support for json data as completion

* Feat: Add Qwen

* feat: add qwen lora example

* feat: update matrix

* fix: add trust_remote_code

* fix: disable gradient checkpointing

* chore: add warning about gradient checkpointing

* fix: config

* fix: turn off sample packing for this example and reduce seq len

* chore: add comment on seq len

.github/workflows/tests.yml

@@ -71,6 +71,7 @@ jobs:
 
       - name: Install dependencies
         run: |
+          pip3 install --extra-index-url https://download.pytorch.org/whl/cu118 -U torch==2.0.1
           pip3 uninstall -y transformers accelerate
           pip3 install -U -e .[flash-attn]
           pip3 install -r requirements-tests.txt

README.md

@@ -77,6 +77,7 @@ Features:
 | XGen     | ✅         | ❓    | ✅     | ❓             | ❓                 | ❓          | ✅            |
 | phi      | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 | RWKV     | ✅         | ❓    | ❓     | ❓             | ❓                 | ❓          | ❓            |
+| Qwen     | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 
 
 ## Quickstart ⚡
@@ -85,14 +86,19 @@ Get started with Axolotl in just a few steps! This quickstart guide will walk yo
 
 **Requirements**: Python >=3.9 and Pytorch >=2.0.
 
+`pip3 install "axolotl[flash-attn,deepspeed] @ git+https://github.com/OpenAccess-AI-Collective/axolotl"`
+
+### For developers
 ```bash
 git clone https://github.com/OpenAccess-AI-Collective/axolotl
 cd axolotl
 
 pip3 install packaging
 pip3 install -e '.[flash-attn,deepspeed]'
-pip3 install -U git+https://github.com/huggingface/peft.git
+```
 
+### Usage
+```bash
 # finetune lora
 accelerate launch -m axolotl.cli.train examples/openllama-3b/lora.yml
 
@@ -494,6 +500,7 @@ is_falcon_derived_model:
 is_llama_derived_model:
 # Please note that if you set this to true, `padding_side` will be set to "left" by default
 is_mistral_derived_model:
+is_qwen_derived_model:
 
 # optional overrides to the base model configuration
 model_config:
@@ -538,6 +545,8 @@ datasets:
 
     # Optional[str] fastchat conversation type, only used with type: sharegpt
     conversation:  # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
+    field_human: # Optional[str]. Human key to use for conversation.
+    field_model: # Optional[str]. Assistant key to use for conversation.
 
   # Custom user prompt
   - path: repo
@@ -668,7 +677,8 @@ gradient_accumulation_steps: 1
 micro_batch_size: 2
 eval_batch_size:
 num_epochs: 4
-warmup_steps: 100
+warmup_steps: 100  # cannot use with warmup_ratio
+warmup_ratio: 0.05  # cannot use with warmup_steps
 learning_rate: 0.00003
 lr_quadratic_warmup:
 logging_steps:

examples/llama-2/tiny-llama.yml

@@ -4,20 +4,19 @@ model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
 is_llama_derived_model: true
 
-load_in_8bit: false
+load_in_8bit: true
 load_in_4bit: false
 strict: false
 
 datasets:
-  - path: mhenrichsen/context-aware-splits-english
+  - path: mhenrichsen/alpaca_2k_test
     type: alpaca
 dataset_prepared_path:
-val_set_size: 200
-output_dir: ./tiny-llama
+val_set_size: 0.05
+output_dir: ./lora-out
 
-sequence_len: 8192
+sequence_len: 4096
 sample_packing: true
-pad_to_sequence_len: true
 
 adapter: lora
 lora_model_dir:
@@ -33,9 +32,9 @@ wandb_watch:
 wandb_run_id:
 wandb_log_model:
 
-gradient_accumulation_steps: 1
-micro_batch_size: 8
-num_epochs: 3
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
 optimizer: adamw_bnb_8bit
 lr_scheduler: cosine
 learning_rate: 0.0002
@@ -54,13 +53,13 @@ logging_steps: 1
 xformers_attention:
 flash_attention: true
 
-warmup_steps: 50
+warmup_steps: 10
 eval_steps: 0.05
 eval_table_size:
-save_steps: 0.50
+save_steps:
 debug:
 deepspeed:
-weight_decay: 0.1
+weight_decay: 0.0
 fsdp:
 fsdp_config:
 special_tokens:

examples/phi/phi-ft.yml

@@ -1,5 +1,5 @@
 base_model: microsoft/phi-1_5
-model_type: MixFormerSequentialForCausalLM
+model_type: PhiForCausalLM
 tokenizer_type: AutoTokenizer
 is_llama_derived_model: false
 trust_remote_code: true

examples/qwen/lora.yml

@@ -0,0 +1,68 @@
+base_model: Qwen/Qwen-7B
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+is_qwen_derived_model: true
+trust_remote_code: true
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./lora-out
+
+sequence_len: 2048  # supports up to 8192
+sample_packing: false
+pad_to_sequence_len:
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_run_id:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+fp16: false
+tf32: false
+
+gradient_checkpointing: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+eval_steps: 0.05
+eval_table_size:
+eval_table_max_new_tokens: 128
+save_steps:
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/qwen/qlora.yml

@@ -0,0 +1,68 @@
+base_model: Qwen/Qwen-7B
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+is_qwen_derived_model: true
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./lora-out
+
+sequence_len: 2048  # supports up to 8192
+sample_packing: false
+pad_to_sequence_len:
+
+adapter: qlora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_run_id:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+fp16: false
+tf32: false
+
+gradient_checkpointing: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+eval_steps: 0.05
+eval_table_size:
+eval_table_max_new_tokens: 128
+save_steps:
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

requirements.txt

@@ -1,22 +1,20 @@
---extra-index-url https://download.pytorch.org/whl/cu118
 --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
-torch==2.0.1
-auto-gptq==0.4.2
+auto-gptq==0.5.1
 packaging
 peft==0.6.0
-transformers @ git+https://github.com/huggingface/transformers.git@acc394c4f5e1283c19783581790b3dc3105a3697
+transformers==4.35.1
 bitsandbytes>=0.41.1
-accelerate @ git+https://github.com/huggingface/accelerate@80da9cfb09bb3cc9f1b385cb55d6b90d025a5fd9
+accelerate==0.24.1
 deepspeed
 addict
 fire
 PyYAML>=6.0
 datasets>=2.14.0
-flash-attn>=2.3.0
+flash-attn==2.3.3
 sentencepiece
 wandb
 einops
-xformers>=0.0.22
+xformers==0.0.22
 optimum==1.13.2
 hf_transfer
 colorama

src/axolotl/core/trainer_builder.py

@@ -461,11 +461,14 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
         return AxolotlTrainer
 
     def build(self, total_num_steps):
-        warmup_steps = (
-            self.cfg.warmup_steps
-            if self.cfg.warmup_steps is not None
-            else min(int(0.03 * total_num_steps), 100)
-        )
+        warmup_steps = None
+        if self.cfg.warmup_steps is not None:
+            warmup_steps = self.cfg.warmup_steps
+        elif self.cfg.warmup_ratio is not None:
+            warmup_steps = max(int(self.cfg.warmup_ratio * total_num_steps), 0)
+        else:
+            warmup_steps = min(int(0.03 * total_num_steps), 100)
+
         logging_steps = (
             self.cfg.logging_steps
             if self.cfg.logging_steps is not None
@@ -658,7 +661,9 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
             self.cfg.sample_packing if self.cfg.sample_packing else False
         )
         training_arguments_kwargs["eval_sample_packing"] = (
-            self.cfg.sample_packing if self.cfg.sample_packing else False
+            self.cfg.sample_packing
+            if self.cfg.eval_sample_packing is not False
+            else False
         )
         training_arguments_kwargs[
             "sample_packing_seq_len_multiplier"

src/axolotl/models/phi/__init__.py

@@ -3,4 +3,6 @@ MixFormers model architecture used for phi models
 """
 
 from .configuration_mixformer_sequential import MixFormerSequentialConfig  # noqa
+from .configuration_phi import PhiConfig  # noqa
 from .modeling_mixformer_sequential import MixFormerSequentialForCausalLM  # noqa
+from .modeling_phi import PhiForCausalLM  # noqa

src/axolotl/models/phi/configuration_phi.py

@@ -0,0 +1,65 @@
+# pylint: skip-file
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT license.
+
+import math
+from typing import Optional
+
+from transformers import PretrainedConfig
+
+
+class PhiConfig(PretrainedConfig):
+    """Phi configuration."""
+
+    model_type = "phi"
+    attribute_map = {
+        "max_position_embeddings": "n_positions",
+        "hidden_size": "n_embd",
+        "num_attention_heads": "n_head",
+        "num_hidden_layers": "n_layer",
+    }
+
+    def __init__(
+        self,
+        vocab_size: int = 50304,
+        n_positions: int = 2048,
+        n_embd: int = 1024,
+        n_layer: int = 20,
+        n_inner: Optional[int] = None,
+        n_head: int = 16,
+        n_head_kv: Optional[int] = None,
+        rotary_dim: Optional[int] = 32,
+        activation_function: Optional[str] = "gelu_new",
+        flash_attn: bool = False,
+        flash_rotary: bool = False,
+        fused_dense: bool = False,
+        attn_pdrop: float = 0.0,
+        embd_pdrop: float = 0.0,
+        resid_pdrop: float = 0.0,
+        layer_norm_epsilon: float = 1e-5,
+        initializer_range: float = 0.02,
+        tie_word_embeddings: bool = False,
+        pad_vocab_size_multiple: int = 64,
+        **kwargs
+    ) -> None:
+        self.vocab_size = int(
+            math.ceil(vocab_size / pad_vocab_size_multiple) * pad_vocab_size_multiple
+        )
+        self.n_positions = n_positions
+        self.n_embd = n_embd
+        self.n_layer = n_layer
+        self.n_inner = n_inner
+        self.n_head = n_head
+        self.n_head_kv = n_head_kv
+        self.rotary_dim = min(rotary_dim, n_embd // n_head)
+        self.activation_function = activation_function
+        self.flash_attn = flash_attn
+        self.flash_rotary = flash_rotary
+        self.fused_dense = fused_dense
+        self.attn_pdrop = attn_pdrop
+        self.embd_pdrop = embd_pdrop
+        self.resid_pdrop = resid_pdrop
+        self.layer_norm_epsilon = layer_norm_epsilon
+        self.initializer_range = initializer_range
+
+        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)

src/axolotl/prompt_strategies/completion.py

@@ -1,6 +1,7 @@
 """
 Basic completion text
 """
+import json
 from collections import defaultdict
 from typing import Any, Dict, Generator, Optional, Tuple
 
@@ -64,6 +65,19 @@ class CompletionPromptTokenizingStrategy(InstructionPromptTokenizingStrategy):
         return next(iter(self.prompter.build_prompt(instruction, input, response)))
 
 
+class CompletionJSONPromptTokenizationStrategy(CompletionPromptTokenizingStrategy):
+    """
+    Strategy to return the stringified JSON of the entire row as the training data
+    """
+
+    def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
+        return (
+            json.dumps(prompt),
+            "",
+            "",
+        )
+
+
 class CompletionPrompter:
     """
     Prompter for completion
@@ -82,7 +96,7 @@ def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):
     strat = CompletionPromptTokenizingStrategy(
         CompletionPrompter(),
         tokenizer,
-        cfg.train_on_inputs,
+        True,
         cfg.sequence_len,
         max_length=cfg.sequence_len * 64,
     )
@@ -90,3 +104,15 @@ def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):
         strat.field = ds_cfg["field"]
 
     return strat
+
+
+def load_json(tokenizer, cfg):
+    strat = CompletionJSONPromptTokenizationStrategy(
+        CompletionPrompter(),
+        tokenizer,
+        True,
+        cfg.sequence_len,
+        max_length=cfg.sequence_len * 64,
+    )
+
+    return strat

src/axolotl/utils/config.py

@@ -122,6 +122,19 @@ def normalize_config(cfg):
         or (cfg.model_type and "mistral" in cfg.model_type.lower())
     )
 
+    cfg.is_qwen_derived_model = (
+        (
+            hasattr(model_config, "model_type")
+            and model_config.model_type
+            in [
+                "qwen",
+            ]
+        )
+        or cfg.is_qwen_derived_model
+        or "qwen" in cfg.base_model.lower()
+        or (cfg.model_type and "qwen" in cfg.model_type.lower())
+    )
+
     if isinstance(cfg.learning_rate, str):
         cfg.learning_rate = float(cfg.learning_rate)
 
@@ -165,7 +178,11 @@ def validate_config(cfg):
             "batch_size is not recommended. Please use gradient_accumulation_steps instead.",
             "To calculate the equivalent gradient_accumulation_steps, divide batch_size / micro_batch_size / number of gpus.",
         )
-    if cfg.eval_batch_size != cfg.micro_batch_size:
+    if (
+        cfg.eval_batch_size
+        and cfg.micro_batch_size
+        and cfg.eval_batch_size != cfg.micro_batch_size
+    ):
         LOG.warning(
             "eval_batch_size != micro_batch_size. This can lead to VRAM instability."
         )
@@ -372,6 +389,14 @@ def validate_config(cfg):
     if cfg.rope_scaling:
         LOG.warning("`rope_scaling` should now be be a key under `model_config`")
 
+    if cfg.warmup_steps and cfg.warmup_ratio:
+        raise ValueError("warmup_steps and warmup_ratio are mutually exclusive")
+
+    if cfg.is_qwen_derived_model and cfg.gradient_checkpointing:
+        LOG.warning(
+            "Gradient checkpointing is broken for Qwen models for transformers>=4.35.0, except main branch."
+        )
+
     # TODO
     # MPT 7b
     # https://github.com/facebookresearch/bitsandbytes/issues/25

src/axolotl/utils/data.py

@@ -79,6 +79,14 @@ def prepare_dataset(cfg, tokenizer):
         train_dataset, eval_dataset = process_datasets_for_packing(
             cfg, train_dataset, eval_dataset, tokenizer
         )
+
+    if eval_dataset and cfg.sample_packing and cfg.eval_sample_packing is not False:
+        total_eval_steps = calculate_total_num_steps(cfg, eval_dataset, update=False)
+        if total_eval_steps == 0:
+            raise ValueError(
+                "eval dataset split is too small for sample_packing. You should set `eval_sample_packing: False`. "
+            )
+
     if cfg.max_steps:
         total_num_steps = min(
             calculate_total_num_steps(cfg, train_dataset), cfg.max_steps
@@ -234,7 +242,14 @@ def load_tokenized_prepared_datasets(
             local_path = Path(config_dataset.path)
             if local_path.exists():
                 if local_path.is_dir():
-                    ds = load_from_disk(config_dataset.path)
+                    # TODO dirs with arrow or parquet files could be loaded with `load_from_disk`
+                    ds = load_dataset(
+                        config_dataset.path,
+                        name=config_dataset.name,
+                        data_files=config_dataset.data_files,
+                        streaming=False,
+                        split=None,
+                    )
                 elif local_path.is_file():
                     ds_type = get_ds_type(config_dataset)
 

src/axolotl/utils/models.py

@@ -84,6 +84,18 @@ def load_tokenizer(cfg):
     if cfg.is_mistral_derived_model and cfg.flash_attention and not cfg.sample_packing:
         tokenizer.padding_side = "left"
 
+    # Qwen base only has single token, so we need to set the special tokens
+    if cfg.is_qwen_derived_model:
+        token_ids = ["bos_token_id", "eos_token_id", "pad_token_id", "unk_token_id"]
+        for attr_name in token_ids:
+            if getattr(tokenizer, attr_name) is None:
+                setattr(tokenizer, attr_name, tokenizer.eod_id)
+
+        token_names = ["bos_token", "eos_token", "pad_token", "unk_token"]
+        for attr_name in token_names:
+            if getattr(tokenizer, attr_name) is None:
+                setattr(tokenizer, attr_name, "<|endoftext|>")
+
     if cfg.special_tokens:
         for k, val in cfg.special_tokens.items():
             tokenizer.add_special_tokens(
@@ -288,10 +300,10 @@ def load_model(
         #         device=cfg.device,
         #     )
         #     model.train() # sets to train instead of eval mode
-        elif model_type == "MixFormerSequentialForCausalLM":
-            from axolotl.models.phi import MixFormerSequentialForCausalLM
+        elif model_type == "PhiForCausalLM":
+            from axolotl.models.phi import PhiForCausalLM
 
-            model = MixFormerSequentialForCausalLM.from_pretrained(
+            model = PhiForCausalLM.from_pretrained(
                 base_model,
                 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,

src/axolotl/utils/samplers/multipack.py

@@ -182,7 +182,7 @@ class MultipackBatchSampler(BatchSampler):
 
         # shave off 1% + 1 for dealing with variance in packing from random sampler to sampler
         return max(
-            1,
+            0,
             (
                 world_size
                 * math.floor(

src/axolotl/utils/trainer.py

@@ -141,7 +141,7 @@ def process_datasets_for_packing(cfg, train_dataset, eval_dataset, tokenizer):
     return train_dataset, eval_dataset
 
 
-def calculate_total_num_steps(cfg, train_dataset):
+def calculate_total_num_steps(cfg, train_dataset, update=True):
     if not cfg.total_num_tokens:
         total_num_tokens = np.sum(
             train_dataset.data.column("input_ids")
@@ -150,7 +150,8 @@ def calculate_total_num_steps(cfg, train_dataset):
             .values
         )
         LOG.debug(f"total_num_tokens: {total_num_tokens}", main_process_only=True)
-        cfg.total_num_tokens = total_num_tokens
+        if update:
+            cfg.total_num_tokens = total_num_tokens
 
     if not cfg.total_supervised_tokens:
         total_supervised_tokens = (
@@ -163,7 +164,8 @@ def calculate_total_num_steps(cfg, train_dataset):
             f"`total_supervised_tokens: {total_supervised_tokens}`",
             main_process_only=True,
         )
-        cfg.total_supervised_tokens = total_supervised_tokens
+        if update:
+            cfg.total_supervised_tokens = total_supervised_tokens
 
     if cfg.sample_packing:
         # we have to drop anything longer then sequence len otherwise
@@ -232,7 +234,8 @@ def calculate_total_num_steps(cfg, train_dataset):
             sample_packing_eff_est = (
                 math.ceil(sample_packing_actual_eff_all * 100.0) / 100.0
             )
-            cfg.sample_packing_eff_est = sample_packing_eff_est
+            if update:
+                cfg.sample_packing_eff_est = sample_packing_eff_est
             LOG.debug(
                 f"sample_packing_eff_est: {cfg.sample_packing_eff_est}",
                 main_process_only=True,

tests/e2e/test_phi.py

@@ -31,7 +31,7 @@ class TestPhi(unittest.TestCase):
             {
                 "base_model": "microsoft/phi-1_5",
                 "trust_remote_code": True,
-                "model_type": "MixFormerSequentialForCausalLM",
+                "model_type": "PhiForCausalLM",
                 "tokenizer_type": "AutoTokenizer",
                 "sequence_len": 512,
                 "sample_packing": False,
@@ -76,7 +76,7 @@ class TestPhi(unittest.TestCase):
             {
                 "base_model": "microsoft/phi-1_5",
                 "trust_remote_code": True,
-                "model_type": "MixFormerSequentialForCausalLM",
+                "model_type": "PhiForCausalLM",
                 "tokenizer_type": "AutoTokenizer",
                 "sequence_len": 512,
                 "sample_packing": True,

tests/test_validation.py

@@ -649,3 +649,33 @@ class ValidationTest(unittest.TestCase):
         )
 
         validate_config(cfg)
+
+    def test_warmup_step_no_conflict(self):
+        cfg = DictDefault(
+            {
+                "warmup_steps": 10,
+                "warmup_ratio": 0.1,
+            }
+        )
+
+        with pytest.raises(
+            ValueError,
+            match=r".*warmup_steps and warmup_ratio are mutually exclusive*",
+        ):
+            validate_config(cfg)
+
+        cfg = DictDefault(
+            {
+                "warmup_steps": 10,
+            }
+        )
+
+        validate_config(cfg)
+
+        cfg = DictDefault(
+            {
+                "warmup_ratio": 0.1,
+            }
+        )
+
+        validate_config(cfg)