Benign huggingface Download

xtqzcm/hf-transfer-20261014t012336-bce9694a11b9 main

Long base64 string literalMegabyte-scale size constant in source
SHA-256ccbcdb19184319b485fe7b513df1e824f0ac2016c914e603a1064776d0c77405

Evidence

imports deepcopy from copy lines 1–12
1from __future__ import annotations
2
3from copy import deepcopy
4from typing import Any
5
6from transformers import Qwen3Config
7
8
⋯4 lines
Megabyte-scale size constant in source lines 50–56
50:34… xels: int = 65536,
51 simplear_qwen3_vit_max_pixels: int = 1048576,
52 **kwargs,
53 ):
54 if not 0.0 <= simplear_vision_feature_noise_alpha <= 1.0:
55 raise ValueError(
56 …
imports math lines 99–102
99:22… {simplear_ar_head_attention!r}."
100 )
101 import math
102 if not math.isfinite(simplear_feature_ar_loss_weight) or simplear_feature_ar_loss_weigh …
imports .alignment_options.ALIGNMENT_DEFAULTS from .alignment_options lines 115–122
115:41… ls = simplear_qwen3_vit_max_pixels
116 from .clip_pal import PAL_DEFAULTS
117 for key, default in PAL_DEFAULTS.items():
118 setattr(self, key, kwargs.get(key, default))
119 from .alignment_options import ALIGNMENT_DEFAULTS, resolve_alignment_type
120 for key, default in ALIGNMENT_DEFAULTS.items():
121 setattr(self, key, kwargs.get(key, default))
122 …
Python module-level function call lines 124–128
124:27… e = resolve_alignment_type(self)
125 self.architectures = ["SimpleARWithARHeadForCausalLM"]
126
127
128SimpleARWithARHeadConfig.register_for_auto_class()

Keyboard shortcuts on this page: j for the next sample, k for the previous one, x to go back to the feed, d to download the original bytes, r to re-queue the sample for analysis.