File as GitHub Issue repo: linkedin/Liger-Kernel

Push this scan report to `linkedin/Liger-Kernel`

Click the green button below to open GitHub’s new-issue form, pre-filled with the report title, summary table, top findings, and an embedded score-card image. No authentication needed — you review on GitHub before submitting. Repobility is credited as the scanner.

Embedded score card image

This image will render at the top of the issue body. Hosted on Repobility, refreshes automatically after re-scans.

Issue title

Phantom test coverage — test files without real assertions

Curate findings to include

Pick exactly which findings appear in the issue body. By default the top 5 are included. Uncheck noise, check what matters.

Top 5 (default)

Severity	Rule	Title	File:line
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_tvd.py:185`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_tvd.py:164`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_tvd.py:155`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:315`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:310`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:296`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:288`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:279`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_jsd.py:272`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_geglu.py:49`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_cro…:558`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_dyt.py:127`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_dyt.py:60`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_sparsemax.py:87`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_sparsemax.py:45`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_poly_norm.py:103`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_softmax.py:88`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_softmax.py:42`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_multi_token_atte…:286`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_multi_token_atte…:221`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_multi_token_atte…:136`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_multi_token_atte…:66`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_add_rms_no…:126`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_rms_norm.py:304`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_rms_norm.py:150`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_jsd…:384`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_jsd…:330`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_jsd…:257`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_jsd…:180`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_linear_jsd…:106`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_flex_attention.py:289`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_flex_attention.py:105`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_neighborho…:496`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_neighborho…:455`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_neighborho…:244`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_fused_neighborho…:112`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:879`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:844`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:784`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:720`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:686`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:652`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:626`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:600`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:574`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_cross_entropy.py:554`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:1232`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:1162`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:896`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:409`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:319`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:248`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_grpo_loss.py:213`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_mhc.py:479`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_mhc.py:321`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_mhc.py:252`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_mhc.py:207`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_mhc.py:91`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_kl_div.py:91`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_kl_div.py:82`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/transformers/test_swiglu.py:485`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/fp32/test_mini_models_…:1782`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/fp32/test_mini_models_…:1624`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/fp32/test_mini_models.…:2058`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/fp32/test_mini_models_…:1926`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/bf16/test_mini_models.…:2219`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/bf16/test_mini_models_…:1676`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/bf16/test_mini_models_…:1517`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/convergence/bf16/test_mini_models_…:2059`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_dpo_loss.py:842`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_dpo_loss.py:602`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_grpo_loss.py:832`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_grpo_loss.py:808`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_grpo_loss.py:543`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_cpo_loss.py:265`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_jsd_loss.py:366`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_jsd_loss.py:247`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_simpo_loss.py:178`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_cosine_loss.py:258`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_cosine_loss.py:167`
HIGH	`TEST001`	Phantom test coverage — test files without real assertions	`test/chunked_loss/test_orpo_loss.py:226`
HIGH	`ERRH001`	Bare except: pass — silent failure	`src/liger_kernel/ops/backends/_ascend/u…:172`
HIGH	`SEC013`	[SEC013] Path Traversal — User Input in File Path: User-controlled input used in file pat…	`src/liger_kernel/transformers/model/qwe…:95`
HIGH	`SEC013`	[SEC013] Path Traversal — User Input in File Path: User-controlled input used in file pat…	`src/liger_kernel/transformers/model/qwe…:92`
HIGH	`SEC013`	[SEC013] Path Traversal — User Input in File Path: User-controlled input used in file pat…	`src/liger_kernel/transformers/model/lla…:63`
MED	`ERR001`	[ERR001] Silent Exception Swallowing: Silently swallowing all exceptions hides bugs. Even…	`src/liger_kernel/ops/backends/_ascend/u…:172`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:4`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:352`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:285`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:875`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:207`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/ops/backends/_ascend/o…:1`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/simpo_los…:72`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/simpo_los…:35`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/orpo_loss…:63`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/jsd_loss.…:43`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/grpo_loss…:193`
MED	`AIC003`	Duplicated implementation block across source files	`src/liger_kernel/chunked_loss/fused_lin…:6`
MED	`QUAL002`	Python mutable default argument	`benchmark/scripts/utils.py:260`
MED	`TEST002`	Function is stub-only (pass/raise NotImplementedError)	`test/transformers/test_monkey_patch.py:363`
MED	`TEST002`	Function is stub-only (pass/raise NotImplementedError)	`test/transformers/test_monkey_patch.py:293`
MED	`TEST002`	Function is stub-only (pass/raise NotImplementedError)	`test/utils.py:829`
MED	`ERRH002`	Bare except — overly broad	`/tank0/claude-archive/community/linkedi…:16`
MED	`LOG001`	PII printed to stdout/stderr	`src/liger_kernel/transformers/model/int…:70`
MED	`LOG001`	PII printed to stdout/stderr	`benchmark/scripts/benchmark_grpo_loss.py:203`
MED	`LOG001`	PII printed to stdout/stderr	`examples/huggingface/launch_on_modal.py:39`
MED	`LOG001`	PII printed to stdout/stderr	`examples/medusa/train.py:308`
MED	`LOG001`	PII printed to stdout/stderr	`examples/medusa/train.py:307`
MED	`SUPC002`	Supply chain — npm install without lockfile	`examples/huggingface/launch_on_modal.py:11`
MED	`CRYP001`	Crypto — plaintext HTTP for sensitive endpoint	`src/liger_kernel/chunked_loss/grpo_loss…:10`
MED	`SEC005`	[SEC005] Command Injection Risk: Unsafe shell execution or eval of user input.	`dev/modal/tests.py:23`
MED	`SEC005`	[SEC005] Command Injection Risk: Unsafe shell execution or eval of user input.	`dev/modal/benchmarks.py:22`
MED	`CRYP001`	Crypto — plaintext HTTP for sensitive endpoint	`examples/medusa/train.py:9`
LOW	`QUAL005`	Cluster of TODOs in one file	`src/liger_kernel/ops/cross_entropy.py:287`
LOW	`QUAL005`	Cluster of TODOs in one file	`src/liger_kernel/ops/cross_entropy.py:236`
LOW	`AIC005`	Duplicate top-level symbol appears in a patch-style file	`src/liger_kernel/transformers/model/hun…:1`
LOW	`AIC002`	Source file name looks like an AI patch artifact	`src/liger_kernel/transformers/model/hun…:1`

Reset to top 5 117 findings available (after auto-suppression of test files + won't-fix)

Issue body (markdown)

## Code-quality scan: `linkedin/Liger-Kernel`

**Score: 86/100 (B+)**  ·  124 findings  ·  scanned 2026-05-15 18:26 UTC  ·  86,760 LOC

| Severity | Count |
|---|---|
| CRITICAL | 0 |
| HIGH | 85 |
| MEDIUM | 28 |
| LOW | 4 |

📊 [Full filterable report](https://repobility.com/scan/fedca033-f0ab-4858-8e75-ec394950c9b6/)  ·  ![scorecard](https://repobility.com/scan/fedca033-f0ab-4858-8e75-ec394950c9b6/report.png?v=1778869607-s2)

### Top findings

1. **HIGH** `TEST001` — Phantom test coverage — test files without real assertions
   `test/transformers/test_tvd.py:185`
2. **HIGH** `TEST001` — Phantom test coverage — test files without real assertions
   `test/transformers/test_tvd.py:164`
3. **HIGH** `TEST001` — Phantom test coverage — test files without real assertions
   `test/transformers/test_tvd.py:155`
4. **HIGH** `TEST001` — Phantom test coverage — test files without real assertions
   `test/transformers/test_jsd.py:315`
5. **HIGH** `TEST001` — Phantom test coverage — test files without real assertions
   `test/transformers/test_jsd.py:310`

---

_Filed automatically. Close this issue if not useful — we won't refile. Full report: https://repobility.com/scan/fedca033-f0ab-4858-8e75-ec394950c9b6/_

Megaproject â high spam risk

Could not determine 'linkedin/Liger-Kernel' star count (GitHub API rate-limited or unreachable). When in doubt about repo size, prefer opening a focused PR or a discussion rather than an issue.

Already filed

95/124 findings (77%) on this scan are already flagged as test-file, won't-fix, or suppressed. The scan is too noisy to file as a single issue. Curate down to specific actionable findings, or address the FP source first.

Open in GitHub to file this issue Filing guard overridden Cancel

The button opens GitHubâs new-issue page in a new tab. You will see the title + body pre-filled â review, edit if you want, then click GitHubâs "Submit new issue" button. Repobility never posts anything on your behalf.

For real security findings on big repos: use the project's SECURITY.md or private advisory flow instead of a public issue.

Push this scan report to linkedin/Liger-Kernel

Embedded score card image

Issue title

Curate findings to include

Issue body (markdown)

Push this scan report to `linkedin/Liger-Kernel`