From 20adb4465ab9f698e2f0910433d292a04fccf2d5 Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Fri, 19 Dec 2025 19:35:41 -0800 Subject: [PATCH] Nightly (#3753) * Update _utils.py * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [FIX] [Transformers] VLM input embeds fix for gradients (#3715) * Fix get_input_embeds call for VLMs * patch input_require_grads instead * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * cleanup old patch * cleanup old patch * cleanup * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Apply suggestion from @danielhanchen * use logger instead of prints * Move unsloth present set * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Daniel Han * Update rope_embedding.py * Fixes * Update _utils.py * Update import_fixes.py * Update rl_replacements.py * fix_openenv_no_vllm * Fix * Update __init__.py * Update __init__.py * Update __init__.py * Update import_fixes.py * Update import_fixes.py * Update import_fixes.py * logger * Update __init__.py * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update __init__.py * Update import_fixes.py * Update __init__.py * Update import_fixes.py * Update import_fixes.py * Update import_fixes.py * Update import_fixes.py * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update import_fixes.py * Update unsloth/import_fixes.py Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Update save.py * [fbgemm] Silence tma fbgemm (#3735) * Silence fbgemm TMA print Also safer .push_to_hub * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update loader.py * Update save.py * Update save.py * Update _utils.py * Update _utils.py * Diffusers warnings * Update pyproject.toml * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [hf_hub] Token login (#3739) * login on token * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * cleanup old code * safer imports * cleanup * Return token after login * correct return types * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Apply suggestion from @danielhanchen * add back imports * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * finish return token --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Daniel Han * Do not overwrite slots (#3752) * Do not overwrite slots * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Datta Nimmaturi Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> --- unsloth/import_fixes.py | 2 -- unsloth/models/_utils.py | 21 +++++++++++++++++++++ unsloth/models/llama.py | 3 +-- unsloth/models/loader.py | 23 +++-------------------- unsloth/models/vision.py | 3 +-- 5 files changed, 26 insertions(+), 26 deletions(-) diff --git a/unsloth/import_fixes.py b/unsloth/import_fixes.py index f3aae7f523..efc7a7f4cd 100644 --- a/unsloth/import_fixes.py +++ b/unsloth/import_fixes.py @@ -72,8 +72,6 @@ class HideLoggingMessage(logging.Filter): class HidePrintMessage: - __slots__ = ("_original_stream", "_hidden_texts") - def __init__(self, original_stream): self._original_stream = original_stream self._hidden_texts = [] diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py index d769f815bd..9929277719 100644 --- a/unsloth/models/_utils.py +++ b/unsloth/models/_utils.py @@ -72,6 +72,7 @@ __all__ = [ "patch_hf_quantizer", "verify_fp8_support_if_applicable", "_get_inference_mode_context_manager", + "hf_login", ] import torch @@ -2344,3 +2345,23 @@ def _get_inference_mode_context_manager(model: torch.nn.Module): return torch.no_grad() else: return torch.inference_mode() + + +def hf_login(token: Optional[str] = None) -> Optional[str]: + if token is None: + try: + from huggingface_hub import get_token + + token = get_token() + if token is None: + return None + except: + return None + try: + from huggingface_hub import login + + login(token = token) + return token + except Exception as e: + logger.info(f"Failed to login to huggingface using token with error: {e}") + return token diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 4c9337ccf9..1d7695b9aa 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -2130,8 +2130,7 @@ class FastLlamaModel: "Unsloth: `unsloth_vllm_standby` is True, but environment variable `UNSLOTH_VLLM_STANDBY` is not set to 1!" ) - if token is None: - token = get_token() + token = hf_login(token) if model_patcher is None: model_patcher = FastLlamaModel SUPPORTS_BFLOAT16 = is_bfloat16_supported() diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py index a1e5756060..fdfc313732 100644 --- a/unsloth/models/loader.py +++ b/unsloth/models/loader.py @@ -20,6 +20,7 @@ from ._utils import ( HAS_FLASH_ATTENTION_SOFTCAPPING, USE_MODELSCOPE, get_transformers_model_type, + hf_login, ) from .granite import FastGraniteModel from .llama import FastLlamaModel, logger @@ -151,15 +152,7 @@ class FastLanguageModel(FastLlamaModel): **kwargs, ): # Login to allow private models - if token is None: - token = get_token() - if token is not None: - try: - from huggingface_hub import login - - login(token = token) - except: - pass + token = hf_login(token) if load_in_8bit or full_finetuning or qat_scheme is not None: return FastModel.from_pretrained( model_name = model_name, @@ -195,8 +188,6 @@ class FastLanguageModel(FastLlamaModel): **kwargs, ) - if token is None: - token = get_token() if isinstance(dtype, str) and dtype in ["float16", "bfloat16"]: dtype = getattr(torch, dtype) assert ( @@ -687,16 +678,8 @@ class FastModel(FastBaseModel): *args, **kwargs, ): - if token is None: - token = get_token() # Login to allow private models - if token is not None: - try: - from huggingface_hub import login - - login(token = token) - except: - pass + token = hf_login(token) if whisper_language is not None: assert type(whisper_language) is str if whisper_task is not None: diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py index 138b1b633a..b78b190bcb 100644 --- a/unsloth/models/vision.py +++ b/unsloth/models/vision.py @@ -397,8 +397,7 @@ class FastBaseModel: "Unsloth: WARNING `trust_remote_code` is True.\n" "Are you certain you want to do remote code execution?" ) - if token is None: - token = get_token() + token = hf_login(token) SUPPORTS_BFLOAT16 = is_bfloat16_supported() if DEVICE_TYPE == "cuda":