tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

attempt to also run e2e tests that needs gpus (#1070)

Browse Source 
* attempt to also run e2e tests that needs gpus

* fix stray quote

* checkout specific github ref

* dockerfile for tests with proper checkout

ensure wandb is dissabled for docker pytests
clear wandb env after testing
clear wandb env after testing
make sure to provide a default val for pop
tryin skipping wandb validation tests
explicitly disable wandb in the e2e tests
explicitly report_to None to see if that fixes the docker e2e tests
split gpu from non-gpu unit tests
skip bf16 check in test for now
build docker w/o cache since it uses branch name ref
revert some changes now that caching is fixed
skip bf16 check if on gpu w support

* pytest skip for auto-gptq requirements

* skip mamba tests for now, split multipack and non packed lora llama tests

* split tests that use monkeypatches

* fix relative import for prev commit

* move other tests using monkeypatches to the correct run
This commit is contained in:
Wing Lian
2024-01-09 21:23:23 -05:00
committed by GitHub
parent 9be92d1448
commit 788649fe95
13 changed files with 214 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
tests/e2e/test_lora_llama.py
View File

@@ -65,96 +65,3 @@ class TestLoraLlama(unittest.TestCase):
train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
assert (Path(temp_dir) / "adapter_model.bin").exists()
@with_temp_dir
def test_lora_packing(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "JackFram/llama-68m",
"tokenizer_type": "LlamaTokenizer",
"sequence_len": 1024,
"sample_packing": True,
"flash_attention": True,
"load_in_8bit": True,
"adapter": "lora",
"lora_r": 32,
"lora_alpha": 64,
"lora_dropout": 0.05,
"lora_target_linear": True,
"val_set_size": 0.1,
"special_tokens": {
"unk_token": "<unk>",
"bos_token": "<s>",
"eos_token": "</s>",
},
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"type": "alpaca",
},
],
"num_epochs": 2,
"micro_batch_size": 8,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"lr_scheduler": "cosine",
"bf16": True,
}
)
normalize_config(cfg)
cli_args = TrainerCliArgs()
dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
assert (Path(temp_dir) / "adapter_model.bin").exists()
@with_temp_dir
def test_lora_gptq(self, temp_dir):
# pylint: disable=duplicate-code
cfg = DictDefault(
{
"base_model": "TheBlokeAI/jackfram_llama-68m-GPTQ",
"model_type": "AutoModelForCausalLM",
"tokenizer_type": "LlamaTokenizer",
"sequence_len": 1024,
"sample_packing": True,
"flash_attention": True,
"load_in_8bit": True,
"adapter": "lora",
"gptq": True,
"gptq_disable_exllama": True,
"lora_r": 32,
"lora_alpha": 64,
"lora_dropout": 0.05,
"lora_target_linear": True,
"val_set_size": 0.1,
"special_tokens": {
"unk_token": "<unk>",
"bos_token": "<s>",
"eos_token": "</s>",
},
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"type": "alpaca",
},
],
"num_epochs": 2,
"save_steps": 0.5,
"micro_batch_size": 8,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"lr_scheduler": "cosine",
}
)
normalize_config(cfg)
cli_args = TrainerCliArgs()
dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
assert (Path(temp_dir) / "adapter_model.bin").exists()