
+    TrainerCallback
+):  # pylint: disable=too-few-public-methods disable=unused-argument
+    """Callback to track GPU utilization"""
+
+    def __init__(self, cfg):
+        self.cfg = cfg
+        self.logged = False
+
+    def on_step_end(
+        self,
+        args: TrainingArguments,
+        state: TrainerState,
+        control: TrainerControl,
+        **kwargs,
+    ):
+        if not self.logged and state.global_step > 1:
+            log_gpu_memory_usage(LOG, "while training", self.cfg.device)
+            self.logged = True
+        return control

src/axolotl/utils/collators.py

@@ -0,0 +1,121 @@
+"""
+DataCollator for axolotl to pad labels and position_ids for packed sequences
+"""
+from dataclasses import dataclass
+from typing import Any, Optional, Union
+
+import numpy as np
+from transformers import PreTrainedTokenizerBase
+from transformers.utils import PaddingStrategy
+
+
+@dataclass
+class DataCollatorForSeq2Seq:
+    """
+    Data collator that will dynamically pad the inputs received, as well as the labels and position_ids
+
+    Args:
+        tokenizer ([`PreTrainedTokenizer`] or [`PreTrainedTokenizerFast`]):
+            The tokenizer used for encoding the data.
+        model ([`PreTrainedModel`]):
+            The model that is being trained. If set and has the *prepare_decoder_input_ids_from_labels*, use it to
+            prepare the *decoder_input_ids*
+
+            This is useful when using *label_smoothing* to avoid calculating loss twice.
+        padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `True`):
+            Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
+            among:
+
+            - `True` or `'longest'` (default): Pad to the longest sequence in the batch (or no padding if only a single
+              sequence is provided).
+            - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
+              acceptable input length for the model if that argument is not provided.
+            - `False` or `'do_not_pad'`: No padding (i.e., can output a batch with sequences of different lengths).
+        max_length (`int`, *optional*):
+            Maximum length of the returned list and optionally padding length (see above).
+        pad_to_multiple_of (`int`, *optional*):
+            If set will pad the sequence to a multiple of the provided value.
+
+            This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
+            7.5 (Volta).
+        label_pad_token_id (`int`, *optional*, defaults to -100):
+            The id to use when padding the labels (-100 will be automatically ignored by PyTorch loss functions).
+        return_tensors (`str`):
+            The type of Tensor to return. Allowable values are "np", "pt" and "tf".
+    """
+
+    tokenizer: PreTrainedTokenizerBase
+    model: Optional[Any] = None
+    padding: Union[bool, str, PaddingStrategy] = True
+    max_length: Optional[int] = None
+    pad_to_multiple_of: Optional[int] = None
+    label_pad_token_id: int = -100
+    position_pad_token_id: int = 0
+    return_tensors: str = "pt"
+
+    def __call__(self, features, return_tensors=None):
+        labels = None
+        if return_tensors is None:
+            return_tensors = self.return_tensors
+
+        for feature_name, pad_token_id in [
+            ("labels", self.label_pad_token_id),
+            ("position_ids", self.position_pad_token_id),
+        ]:
+            feat = (
+                [feature[feature_name] for feature in features]
+                if feature_name in features[0].keys()
+                else None
+            )
+            labels = feat if feat and feature_name == "labels" else labels
+            # We have to pad the labels before calling `tokenizer.pad` as this method won't pad them and needs them of the
+            # same length to return tensors.
+            if feat is not None:
+                max_feature_length = max(len(l) for l in feat)  # noqa: E741
+                if self.pad_to_multiple_of is not None:
+                    max_feature_length = (
+                        (max_feature_length + self.pad_to_multiple_of - 1)
+                        // self.pad_to_multiple_of
+                        * self.pad_to_multiple_of
+                    )
+
+                padding_side = self.tokenizer.padding_side
+                for feature in features:
+                    remainder = [pad_token_id] * (
+                        max_feature_length - len(feature[feature_name])
+                    )
+                    if isinstance(feature[feature_name], list):
+                        feature[feature_name] = (
+                            feature[feature_name] + remainder
+                            if padding_side == "right"
+                            else remainder + feature[feature_name]
+                        )
+                    elif padding_side == "right":
+                        feature[feature_name] = np.concatenate(
+                            [feature[feature_name], remainder]
+                        ).astype(np.int64)
+                    else:
+                        feature[feature_name] = np.concatenate(
+                            [remainder, feature[feature_name]]
+                        ).astype(np.int64)
+
+        features = self.tokenizer.pad(
+            features,
+            padding=self.padding,
+            max_length=self.max_length,
+            pad_to_multiple_of=self.pad_to_multiple_of,
+            return_tensors=return_tensors,
+        )
+
+        # prepare decoder_input_ids
+        if (
+            labels is not None
+            and self.model is not None
+            and hasattr(self.model, "prepare_decoder_input_ids_from_labels")
+        ):
+            decoder_input_ids = self.model.prepare_decoder_input_ids_from_labels(
+                labels=features["labels"]
+            )
+            features["decoder_input_ids"] = decoder_input_ids
+
+        return features

src/axolotl/utils/config.py

@@ -1,13 +1,84 @@
-"""Module for validating config files"""
+"""Module for working with config dicts"""
 
 import logging
+import os
 
 import torch
 
+from axolotl.utils.bench import log_gpu_memory_usage
+
 LOG = logging.getLogger("axolotl")
 
 
+def choose_device(cfg):
+    def get_device():
+        try:
+            if torch.cuda.is_available():
+                return f"cuda:{cfg.local_rank}"
+
+            if torch.backends.mps.is_available():
+                return "mps"
+
+            raise SystemError("No CUDA/mps device found")
+        except Exception:  # pylint: disable=broad-exception-caught
+            return "cpu"
+
+    cfg.device = get_device()
+    if cfg.device_map != "auto":
+        if cfg.device.startswith("cuda"):
+            cfg.device_map = {"": cfg.local_rank}
+        else:
+            cfg.device_map = {"": cfg.device}
+
+    # in `accelerate launch`, we need to not pass through any device map and let
+    # accelerate figure out which parts of the model to put on which gpu
+    accelerate_vars = [var for var in os.environ if var.startswith("ACCELERATE_USE_")]
+    if accelerate_vars:
+        cfg.device_map = None
+
+
+def normalize_config(cfg):
+    # setup some derived config / hyperparams
+    cfg.gradient_accumulation_steps = cfg.gradient_accumulation_steps or (
+        cfg.batch_size // cfg.micro_batch_size
+    )
+    cfg.batch_size = (
+        cfg.batch_size or cfg.micro_batch_size * cfg.gradient_accumulation_steps
+    )
+    cfg.world_size = int(os.environ.get("WORLD_SIZE", 1))
+    cfg.local_rank = int(os.environ.get("LOCAL_RANK", 0))
+    choose_device(cfg)
+    cfg.ddp = cfg.ddp if cfg.ddp is not None else cfg.world_size != 1
+    if cfg.ddp:
+        cfg.device_map = {"": int(os.environ.get("LOCAL_RANK", 0))}
+        cfg.batch_size = cfg.batch_size * cfg.world_size
+
+    if cfg.device == "mps":
+        cfg.load_in_8bit = False
+        cfg.tf32 = False
+        if cfg.bf16:
+            cfg.fp16 = True
+        cfg.bf16 = False
+    else:
+        torch.backends.cuda.matmul.allow_tf32 = cfg.tf32 or False
+
+    log_gpu_memory_usage(LOG, "baseline", cfg.device)
+
+
 def validate_config(cfg):
+    if cfg.max_packed_sequence_len and cfg.sample_packing:
+        raise ValueError(
+            "please set only one of max_packed_sequence_len (deprecated soon) or sample_packing"
+        )
+    if cfg.max_packed_sequence_len:
+        LOG.warning(
+            str(
+                PendingDeprecationWarning(
+                    "max_packed_sequence_len will be deprecated in favor of sample_packing"
+                )
+            )
+        )
+
     if cfg.gradient_accumulation_steps and cfg.batch_size:
         raise ValueError(
             "please set only one of gradient_accumulation_steps or batch_size"
@@ -97,6 +168,24 @@ def validate_config(cfg):
             "push_to_hub_model_id is deprecated. Please use hub_model_id instead."
         )
 
+    if cfg.gptq and cfg.model_revision:
+        raise ValueError(
+            "model_revision is not supported for GPTQ models. "
+            + "Please download the model from HuggingFace Hub manually for correct branch, "
+            + "point to its path, and remove model_revision from the config."
+        )
+
+    if cfg.sample_packing and cfg.sdp_attention:
+        # incompatible due to bug w/ accelerate causing 0.0 loss when using llama2
+        raise ValueError(
+            "sample_packing not compatible with sdp_attention. Use flash_attention"
+        )
+
+    if cfg.sample_packing and cfg.xformers_attention:
+        raise ValueError(
+            "sample_packing not compatible with xformers_attention. Use flash_attention"
+        )
+
     # TODO
     # MPT 7b
     # https://github.com/facebookresearch/bitsandbytes/issues/25

src/axolotl/utils/data.py

@@ -1,13 +1,19 @@
 """Module containing data utilities"""
 import functools
-import itertools
+import hashlib
 import logging
 from hashlib import md5
 from pathlib import Path
-from typing import List, Tuple, Union
+from typing import Tuple, Union
 
 import torch
-from datasets import Dataset, DatasetDict, load_dataset, load_from_disk
+from datasets import (
+    Dataset,
+    DatasetDict,
+    concatenate_datasets,
+    load_dataset,
+    load_from_disk,
+)
 from huggingface_hub import hf_hub_download
 from transformers import PreTrainedTokenizerBase
 
@@ -35,6 +41,7 @@ from axolotl.prompters import (
     ShareGPTPrompter,
     SummarizeTLDRPrompter,
 )
+from axolotl.utils.distributed import barrier, is_main_process
 
 LOG = logging.getLogger("axolotl")
 
@@ -109,6 +116,7 @@ def load_tokenized_prepared_datasets(
             local_path = Path(d.path)
             if local_path.exists():
                 if local_path.is_dir():
+                    # TODO dirs with arrow or parquet files could be loaded with `load_from_disk`
                     ds = load_dataset(
                         d.path,
                         name=d.name,
@@ -262,20 +270,12 @@ def load_tokenized_prepared_datasets(
                 raise ValueError(
                     f"unhandled prompt tokenization strategy: {d.type} {suffix}"
                 )
-        LOG.info("tokenizing, merging, and shuffling master dataset")
+        LOG.info("merging datasets")
+        dataset = concatenate_datasets(datasets)
 
-        samples: List[int] = []
-        chunk_size = 1000
-        for d in datasets:
-            d_iter = iter(d)
-            while True:
-                chunk = list(itertools.islice(d_iter, chunk_size))
-                if not chunk:
-                    break
-                samples.extend(chunk)
-
-        LOG.info("shuffle")
-        dataset = Dataset.from_list(samples).shuffle(seed=seed)
+        if len(datasets) > 1:
+            LOG.info("shuffle merged datasets")
+            dataset = dataset.shuffle(seed=seed)
         if cfg.local_rank == 0:
             LOG.info(f"Saving merged prepared dataset to disk... {prepared_ds_path}")
             dataset.save_to_disk(prepared_ds_path)
@@ -374,11 +374,12 @@ def load_prepare_datasets(
             dataset = Dataset.from_list(list(constant_len_dataset))
 
             # filter out bad data
+            # TODO convert to dataset.filter(...)
             dataset = Dataset.from_list(
                 [
                     d
                     for d in dataset
-                    if len(d["input_ids"]) < cfg.sequence_len
+                    if len(d["input_ids"]) <= cfg.sequence_len
                     and len(d["input_ids"]) > 0
                     and len(d["input_ids"]) == len(d["attention_mask"])
                     and len(d["input_ids"]) == len(d["labels"])
@@ -413,7 +414,51 @@ def load_prepare_datasets(
         )
 
     if cfg.val_set_size:
-        dataset = dataset.train_test_split(test_size=cfg.val_set_size, shuffle=False)
+        # ensure we end up with the same fingerprint by doing rank0 first and being able to cache
+        to_hash_train = (
+            dataset._fingerprint  # pylint: disable=protected-access
+            + "|"
+            + str(cfg.val_set_size)
+            + "|"
+            + "train"
+            + "|"
+            + str(cfg.seed or 42)
+        )
+        to_hash_test = (
+            dataset._fingerprint  # pylint: disable=protected-access
+            + "|"
+            + str(cfg.val_set_size)
+            + "|"
+            + "test"
+            + "|"
+            + str(cfg.seed or 42)
+        )
+        train_fingerprint = hashlib.md5(
+            to_hash_train.encode(), usedforsecurity=False
+        ).hexdigest()
+        test_fingerprint = hashlib.md5(
+            to_hash_test.encode(), usedforsecurity=False
+        ).hexdigest()
+
+        if is_main_process():
+            dataset = dataset.train_test_split(
+                test_size=cfg.val_set_size,
+                shuffle=False,
+                seed=cfg.seed or 42,
+                train_new_fingerprint=train_fingerprint,
+                test_new_fingerprint=test_fingerprint,
+            )
+        barrier()
+        if not is_main_process():
+            dataset = dataset.train_test_split(
+                test_size=cfg.val_set_size,
+                shuffle=False,
+                seed=cfg.seed or 42,
+                train_new_fingerprint=train_fingerprint,
+                test_new_fingerprint=test_fingerprint,
+            )
+        barrier()
+
         train_dataset = dataset["train"]
         eval_dataset = dataset["test"]
     else:

src/axolotl/utils/dataloader.py

@@ -0,0 +1,288 @@
+# pylint: skip-file
+import hashlib
+import itertools
+import logging
+import math
+from typing import Any, Callable, List, Union
+
+import numba
+import numpy as np
+from torch.utils.data import DistributedSampler, Sampler
+
+LOG = logging.getLogger("axolotl.utils.dataloader")
+
+
+@numba.njit
+def ffd_check(a: np.ndarray, c: int, n: int):
+    # First-fit-decreasing bin packing
+    # Check if a[] could fit in n bins with capacity c
+    # https://en.wikipedia.org/wiki/First-fit-decreasing_bin_packing
+
+    a = np.sort(a)[::-1]
+    bins = np.full((n,), c, dtype=a.dtype)
+    for size in a:
+        not_found = True
+        for idx in range(n):
+            if bins[idx] >= size:
+                bins[idx] -= size
+                not_found = False
+                break
+
+        if not_found:
+            return False
+
+    return True
+
+
+@numba.njit
+def ffd_with_result(a: np.ndarray, c: int, start_index: int):
+    # First-fit-decreasing bin packing (with result return)
+
+    indices = np.argsort(a)[::-1]
+    a = a[indices]
+
+    bins: List[Any] = []
+    bins_result: List[Any] = []
+    for a_id, size in enumerate(a):
+        add_new = True
+        for idx in range(len(bins)):
+            if bins[idx] >= size:
+                bins[idx] -= size
+                bins_result[idx].append(indices[a_id] + start_index)
+                add_new = False
+                break
+
+        if add_new:
+            bins.append(c - size)
+            bins_result.append([indices[a_id] + start_index])
+
+    return bins_result, len(a)
+
+
+@numba.njit
+def allocate(
+    lengths: np.ndarray, lengths_cumsum: np.ndarray, rank: int, c: int, n: int
+):
+    """
+    :param lengths: array of lengths of each sample
+    :param lengths_cumsum: cumulative sum of consecutive lengths
+    :param rank: rank for this process
+    :param c: length of tokens per batch
+    :param n: number of ranks
+    :return:
+    """
+    # Dynamic batch allocator, similar to Multifit
+    # https://en.wikipedia.org/wiki/Multifit_algorithm
+    # ~99.5% efficiency on OpenChat training set (12 * 2048 ctx len)
+
+    s = 0
+    start_index = 0
+    result = []
+    result_totseqs = []
+
+    while True:
+        # binary search [left, right)
+        left = 1
+        right = 1 + np.searchsorted(lengths_cumsum[start_index:], s + c * n, "right")
+
+        while right - left > 1:
+            mid = (left + right) // 2
+            if ffd_check(lengths[start_index : start_index + mid], c, n):
+                left = mid
+            else:
+                right = mid
+
+        # use length left
+        batch, tot_seqs = ffd_with_result(
+            lengths[start_index : start_index + left], c, start_index
+        )
+        if len(batch) < n:
+            break
+
+        start_index += left
+        s = lengths_cumsum[start_index - 1]
+
+        # add local rank
+        result.append(batch[rank])
+        # add total seqs for all ranks
+        result_totseqs.append(tot_seqs)
+        # yield batch[rank], tot_seqs, s, len(result) * c * n
+    return result, result_totseqs, s, len(result) * c * n
+
+
+def chunk(iterable, n):
+    """
+    Chunk data into tuples of length n
+    """
+    # batched('ABCDEFG', 3) --> ABC DEF G
+    if n < 1:
+        raise ValueError("n must be at least one")
+    it = iter(iterable)
+    while batch := tuple(itertools.islice(it, n)):
+        yield batch
+
+
+def hash_indices(lst: List[int]) -> str:
+    # Convert the list of integers to a string representation
+    concatenated = ",".join(map(str, lst))
+
+    # Generate the hash
+    sha256 = hashlib.sha256()
+    sha256.update(concatenated.encode())
+
+    return sha256.hexdigest()
+
+
+class MultipackDistributedDataloader:
+    """Unpadded data loading using Multipack.
+    Adapted from https://github.com/imoneoi/openchat/blob/v3_fix_mle_loss/ochat/training_deepspeed/multipack_dataloader.py
+    Approximate (at most ~1.22x) the optimal solution of the identical-machines scheduling problem, which is NP-hard.
+    """
+
+    def __init__(
+        self,
+        dataset: Any,
+        collate_fn: Callable,
+        seq_max_length: int = 2048,
+        batch_size: int = 1,
+        sampler: Union[Sampler, DistributedSampler] = None,
+        packing_efficiency_estimate: float = 1.0,
+        sample_packing_seq_len_multiplier: int = 1,
+        device_count: int = 1,
+    ):
+        # Dataset
+        self.dataset = dataset
+        self.lengths = (
+            dataset.data.column("position_ids")
+            .to_pandas()
+            .apply(lambda x: x[-1] + 1)
+            .values
+        )
+        assert isinstance(self.lengths, np.ndarray)
+        assert batch_size % sample_packing_seq_len_multiplier == 0
+        assert batch_size >= sample_packing_seq_len_multiplier
+        self.sampler = sampler
+        self.batch_size = batch_size
+        self.sample_packing_seq_len_multiplier = sample_packing_seq_len_multiplier
+        self.seq_max_length = seq_max_length
+        self.batch_max_length = batch_size * seq_max_length
+        self.collate_fn = collate_fn
+
+        self.num_replicas = 1
+        self.rank = 0
+
+        # statistics
+        self.eff_total_used = 0
+        self.eff_total_slots = 0
+        self.packing_efficiency_estimate = packing_efficiency_estimate or 1.0
+        self.device_count = device_count
+
+    def generate_batches(self, set_stats=False):
+        LOG.info("generating packed batches")
+        if self.sampler:
+            indices = [idx for idx in self.sampler]
+        else:
+            indices = range(0, len(self.dataset))
+
+        LOG.info(hash_indices(indices))
+        lengths = self.lengths[indices]
+        lengths_cumsum = np.cumsum(lengths)
+
+        batches, totseqs, total_used, total_slots = allocate(
+            lengths=lengths,
+            lengths_cumsum=lengths_cumsum,
+            rank=self.rank,
+            # c=self.batch_max_length,
+            c=self.seq_max_length * self.sample_packing_seq_len_multiplier,
+            n=self.num_replicas,
+        )
+
+        batches = [[indices[b_idx] for b_idx in batch] for batch in batches]
+
+        # statistics
+        if set_stats:
+            self.eff_total_used += total_used
+            self.eff_total_slots += total_slots
+
+        return batches, totseqs
+
+    def __iter__(self):
+        if hasattr(self.sampler, "set_epoch"):
+            new_epoch = self.sampler.epoch + 1
+            self.sampler.set_epoch(new_epoch)
+            LOG.info(f"calling sampler.set_epoch({new_epoch})")
+        all_batches, _ = self.generate_batches(set_stats=True)
+        features = self.dataset.features.keys()
+        len_remaining = self._len_est()
+        for batches in chunk(
+            all_batches, self.batch_size // self.sample_packing_seq_len_multiplier
+        ):
+            chunked_data = []
+            attn_mask_cum_idx = 0
+            for batch in batches:
+                concatenated = {}
+                batched_data = [self.dataset[batch_idx] for batch_idx in batch]
+                for feature in features:
+                    if feature == "attention_mask":
+                        arrays = [
+                            (attn_mask_cum_idx + idx + 1) * np.array(item[feature])
+                            for idx, item in enumerate(batched_data)
+                            if feature in item
+                        ]
+                        attn_mask_cum_idx += len(batched_data)
+                        concatenated[feature] = np.concatenate(arrays)
+                    else:
+                        arrays = [
+                            np.array(item[feature])
+                            for item in batched_data
+                            if feature in item
+                        ]
+                        concatenated[feature] = np.concatenate(arrays)
+                chunked_data.append(concatenated)
+            yield self.collate_fn(chunked_data)
+            len_remaining -= 1
+            if not len_remaining:
+                return
+
+    def _len_est(self):
+        lengths_sum = np.sum(self.lengths)
+        lengths_sum_per_device = lengths_sum // self.device_count
+        LOG.info(
+            f"packing_efficiency_estimate: {self.packing_efficiency_estimate} "
+            f"total_num_tokens per device: {lengths_sum_per_device}"
+        )
+
+        # shave off 1% + 1 for dealing with variance in packing from random sampler to sampler
+        return (
+            math.floor(
+                0.99
+                * lengths_sum_per_device
+                / self.packing_efficiency_estimate
+                // self.seq_max_length
+                // self.batch_size
+            )
+            - 1
+        )
+
+    def __len__(self):
+        # this doesn't return the actual length b/c with distributed samplers, not all dataloaders get
+        # the same share of total tokens
+        # if not self.eff_total_used:
+        #     batches, _ = self.generate_batches(set_stats=True)
+        # LOG.info(
+        #     f"packing_efficiency_estimate: {self.packing_efficiency_estimate} "
+        #     f"actual packing efficiency: {self.efficiency()}"
+        # )
+        return max(1, self._len_est())
+
+    def len_w_stats(self):
+        if not self.eff_total_used:
+            batches, _ = self.generate_batches(set_stats=True)
+        LOG.info(
+            f"packing_efficiency_estimate: {self.packing_efficiency_estimate} "
+            f"actual packing efficiency: {self.efficiency()}"
+        )
+        return max(1, self._len_est())
+
+    def efficiency(self):
+        return self.eff_total_used / self.eff_total_slots

src/axolotl/utils/dict.py

@@ -10,3 +10,6 @@ class DictDefault(Dict):
 
     def __missing__(self, key):
         return None
+
+    def __or__(self, other):
+        return DictDefault(super().__or__(other))

src/axolotl/utils/distributed.py

@@ -0,0 +1,41 @@
+"""
+utility helpers for distributed checks
+"""
+import torch.distributed as dist
+from accelerate import Accelerator
+
+accelerate = None  # pylint: disable=invalid-name
+
+
+def load_accelerate():
+    global accelerate  # pylint: disable=global-statement
+    accelerate = Accelerator()
+
+
+def is_distributed():
+    """
+    Check if distributed training is initialized.
+    """
+    global accelerate  # pylint: disable=global-statement
+    if not accelerate:
+        accelerate = Accelerator()
+    return dist.is_available() and dist.is_initialized()
+
+
+def barrier():
+    """
+    Acts as a barrier to wait for all processes. This ensures that all processes
+    reach the barrier before proceeding further.
+    """
+    if is_distributed():
+        dist.barrier()
+
+
+def is_main_process():
+    """
+    Check if the current process is the main process.
+    If not in distributed mode, always return True.
+    """
+    if not is_distributed():
+        return True
+    return dist.get_rank() == 0

src/axolotl/utils/models.py

@@ -22,6 +22,7 @@ from transformers import (  # noqa: F401
 )
 
 from axolotl.prompt_tokenizers import LLAMA_DEFAULT_PAD_TOKEN
+from axolotl.utils.bench import log_gpu_memory_usage
 
 LOG = logging.getLogger("axolotl")
 
@@ -31,31 +32,27 @@ if TYPE_CHECKING:
     from axolotl.utils.dict import DictDefault  # noqa: F401
 
 
-def load_tokenizer(
-    tokenizer_config,
-    tokenizer_type,
-    cfg,
-):
+def load_tokenizer(cfg):
+    tokenizer_kwargs = {}
     use_fast = True  # this is the default
+
     if cfg.tokenizer_use_fast is not None:
         use_fast = cfg.tokenizer_use_fast
-    if tokenizer_type:
-        tokenizer = getattr(transformers, tokenizer_type).from_pretrained(
-            tokenizer_config,
-            trust_remote_code=cfg.trust_remote_code or False,
-            use_fast=use_fast,
-        )
-    else:
-        tokenizer = AutoTokenizer.from_pretrained(
-            tokenizer_config,
-            trust_remote_code=cfg.trust_remote_code or False,
-            use_fast=use_fast,
-        )
+    if cfg.tokenizer_legacy is not None:
+        # True is the default w/ https://github.com/huggingface/transformers/pull/25224
+        tokenizer_kwargs["legacy"] = cfg.tokenizer_legacy
 
-    LOG.debug(f"EOS: {tokenizer.eos_token_id} / {tokenizer.eos_token}")
-    LOG.debug(f"BOS: {tokenizer.bos_token_id} / {tokenizer.bos_token}")
-    LOG.debug(f"PAD: {tokenizer.pad_token_id} / {tokenizer.pad_token}")
-    LOG.debug(f"UNK: {tokenizer.unk_token_id} / {tokenizer.unk_token}")
+    tokenizer_cls = AutoTokenizer
+    if cfg.tokenizer_type:
+        tokenizer_cls = getattr(transformers, cfg.tokenizer_type)
+
+    tokenizer_config = cfg.tokenizer_config or cfg.base_model_config
+    tokenizer = tokenizer_cls.from_pretrained(
+        tokenizer_config,
+        trust_remote_code=cfg.trust_remote_code or False,
+        use_fast=use_fast,
+        **tokenizer_kwargs,
+    )
 
     if tokenizer.__class__.__name__ in [
         "LlamaTokenizer",
@@ -63,6 +60,11 @@ def load_tokenizer(
     ]:
         tokenizer.pad_token = LLAMA_DEFAULT_PAD_TOKEN
 
+    LOG.debug(f"EOS: {tokenizer.eos_token_id} / {tokenizer.eos_token}")
+    LOG.debug(f"BOS: {tokenizer.bos_token_id} / {tokenizer.bos_token}")
+    LOG.debug(f"PAD: {tokenizer.pad_token_id} / {tokenizer.pad_token}")
+    LOG.debug(f"UNK: {tokenizer.unk_token_id} / {tokenizer.unk_token}")
+
     if tokenizer.__class__.__name__ == "GPTNeoXTokenizerFast":
         tokenizer.add_special_tokens({"pad_token": "[PAD]"})
         os.environ["TOKENIZERS_PARALLELISM"] = "false"
@@ -77,17 +79,21 @@ def load_tokenizer(
 
 
 def load_model(
-    base_model, base_model_config, model_type, tokenizer, cfg, adapter="lora"
-):
-    # type: (str, str, str, PreTrainedTokenizerBase, DictDefault, Optional[str]) -> Tuple[PreTrainedModel, Optional[PeftConfig]]
+    cfg, tokenizer
+):  # type: (DictDefault, PreTrainedTokenizerBase) -> Tuple[PreTrainedModel, Optional[PeftConfig]]
     """
-    Load a model from a base model and a model type.
+    Load a model for a given configuration and tokenizer.
     """
+    base_model = cfg.base_model
+    base_model_config = cfg.base_model_config
+    model_type = cfg.model_type
 
     # TODO refactor as a kwarg
     load_in_8bit = cfg.load_in_8bit
-    cfg.is_llama_derived_model = "llama" in base_model or (
-        cfg.model_type and "llama" in cfg.model_type.lower()
+    cfg.is_llama_derived_model = (
+        "llama" in base_model
+        or (cfg.model_type and "llama" in cfg.model_type.lower())
+        or cfg.is_llama_derived_model
     )
 
     if cfg.is_llama_derived_model and cfg.flash_attention:
@@ -106,9 +112,7 @@ def load_model(
         LOG.info("patching with xformers attention")
         hijack_llama_attention()
     elif cfg.is_llama_derived_model and cfg.sdp_attention:
-        from axolotl.monkeypatch.llama_attn_hijack_xformers import (
-            hijack_llama_sdp_attention,
-        )
+        from axolotl.monkeypatch.llama_attn_hijack_sdp import hijack_llama_sdp_attention
 
         LOG.info("patching with sdp attention")
         hijack_llama_sdp_attention()
@@ -132,6 +136,14 @@ def load_model(
         LOG.info("patching with xpos rope")
         replace_llama_rope_with_xpos_rope()
 
+    if cfg.is_llama_derived_model and (
+        cfg.max_packed_sequence_len or cfg.sample_packing
+    ):
+        from axolotl.monkeypatch.llama_expand_mask import hijack_expand_mask
+
+        LOG.info("patching _expand_mask")
+        hijack_expand_mask()
+
     if cfg.bf16 or cfg.bfloat16:
         torch_dtype = torch.bfloat16
     elif cfg.load_in_8bit or cfg.fp16 or cfg.float16:
@@ -215,14 +227,16 @@ def load_model(
         elif cfg.is_llama_derived_model and not cfg.trust_remote_code:
             from transformers import LlamaForCausalLM
 
-            config = LlamaConfig.from_pretrained(base_model_config)
+            config = LlamaConfig.from_pretrained(
+                base_model_config, rope_scaling=cfg.rope_scaling
+            )
             model = LlamaForCausalLM.from_pretrained(
                 base_model,
                 config=config,
+                device_map=cfg.device_map,
                 load_in_8bit=cfg.load_in_8bit and cfg.adapter is not None,
                 load_in_4bit=cfg.load_in_4bit and cfg.adapter is not None,
                 torch_dtype=torch_dtype,
-                device_map="auto" if cfg.world_size == 1 else cfg.device_map,
                 **model_kwargs,
             )
         # elif model_type == "GPTNeoXForCausalLM" and cfg.flash_attention:
@@ -254,10 +268,10 @@ def load_model(
         elif model_type and not cfg.trust_remote_code:
             model = getattr(transformers, model_type).from_pretrained(
                 base_model,
+                device_map=cfg.device_map,
                 load_in_8bit=cfg.load_in_8bit and cfg.adapter is not None,
                 load_in_4bit=cfg.load_in_4bit and cfg.adapter is not None,
                 torch_dtype=torch_dtype,
-                device_map=cfg.device_map,
                 trust_remote_code=cfg.trust_remote_code or False,
                 **model_kwargs,
             )
@@ -285,10 +299,10 @@ def load_model(
             model = AutoModelForCausalLM.from_pretrained(
                 base_model,
                 config=config,
+                device_map=cfg.device_map,
                 load_in_8bit=cfg.load_in_8bit and cfg.adapter is not None,
                 load_in_4bit=cfg.load_in_4bit and cfg.adapter is not None,
                 torch_dtype=torch_dtype,
-                device_map=cfg.device_map,
                 trust_remote_code=cfg.trust_remote_code or False,
                 **model_kwargs,
             )
@@ -299,10 +313,10 @@ def load_model(
         LOG.exception(err)
         model = AutoModelForCausalLM.from_pretrained(
             base_model,
+            device_map=cfg.device_map,
             load_in_8bit=cfg.load_in_8bit and cfg.adapter is not None,
             load_in_4bit=cfg.load_in_4bit and cfg.adapter is not None,
             torch_dtype=torch_dtype,
-            device_map=cfg.device_map,
             trust_remote_code=cfg.trust_remote_code or False,
             **model_kwargs,
         )
@@ -324,6 +338,9 @@ def load_model(
         )
         model.config.max_position_embeddings = cfg.sequence_len
 
+    if model.device.type == "cuda":
+        log_gpu_memory_usage(LOG, "after model load", model.device)
+
     if not cfg.gptq and (
         (cfg.adapter == "lora" and load_in_8bit)
         or (cfg.adapter == "qlora" and cfg.load_in_4bit)
@@ -343,7 +360,7 @@ def load_model(
                     if hasattr(module, "weight"):
                         module.to(torch_dtype)
 
-    model, lora_config = load_adapter(model, cfg, adapter)
+    model, lora_config = load_adapter(model, cfg, cfg.adapter)
 
     if cfg.ddp and not load_in_8bit:
         model.to(f"cuda:{cfg.local_rank}")
@@ -382,6 +399,9 @@ def load_model(
     if cfg.flash_optimum:
         model = BetterTransformer.transform(model)
 
+    if cfg.adapter is not None:
+        log_gpu_memory_usage(LOG, "after adapters", model.device)
+
     # TODO resume_from_checkpoint handling
     return model, lora_config
 
@@ -391,6 +411,8 @@ def load_adapter(model, cfg, adapter):
 
     if adapter is None:
         return model, None
+    if hasattr(model, "enable_input_require_grads"):
+        model.enable_input_require_grads()
     if adapter in ["lora", "qlora"]:
         return load_lora(model, cfg)
     if adapter == "llama-adapter":

src/axolotl/utils/trainer.py

@@ -1,26 +1,33 @@
 """Module containing the Trainer class and related functions"""
-
 import importlib
 import logging
 import math
 import os
 import sys
+from contextlib import contextmanager
 from dataclasses import dataclass, field
+from functools import partial
 from pathlib import Path
-from typing import Optional
+from typing import Optional, Union
 
 import bitsandbytes as bnb
+import numpy as np
 import torch.cuda
 import transformers
+from datasets import Dataset, set_caching_enabled
 from torch import nn
 from torch.optim.lr_scheduler import OneCycleLR
+from torch.utils.data import DataLoader, DistributedSampler, RandomSampler
 from transformers import EarlyStoppingCallback, Trainer, TrainingArguments
 from transformers.trainer_pt_utils import get_parameter_names
 
 from axolotl.utils.callbacks import (
+    GPUStatsCallback,
     SaveBetterTransformerModelCallback,
     SavePeftModelCallback,
 )
+from axolotl.utils.collators import DataCollatorForSeq2Seq
+from axolotl.utils.dataloader import MultipackDistributedDataloader
 from axolotl.utils.schedulers import (
     InterpolatingLogScheduler,
     get_cosine_schedule_with_quadratic_warmup,
@@ -29,6 +36,68 @@ from axolotl.utils.schedulers import (
 LOG = logging.getLogger("axolotl")
 
 
+@torch.jit.script
+def weighted_cross_entropy(
+    logits: torch.Tensor, labels: torch.Tensor, weights: torch.Tensor
+):
+    # Flatten the logits, labels, and weights tensors
+    logits = logits.view(
+        -1, logits.size(-1)
+    )  # logits becomes of shape [batch_size*sequence_length, vocab_size]
+    labels = labels.view(-1)  # labels becomes of shape [batch_size*sequence_length]
+    weights = weights.view(-1)  # weights becomes of shape [batch_size*sequence_length]
+
+    # Compute the unweighted cross entropy loss
+    losses = torch.nn.functional.cross_entropy(logits, labels, reduction="none")
+
+    # Apply the weights to the losses and compute their sum
+    return (weights * losses).sum()
+
+
+@torch.jit.script
+def create_weighted_mask(labels: torch.Tensor):
+    # Check if the tensor is 2D. If not, unsqueeze it to make it 2D
+    if len(labels.shape) == 1:
+        labels = labels.unsqueeze(0)
+
+    weights = torch.zeros_like(labels).float()
+    for i in range(labels.shape[0]):
+        mask = labels[i] != -100
+
+        # Create a tensor to track group ids
+        group_ids = torch.zeros_like(labels[i]).int()
+        curr_group_id = 0
+
+        for j in range(1, len(labels[i])):
+            if mask[j] and not mask[j - 1]:  # switch from masked to unmasked label
+                curr_group_id += 1  # start new group
+            group_ids[j] = (
+                curr_group_id if mask[j] else 0
+            )  # assign group id if unmasked label
+
+        # Count only unmasked labels in each group
+        group_counts = torch.bincount(group_ids[mask])
+
+        mask_weights = torch.zeros_like(labels[i]).float()
+        mask_weights[mask] = 1.0 / group_counts[group_ids[mask]]
+
+        weights[i] = mask_weights
+
+    return weights.squeeze()  # squeeze the output to match the input dimension
+
+
+def trainer_weighted_loss(model_output, labels, shift_labels=True):
+    logits = (
+        model_output["logits"] if isinstance(model_output, dict) else model_output[0]
+    )
+    if shift_labels:
+        logits = logits[..., :-1, :].contiguous()
+        labels = labels[..., 1:].contiguous()
+
+    weights = create_weighted_mask(labels)
+    return weighted_cross_entropy(logits, labels, weights)
+
+
 @dataclass
 class AxolotlTrainingArguments(TrainingArguments):
     """
@@ -39,6 +108,22 @@ class AxolotlTrainingArguments(TrainingArguments):
         default=False,
         metadata={"help": "Use quadratic warmup for cosine scheduling."},
     )
+    sample_packing: bool = field(
+        default=False,
+        metadata={"help": "Use sample packing for efficient training."},
+    )
+    sample_packing_efficiency: float = field(
+        default=1.0,
+        metadata={"help": "Sample packing efficiency for calculating batch length."},
+    )
+    max_seq_length: int = field(
+        default=2048,
+        metadata={"help": "The maximum sequence length the model can handle"},
+    )
+    sample_packing_seq_len_multiplier: int = field(
+        default=1,
+        metadata={"help": "the multiplier for the max len for packed sequences"},
+    )
 
 
 class AxolotlTrainer(Trainer):
@@ -76,6 +161,64 @@ class AxolotlTrainer(Trainer):
                 return super().create_scheduler(num_training_steps, optimizer)
         return self.lr_scheduler
 
+    def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
+        if self.args.world_size > 1 and self.args.sample_packing:
+            return DistributedSampler(
+                self.train_dataset,
+                num_replicas=self.args.world_size,
+                rank=self.args.process_index,
+                seed=self.args.seed,
+            )
+        return super()._get_train_sampler()
+
+    def get_train_dataloader(self) -> Union[DataLoader, MultipackDistributedDataloader]:
+        if self.args.sample_packing:
+            train_sampler = self._get_train_sampler()
+            return self.accelerator.prepare(
+                MultipackDistributedDataloader(
+                    self.train_dataset,
+                    batch_size=self._train_batch_size,
+                    seq_max_length=self.args.max_seq_length,
+                    collate_fn=self.data_collator,
+                    sampler=train_sampler,
+                    packing_efficiency_estimate=self.args.sample_packing_efficiency,
+                    sample_packing_seq_len_multiplier=self.args.sample_packing_seq_len_multiplier,
+                    device_count=int(os.environ.get("WORLD_SIZE", 1)),
+                )
+            )
+        return super().get_train_dataloader()
+
+    def get_eval_dataloader(
+        self, eval_dataset: Optional[Dataset] = None
+    ) -> Union[DataLoader, MultipackDistributedDataloader]:
+        if self.args.sample_packing:
+            eval_dataset = (
+                eval_dataset if eval_dataset is not None else self.eval_dataset
+            )
+            eval_sampler = self._get_eval_sampler(eval_dataset)
+            return self.accelerator.prepare(
+                MultipackDistributedDataloader(
+                    eval_dataset,
+                    batch_size=self.args.eval_batch_size,
+                    seq_max_length=self.args.max_seq_length,
+                    collate_fn=self.data_collator,
+                    sampler=eval_sampler,
+                    packing_efficiency_estimate=self.args.sample_packing_efficiency,
+                    sample_packing_seq_len_multiplier=self.args.eval_batch_size,
+                    device_count=int(os.environ.get("WORLD_SIZE", 1)),
+                )
+            )
+        return super().get_eval_dataloader(eval_dataset)
+
+    def compute_loss(self, model, inputs, return_outputs=False):
+        # use one's weighted cross entropy loss calc
+        # if self.args.sample_packing:
+        #     labels = inputs.pop("labels")
+        #     outputs = model(**inputs)
+        #     loss = trainer_weighted_loss(outputs, labels, shift_labels=True)
+        #     return (loss, outputs) if return_outputs else loss
+        return super().compute_loss(model, inputs, return_outputs=return_outputs)
+
 
 class OneCycleLRSchedulerTrainer(AxolotlTrainer):
     """
@@ -106,10 +249,121 @@ class OneCycleLRSchedulerTrainer(AxolotlTrainer):
         return self.lr_scheduler
 
 
-def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
-    total_num_steps = int(
-        math.ceil(len(train_dataset) * cfg.num_epochs / cfg.batch_size)
-    )
+def add_position_ids(sample):
+    sample["position_ids"] = torch.arange(len(sample["input_ids"]))
+    return sample
+
+
+def drop_long_seq(sample, sequence_len=2048):
+    return len(sample["input_ids"]) <= sequence_len
+
+
+@contextmanager
+def disable_datasets_caching():
+    try:
+        set_caching_enabled(False)
+        yield
+    finally:
+        set_caching_enabled(True)
+
+
+def process_datasets_for_packing(cfg, train_dataset, eval_dataset):
+    if cfg.sample_packing:
+        drop_long = partial(drop_long_seq, sequence_len=cfg.sequence_len)
+        train_dataset = train_dataset.filter(drop_long, num_proc=os.cpu_count()).map(
+            add_position_ids, num_proc=os.cpu_count()
+        )
+        if eval_dataset:
+            eval_dataset = eval_dataset.filter(drop_long, num_proc=os.cpu_count()).map(
+                add_position_ids, num_proc=os.cpu_count()
+            )
+    return train_dataset, eval_dataset
+
+
+def calculate_total_num_steps(cfg, train_dataset, tokenizer):
+    if cfg.sample_packing:
+        # we have to drop anything longer then sequence len otherwise
+        # flash attention with position ids fails
+        if not cfg.total_num_tokens:
+            LOG.info("calculating total_num_tokens")
+            total_num_tokens = np.sum(
+                train_dataset.data.column("input_ids")
+                .to_pandas()
+                .apply(lambda x: len(x))  # pylint: disable=unnecessary-lambda
+                .values
+            )
+            LOG.info(f"📝 UPDATE CONFIG WITH: `total_num_tokens: {total_num_tokens}`")
+            cfg.total_num_tokens = total_num_tokens
+
+        if cfg.sample_packing_eff_est:
+            total_num_steps = (
+                # match count to len est in dataloader
+                (
+                    math.floor(
+                        0.99
+                        * cfg.total_num_tokens
+                        / cfg.sample_packing_eff_est
+                        / cfg.sequence_len
+                        // cfg.batch_size
+                        // int(os.environ.get("WORLD_SIZE", 1))
+                    )
+                    - 1
+                )
+                * cfg.num_epochs
+            )
+            LOG.info(
+                f"total_num_tokens: {cfg.total_num_tokens}, total_num_steps: {total_num_steps}"
+            )
+        else:
+            sampler = RandomSampler(train_dataset)
+            data_loader = MultipackDistributedDataloader(
+                train_dataset,
+                batch_size=cfg.micro_batch_size,
+                seq_max_length=cfg.max_packed_sequence_len or cfg.sequence_len,
+                collate_fn=DataCollatorForSeq2Seq(
+                    tokenizer,
+                    return_tensors="pt",
+                    padding="longest",
+                ),
+                sampler=sampler,
+                packing_efficiency_estimate=cfg.sample_packing_eff_est,
+                sample_packing_seq_len_multiplier=cfg.micro_batch_size,
+                device_count=int(os.environ.get("WORLD_SIZE", 1)),
+            )
+            data_loader_len = data_loader.len_w_stats()
+            actual_eff = data_loader.efficiency()
+            LOG.info(f"data_loader_len: {data_loader_len}")
+            total_num_steps = int(
+                math.floor(
+                    data_loader_len
+                    * cfg.micro_batch_size
+                    * cfg.num_epochs
+                    // cfg.batch_size
+                )
+            )
+            LOG.info(
+                f"📝 UPDATE CONFIG WITH: `sample_packing_eff_est: {math.ceil(actual_eff * 100.0) / 100.0}`"
+            )
+            cfg.sample_packing_eff_est = math.ceil(actual_eff * 100.0) / 100.0
+    else:
+        total_num_steps = int(
+            math.ceil(len(train_dataset) * cfg.num_epochs / cfg.batch_size)
+        )
+    LOG.info(f"total_num_steps: {total_num_steps}")
+    return total_num_steps
+
+
+def setup_fsdp_envs(cfg):
+    os.environ["ACCELERATE_USE_FSDP"] = "true"
+    if cfg.fsdp_config.fsdp_sync_module_states:
+        os.environ["FSDP_SYNC_MODULE_STATES"] = "true"
+    if cfg.fsdp_config.fsdp_state_dict_type:
+        os.environ["FSDP_STATE_DICT_TYPE"] = cfg.fsdp_config.fsdp_state_dict_type
+
+
+def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer, total_num_steps):
+    if cfg.fsdp:
+        setup_fsdp_envs(cfg)
     warmup_steps = (
         cfg.warmup_steps
         if cfg.warmup_steps is not None
@@ -189,7 +443,14 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
     if cfg.save_safetensors:
         training_arguments_kwargs["save_safetensors"] = cfg.save_safetensors
 
+    if cfg.sample_packing_eff_est:
+        training_arguments_kwargs[
+            "sample_packing_efficiency"
+        ] = cfg.sample_packing_eff_est
+
     training_args = AxolotlTrainingArguments(  # pylint: disable=unexpected-keyword-arg
+        # max_steps=total_num_steps,  # this is helpful in case we don't actually know total # of steps
+        max_seq_length=cfg.sequence_len,
         per_device_train_batch_size=cfg.micro_batch_size,
         per_device_eval_batch_size=cfg.eval_batch_size
         if cfg.eval_batch_size is not None
@@ -203,7 +464,7 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         eval_steps=cfg.eval_steps if cfg.val_set_size > 0 else None,
         save_steps=cfg.save_steps,
         output_dir=cfg.output_dir,
-        save_total_limit=3,
+        save_total_limit=cfg.save_total_limit if cfg.save_total_limit else 4,
         load_best_model_at_end=(
             cfg.load_best_model_at_end is not False
             and cfg.val_set_size > 0
@@ -221,6 +482,8 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         if cfg.lr_scheduler and cfg.lr_scheduler not in ("one_cycle", "log_sweep")
         else "cosine",
         weight_decay=cfg.weight_decay if cfg.weight_decay is not None else 0.0,
+        sample_packing=cfg.sample_packing if cfg.sample_packing else False,
+        sample_packing_seq_len_multiplier=cfg.micro_batch_size,
         **training_arguments_kwargs,
     )
 
@@ -292,6 +555,7 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         trainer_kwargs["optimizers"] = (optimizer, lr_scheduler)
 
     callbacks = []
+    callbacks.append(GPUStatsCallback(cfg))
     # TODO on_save callback to sync checkpoints to GCP/AWS in background
     if cfg.early_stopping_patience:
         early_stop_cb = EarlyStoppingCallback(
@@ -314,11 +578,11 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
     if cfg.collator_pad_to_longest:
         data_collator_kwargs["padding"] = "longest"
     else:
-        data_collator_kwargs["pad_to_multiple_of"] = 8
+        # A100 is best at 64, while others at 8. Let's use the larger so we don't have to check
+        # https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html
+        data_collator_kwargs["pad_to_multiple_of"] = 64
 
     if cfg.is_llama_derived_model and cfg.landmark_attention:
-        from functools import partial
-
         from axolotl.monkeypatch.llama_landmark_attn import (
             add_mem_tokens,
             get_mem_id,
@@ -346,7 +610,7 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer):
         train_dataset=train_dataset,
         eval_dataset=eval_dataset,
         args=training_args,
-        data_collator=transformers.DataCollatorForSeq2Seq(
+        data_collator=DataCollatorForSeq2Seq(
             tokenizer,
             return_tensors="pt",
             **data_collator_kwargs,

src/axolotl/utils/wandb.py

@@ -9,6 +9,8 @@ def setup_wandb_env_vars(cfg):
     elif cfg.wandb_project and len(cfg.wandb_project) > 0:
         os.environ["WANDB_PROJECT"] = cfg.wandb_project
         cfg.use_wandb = True
+        if cfg.wandb_entity and len(cfg.wandb_entity) > 0:
+            os.environ["WANDB_ENTITY"] = cfg.wandb_entity
         if cfg.wandb_watch and len(cfg.wandb_watch) > 0:
             os.environ["WANDB_WATCH"] = cfg.wandb_watch
         if cfg.wandb_log_model and len(cfg.wandb_log_model) > 0:

tests/monkeypatch/test_llama_attn_hijack_flash.py

@@ -0,0 +1,30 @@
+"""
+Unit tests for the monkeypatch utils
+"""
+import unittest
+
+import torch
+
+from axolotl.monkeypatch.utils import get_cu_seqlens, get_cu_seqlens_from_pos_ids
+
+
+class TestMonkeyPatchUtils(unittest.TestCase):
+    """
+    Unit test class for monkeypatch utils
+    """
+
+    def test_get_cu_seqlens_1d(self):
+        attn_mask = torch.tensor([[1, 1, 1, 1, 2, 2, 2, 3, 3, 3, 3, 3, 4, 4, 0, 0]])
+        target_res = torch.tensor([0, 4, 7, 12, 14, 16], dtype=torch.int32)
+        self.assertTrue(torch.allclose(get_cu_seqlens(attn_mask)[0], target_res))
+
+    def test_get_cu_seqlens_from_pos_ids_1d(self):
+        position_ids = torch.tensor([[0, 1, 2, 3, 0, 1, 2, 0, 1, 2, 3, 4, 0, 1, 0, 0]])
+        target_res = torch.tensor([0, 4, 7, 12, 14, 16], dtype=torch.int32)
+        self.assertTrue(
+            torch.allclose(get_cu_seqlens_from_pos_ids(position_ids)[0], target_res)
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_dict.py

@@ -72,6 +72,13 @@ class DictDefaultTest(unittest.TestCase):
 
         assert cfg.random_key is None, "DictDefault should return None for missing keys"
 
+    def test_dict_or(self):
+        cfg = DictDefault({}) | DictDefault({})
+
+        assert (
+            cfg.random_key is None
+        ), "DictDefault should return None for missing keys after | operation"
+
     def test_dict_nested_missingparentkey(self):
         """
         Due to subclassing Dict, DictDefault will error if we try to access a nested key whose parent key does not exist.

tests/test_expand_mask.py

@@ -0,0 +1,44 @@
+"""
+Unit tests for the monkey patch for expand mask to handle packed sequences
+"""
+import unittest
+
+import torch
+
+from axolotl.monkeypatch.llama_expand_mask import _expand_mask
+
+
+class TestExpandMask(unittest.TestCase):
+    """
+    Test class for attention mask expansion for packed sequences
+    """
+
+    def test_output(self):
+        mask = torch.tensor([[1, 1, 1, 2], [2, 3, 3, 0]])
+        dtype = torch.float32
+        expected_output = torch.tensor(
+            [
+                [
+                    [
+                        [0.0000e00, -3.4028e38, -3.4028e38, -3.4028e38],
+                        [0.0000e00, 0.0000e00, -3.4028e38, -3.4028e38],
+                        [0.0000e00, 0.0000e00, 0.0000e00, -3.4028e38],
+                        [-3.4028e38, -3.4028e38, -3.4028e38, 0.0000e00],
+                    ]
+                ],
+                [
+                    [
+                        [0.0000e00, -3.4028e38, -3.4028e38, -3.4028e38],
+                        [-3.4028e38, 0.0000e00, -3.4028e38, -3.4028e38],
+                        [-3.4028e38, 0.0000e00, 0.0000e00, -3.4028e38],
+                        [-3.4028e38, -3.4028e38, -3.4028e38, -3.4028e38],
+                    ]
+                ],
+            ]
+        )
+        # Check that the output matches the expected output
+        self.assertTrue(torch.allclose(_expand_mask(mask, dtype), expected_output))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_packed_dataset.py

@@ -27,7 +27,7 @@ class TestPacking(unittest.TestCase):
             }
         )
 
-    def test_resets_attention(self):
+    def test_increments_attention(self):
         prompter = AlpacaPrompter("chat")
         strat = AlpacaPromptTokenizingStrategy(
             prompter,
@@ -55,10 +55,14 @@ class TestPacking(unittest.TestCase):
         # first example doesn't have mask reset
         assert example["input_ids"][0] == self.tokenizer.bos_token_id
         assert example["attention_mask"][0] == 1
+        assert example["position_ids"][0] == 0
+        assert example["position_ids"][1] == 1
 
         # but subsequent one does
         assert example["input_ids"][next_bos_index] == self.tokenizer.bos_token_id
-        assert example["attention_mask"][next_bos_index] == 0
+        assert example["attention_mask"][next_bos_index] == 2
+        assert example["position_ids"][next_bos_index] == 0
+        assert example["position_ids"][next_bos_index + 1] == 1
 
 
 if __name__ == "__main__":

tests/test_prompt_tokenizers.py

@@ -4,13 +4,17 @@ import logging
 import unittest
 from pathlib import Path
 
-from transformers import AutoTokenizer
+from transformers import AutoTokenizer, LlamaTokenizer
 
 from axolotl.prompt_strategies.alpaca_chat import NoSystemPrompter
 from axolotl.prompt_strategies.alpaca_w_system import (
     InstructionWSystemPromptTokenizingStrategy,
     SystemDataPrompter,
 )
+from axolotl.prompt_strategies.llama2_chat import (
+    Llama2ChatPrompter,
+    LLama2ChatTokenizingStrategy,
+)
 from axolotl.prompt_tokenizers import (
     AlpacaPromptTokenizingStrategy,
     ShareGPTPromptTokenizingStrategy,
@@ -130,9 +134,95 @@ class InstructionWSystemPromptTokenizingStrategyTest(unittest.TestCase):
             "output": "Hi! How can I help?",
         }
         example = strat.tokenize_prompt(sample)
-        assert example["input_ids"][0:4] == [1, 835, 2184, 29901]  # "<s>### System:"
-        assert example["input_ids"][5:7] == [1509, 20118]  # "use cot"
-        assert example["input_ids"][9] == 11889  # USER
+        assert example["input_ids"][0:5] == [
+            1,
+            28962,
+            1254,
+            12665,
+            29901,
+        ]  # "<s>SYSTEM:"
+        assert example["input_ids"][5:7] == [671, 20118]  # " use cot"
+        assert example["input_ids"][8] == 11889  # USER
+
+
+class Llama2ChatTokenizationTest(unittest.TestCase):
+    """
+    Test class for prompt tokenization strategies with sys prompt from the dataset
+    """
+
+    def setUp(self) -> None:
+        # pylint: disable=duplicate-code
+        self.tokenizer = LlamaTokenizer.from_pretrained("NousResearch/Llama-2-7b-hf")
+        # woraround because official Meta repos are not open
+
+    def test_llama2_chat_integration(self):
+        with open(
+            Path(__file__).parent / "fixtures/conversation.json", encoding="utf-8"
+        ) as fin:
+            data = fin.read()
+            conversation = json.loads(data)
+        with open(
+            Path(__file__).parent / "fixtures/conversation.tokenized_llama2chat.json",
+            encoding="utf-8",
+        ) as fin:
+            data = fin.read()
+            tokenized_conversation = json.loads(data)
+        prompter = Llama2ChatPrompter()
+        strat = LLama2ChatTokenizingStrategy(
+            prompter,
+            self.tokenizer,
+            False,
+            4096,
+        )
+        example = strat.tokenize_prompt(conversation)
+        for fields in ["input_ids", "attention_mask", "labels"]:
+            self.assertEqual(len(example[fields]), len(tokenized_conversation[fields]))
+            self.assertEqual(example[fields], tokenized_conversation[fields])
+
+    def compare_with_transformers_integration(self):
+        # this needs transformers >= v4.31.0
+        from transformers.models.llama.tokenization_llama import B_SYS, E_SYS
+        from transformers.pipelines.conversational import Conversation
+
+        # from transformers.models.llama.tokenization_llama import DEFAULT_SYSTEM_PROMPT
+        # broken as of 23/7/20
+        # see https://github.com/huggingface/transformers/pull/24935
+        # pylint: disable=C0103
+        DEFAULT_SYSTEM_PROMPT = """\
+You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.
+
+If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information."""
+        with open(
+            Path(__file__).parent / "fixtures/conversation.json", encoding="utf-8"
+        ) as fin:
+            data = fin.read()
+            conversation = json.loads(data)
+        with open(
+            Path(__file__).parent / "fixtures/conversation.tokenized_llama2chat.json",
+            encoding="utf-8",
+        ) as fin:
+            data = fin.read()
+            tokenized_conversation = json.loads(data)
+
+        user_input = []
+        answers = []
+        for msg in conversation["conversations"]:
+            if msg["from"] == "human":
+                user_input.append(msg["value"])
+            else:
+                answers.append(msg["value"])
+        hf_conf = Conversation(
+            text=user_input[-1],
+            past_user_inputs=[B_SYS + DEFAULT_SYSTEM_PROMPT + E_SYS + user_input[0]]
+            + user_input[1:-1],
+            generated_responses=answers,
+        )
+        # pylint: disable=W0212
+        hf_tokens = self.tokenizer._build_conversation_input_ids(hf_conf)
+
+        self.assertEqual(
+            hf_tokens, tokenized_conversation["input_ids"][: len(hf_tokens)]
+        )
 
 
 if __name__ == "__main__":

tests/test_prompters.py

@@ -70,7 +70,7 @@ class AlpacaPrompterTest(unittest.TestCase):
             )
         )
         assert "use cot" in res
-        assert res.startswith("### System:")
+        assert res.startswith("SYSTEM:")
         assert "### Instruction:" not in res
         assert "### Input:" not in res
         assert "alpacas" in res

tests/test_tokenizers.py

@@ -13,17 +13,22 @@ class TestTokenizers(unittest.TestCase):
     """
 
     def test_default_use_fast(self):
-        cfg = DictDefault({})
-        tokenizer = load_tokenizer("huggyllama/llama-7b", None, cfg)
+        cfg = DictDefault(
+            {
+                "tokenizer_config": "huggyllama/llama-7b",
+            }
+        )
+        tokenizer = load_tokenizer(cfg)
         assert "Fast" in tokenizer.__class__.__name__
 
     def test_dont_use_fast(self):
         cfg = DictDefault(
             {
+                "tokenizer_config": "huggyllama/llama-7b",
                 "tokenizer_use_fast": False,
             }
         )
-        tokenizer = load_tokenizer("huggyllama/llama-7b", None, cfg)
+        tokenizer = load_tokenizer(cfg)
         assert "Fast" not in tokenizer.__class__.__name__
 
 

tests/test_validation.py

@@ -6,8 +6,8 @@ from typing import Optional
 
 import pytest
 
+from axolotl.utils.config import validate_config
 from axolotl.utils.dict import DictDefault
-from axolotl.utils.validation import validate_config
 
 
 class ValidationTest(unittest.TestCase):
@@ -313,3 +313,27 @@ class ValidationTest(unittest.TestCase):
         )
 
         validate_config(cfg)
+
+    def test_packing(self):
+        cfg = DictDefault(
+            {
+                "max_packed_sequence_len": 2048,
+            }
+        )
+        with self._caplog.at_level(logging.WARNING):
+            validate_config(cfg)
+            assert any(
+                "max_packed_sequence_len will be deprecated in favor of sample_packing"
+                in record.message
+                for record in self._caplog.records
+            )
+
+        cfg = DictDefault(
+            {
+                "max_packed_sequence_len": 2048,
+                "sample_packing": True,
+            }
+        )
+        regex_exp = r".*set only one of max_packed_sequence_len \(deprecated soon\) or sample_packing.*"
+        with pytest.raises(ValueError, match=regex_exp):
+            validate_config(cfg)