Benign huggingface Download

hanseungwook/Llama-3.2-1B-Compressed-Joint-GSM8K-TrimAnswer main

References an answer or ground-truth pathPython getattr dynamic attribute lookup
SHA-256c40c9f216dd3258c12ee49a482d456e7d5e0be48602413bcbdb6f93960cd06cd
MaleculeH₂(DbOs)

Evidence

English function-word token "and" lines 1–16
1:4… Shared greedy GSM8K evaluation used by training and final testing."""
2
3from __future__ import annotations
⋯13 lines
References an answer or ground-truth path lines 16–27
16:45…
17 """Extract GSM8K's final numeric answer from its annotated solution."""
18
19 final = answer_text.split("####")[-1].strip()
⋯8 lines
English function-word token "the" lines 35–38
35:3… """Generate answer-only continuations through the Random Joint latent path."""
36
37 texts, _, _ = greedy_decode_with_metadata(
38 model=mode …
English function-word token "for" lines 82–89
82:71… ice=device)
83 generated: list[list[int]] = [[] for _ in questions]
84 raw_eos = model.backbone.config.eos_token_id
85 eos_token_ids = {
⋯4 lines
Python getattr dynamic attribute lookup lines 287–294
287:29… (metrics)
288 else:
289 metric_values = {
290 name: getattr(metrics, name)
291 for name in (
292 "accuracy",
293 "balanced_accuracy",
294 "valid_count" …

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.