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

transformers v5 upgrade (#3272)

Browse Source 
* Prepare for transformers v5 upgrade

* fix hf cli

* update for hf hub changes

* fix tokenizer apply_chat_template args

* remap include_tokens_per_second

* fix tps

* handle migration for warmup

* use latest hf hub

* Fix scan -> ls

* fix import

* fix for renaming of mistral common tokenizer -> backend

* update for fixed tokenziation for llama

* Skip phi35 tests for now

* remove mistral patch fixed upstream in huggingface/transformers#41439

* use namespacing for patch

* don't rely on sdist for e2e tests for now

* run modal ci without waiting too

* Fix dep for ci

* fix imports

* Fix fp8 check

* fsdp2 fixes

* fix version handling

* update fsdp version tests for new v5 behavior

* Fail multigpu tests after 3 failures

* skip known v5 broken tests for now and cleanup

* bump deps

* unmark skipped test

* re-enable test_fsdp_qlora_prequant_packed test

* increase multigpu ci timeout

* skip broken gemma3 test

* reduce timout back to original 120min now that the hanging test is skipped

* fix for un-necessary collator for pretraining with bsz=1

* fix: safe_serialization deprecated in transformers v5 rc01 (#3318)

* torch_dtype deprecated

* load model in float32 for consistency with tests

* revert some test fixtures back

* use hf cache ls instead of scan

* don't strip fsdp_version

more fdsp_Version fixes for v5
fix version in fsdp_config
fix aliasing
fix fsdp_version check
check fsdp_version is 2 in both places

* Transformers v5 rc2 (#3347)

* bump dep

* use latest fbgemm, grab model config as part of fixture, un-skip test

* import AutoConfig

* don't need more problematic autoconfig when specifying config.json manually

* add fixtures for argilla ultrafeedback datasets

* download phi4-reasoning

* fix arg

* update tests for phi fast tokenizer changes

* use explicit model types for gemma3

---------

Co-authored-by: Wing Lian <wing@axolotl.ai>

* fix: AutoModelForVision2Seq -> AutoModelForImageTextToText

* chore: remove duplicate

* fix: attempt fix gemma3 text mode

* chore: lint

* ga release of v5

* need property setter for name_or_path for mistral tokenizer

* vllm not compatible with transformers v5

* setter for chat_template w mistral too

---------

Co-authored-by: NanoCode012 <nano@axolotl.ai>
Co-authored-by: salman <salman.mohammadi@outlook.com>
This commit is contained in:
Wing Lian
2026-01-27 17:08:24 -05:00
committed by GitHub
parent a531e9d946
commit fc4e37920b
74 changed files with 262 additions and 309 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/e2e/multigpu/solo/test_grpo.py
View File

@@ -220,7 +220,6 @@ def oai_gsm8k_transform(cfg, *args, **kwargs):
"learning_rate": 0.0001,
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"save_safetensors": True,
"bf16": "auto",
"use_tensorboard": True,
"save_first_step": False,
@@ -315,7 +314,6 @@ def oai_gsm8k_transform(cfg, *args, **kwargs):
"learning_rate": 0.0001,
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"save_safetensors": True,
"bf16": "auto",
"use_tensorboard": True,
"save_first_step": False,
@@ -408,7 +406,6 @@ def oai_gsm8k_transform(cfg, *args, **kwargs):
"learning_rate": 0.0001,
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"save_safetensors": True,
"bf16": "auto",
"use_tensorboard": True,
"save_first_step": False,

5
tests/e2e/multigpu/test_fp8_fsdp2.py
View File

@@ -11,7 +11,7 @@ from transformers.testing_utils import get_torch_dist_unique_port
from axolotl.utils.dict import DictDefault
from tests.e2e.utils import most_recent_subdir, require_hopper, require_torch_2_7_0
from tests.e2e.utils import most_recent_subdir, require_torch_2_7_0, supports_fp8
AXOLOTL_ROOT = Path(__file__).parent.parent.parent.parent
@@ -49,7 +49,7 @@ class TestFP8FSDP2:
"""Test class for FP8 mixed precision with FSDP2 functionality."""
@require_torch_2_7_0
@require_hopper
@supports_fp8
def test_fp8_fsdp2_smoke(self, temp_dir):
"""Smoke test for 2-GPU FP8 + torch.compile + FSDP2 training"""
cfg = DictDefault(
@@ -94,7 +94,6 @@ class TestFP8FSDP2:
"reshard_after_forward": True,
},
"use_tensorboard": True,
"save_safetensors": True,
"save_first_step": False,
}
)

1
tests/e2e/multigpu/test_fsdp1.py
View File

@@ -244,6 +244,7 @@ class TestFSDP1:
verify_training_success(temp_dir)
@pytest.mark.skip("broken in transformers v5")
@pytest.mark.parametrize(
"adapter_config",
[

4
tests/e2e/multigpu/test_fsdp2.py
View File

@@ -150,6 +150,10 @@ class TestFSDP2:
},
"use_tensorboard": True,
"bf16": True,
# explicitly disable LORA kernels, as they may be auto-enabled
"lora_mlp_kernel": False,
"lora_qkv_kernel": False,
"lora_o_kernel": False,
}
)

2
tests/e2e/multigpu/test_gemma3.py
View File

@@ -23,6 +23,7 @@ def download_model():
snapshot_download("axolotl-mirrors/gemma-3-4b-pt", repo_type="model")
@pytest.mark.skip(reason="FIXME")
class TestMultiGPUGemma3:
"""
Test case for Gemma3 models using LoRA
@@ -32,6 +33,7 @@ class TestMultiGPUGemma3:
cfg = DictDefault(
{
"base_model": "axolotl-mirrors/gemma-3-4b-pt",
"unfrozen_parameters": ["model.language_model.*", "lm_head"],
"sequence_len": 2048,
"ddp_find_unused_parameters": True,
"sample_packing": True,

1
tests/e2e/multigpu/test_llama.py
View File

@@ -901,7 +901,6 @@ class TestMultiGPULlama:
"flash_attention": True,
"sample_packing": True,
"bf16": True,
"save_safetensors": True,
# "deepspeed": str(AXOLOTL_ROOT / "deepspeed_configs/zero1.json"),
"use_tensorboard": True,
"save_first_step": False,