112 lines
3.4 KiB
Python
112 lines
3.4 KiB
Python
"""
|
|
E2E tests for llama w/ S2 attn
|
|
"""
|
|
|
|
import unittest
|
|
|
|
import pytest
|
|
|
|
from axolotl.common.datasets import load_datasets
|
|
from axolotl.train import train
|
|
from axolotl.utils.config import normalize_config, validate_config
|
|
from axolotl.utils.dict import DictDefault
|
|
|
|
from ..utils import check_model_output_exists, with_temp_dir
|
|
|
|
|
|
@pytest.mark.skip(reason="FIXME?")
|
|
class TestLlamaShiftedSparseAttention(unittest.TestCase):
|
|
"""
|
|
Test case for Llama models using S2 Attn
|
|
"""
|
|
|
|
@with_temp_dir
|
|
def test_lora_s2_attn(self, temp_dir):
|
|
cfg = DictDefault(
|
|
{
|
|
"base_model": "HuggingFaceTB/SmolLM2-135M",
|
|
"tokenizer_type": "AutoTokenizer",
|
|
"sequence_len": 16384,
|
|
"sample_packing": False,
|
|
"flash_attention": True,
|
|
"s2_attention": True,
|
|
"load_in_8bit": True,
|
|
"adapter": "lora",
|
|
"lora_r": 32,
|
|
"lora_alpha": 16,
|
|
"lora_dropout": 0.05,
|
|
"lora_target_linear": True,
|
|
"val_set_size": 0.02,
|
|
"special_tokens": {
|
|
"pad_token": "<|endoftext|>",
|
|
},
|
|
"datasets": [
|
|
{
|
|
"path": "Yukang/LongAlpaca-12k",
|
|
"type": "alpaca",
|
|
},
|
|
],
|
|
"num_epochs": 2,
|
|
"micro_batch_size": 1,
|
|
"gradient_accumulation_steps": 1,
|
|
"output_dir": temp_dir,
|
|
"learning_rate": 0.00001,
|
|
"optimizer": "adamw_torch_fused",
|
|
"lr_scheduler": "cosine",
|
|
"max_steps": 10,
|
|
"save_steps": 5,
|
|
"eval_steps": 5,
|
|
"bf16": "auto",
|
|
"save_first_step": False,
|
|
}
|
|
)
|
|
|
|
cfg = validate_config(cfg)
|
|
normalize_config(cfg)
|
|
dataset_meta = load_datasets(cfg=cfg)
|
|
|
|
train(cfg=cfg, dataset_meta=dataset_meta)
|
|
check_model_output_exists(temp_dir, cfg)
|
|
|
|
@with_temp_dir
|
|
def test_fft_s2_attn(self, temp_dir):
|
|
cfg = DictDefault(
|
|
{
|
|
"base_model": "HuggingFaceTB/SmolLM2-135M",
|
|
"tokenizer_type": "AutoTokenizer",
|
|
"sequence_len": 16384,
|
|
"sample_packing": False,
|
|
"flash_attention": True,
|
|
"s2_attention": True,
|
|
"val_set_size": 0.02,
|
|
"special_tokens": {
|
|
"pad_token": "<|endoftext|>",
|
|
},
|
|
"datasets": [
|
|
{
|
|
"path": "Yukang/LongAlpaca-12k",
|
|
"type": "alpaca",
|
|
},
|
|
],
|
|
"num_epochs": 2,
|
|
"micro_batch_size": 1,
|
|
"gradient_accumulation_steps": 1,
|
|
"output_dir": temp_dir,
|
|
"learning_rate": 0.00001,
|
|
"optimizer": "adamw_torch_fused",
|
|
"lr_scheduler": "cosine",
|
|
"max_steps": 10,
|
|
"save_steps": 5,
|
|
"eval_steps": 5,
|
|
"bf16": "auto",
|
|
"save_first_step": False,
|
|
}
|
|
)
|
|
|
|
cfg = validate_config(cfg)
|
|
normalize_config(cfg)
|
|
dataset_meta = load_datasets(cfg=cfg)
|
|
|
|
train(cfg=cfg, dataset_meta=dataset_meta)
|
|
check_model_output_exists(temp_dir, cfg)
|