FlashML-org/FreeToken

FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently.

Python★ 15,546+242 that day#8 on trending

Analyzed November 2, 2026 at ee43b2a (main) with jscpd v5 in 2.2s.

C63.7/100

Project health185,754 lines of code (L)

Duplication58

6.7% in python, c-header, c, cpp, bash

Dead code94

0.6%

Complexity50

49.7%

The score is jscpd's own --health model: markup, data and text left out, tests kept. The numbers below are the trending scan's.

7.82%duplicated code · 10,928 of 139,719 lines

903clones in code

524code files of 749 scanned

6.63%duplicated over all files · 1,075 clones

Most complex files

Dead code 0.63% unused · 40 findings

python/freetoken/layers/linear_attn/gdn_reference.pyunused-file
python/freetoken/mm/processors/deepseek_v4.pyunused-file
python/freetoken/mm/processors/gemma4.pyunused-file
python/freetoken/mm/processors/qwen_vl.pyunused-file
python/freetoken/mm/processors/minimax_m3.pyunused-file
python/freetoken/mm/processors/deepseek_v41.pyunused-file
python/freetoken/mm/processors/glm5_next.pyunused-file
python/freetoken/mm/processors/muse_glimmer.pyunused-file
python/freetoken/models/deepseek_v4/image_processor.pyunused-file
python/freetoken/moe/fused.py:_torch_fused_topkunused-symbol

The code number leaves out tests, fixtures, docs, examples, data files, styles, translations and vendored code.

Largest clones in the code

9 lines in 21 placesc73 tokens

python/freetoken/kernel/csrc/ggml/ggml-cpu/arch/x86/quants.c:1830–1838python/freetoken/kernel/csrc/ggml/ggml-cpu/arch/x86/quants.c:2102–2110python/freetoken/kernel/csrc/ggml/ggml-cpu/arch/x86/quants.c:2280–2288python/freetoken/kernel/csrc/ggml/ggml-cpu/arch/x86/quants.c:2490–2498python/freetoken/kernel/csrc/ggml/ggml-cpu/arch/x86/quants.c:2724–2732and 16 more

1830void ggml_vec_dot_q3_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int …
1831    assert(n % QK_K == 0);
1832    assert(nrc == 1);
1833    UNUSED(nrc);
1834    UNUSED(bx);
1835    UNUSED(by);
1836    UNUSED(bs);
1837
1838    const uint32_t kmask1 = 0x03030303;

2102void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int …
2103    assert(n % QK_K == 0);
2104    assert(nrc == 1);
2105    UNUSED(nrc);
2106    UNUSED(bx);
2107    UNUSED(by);
2108    UNUSED(bs);
2109
2110    const block_q4_K * GGML_RESTRICT x = vx;

69 lines in 3 placespython497 tokens

python/freetoken/models/llama/attention.py:17–85python/freetoken/models/mistral/attention.py:17–85python/freetoken/models/qwen2/attention.py:17–85

17class LlamaAttention(BaseOP):
18    def __init__(
19        self,
20        config: ModelConfig,
21        layer_id: int,
22        *,
23        has_attn_bias: bool = False,
24        has_qk_norm: bool = False,
25        prefix: str = "",
26    ):

First 10 of 69 lines.

59 lines in 4 placespython323 tokens

python/freetoken/models/llama/weight.py:1–59python/freetoken/models/mistral/weight.py:1–59python/freetoken/models/qwen2/weight.py:1–59python/freetoken/models/qwen3/weight.py:1–59

1from __future__ import annotations
2
3from typing import Iterator
4
5import safetensors
6import torch
7from freetoken.distributed import get_tp_info
8from freetoken.models.loader import MergeRule, iter_merged_tensors, iter_weight_files, shard_tensor
9from freetoken.utils import cached_load_hf_config
10from tqdm import tqdm

First 10 of 59 lines.

7 more

69 lines in 2 placesc758 tokens
python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:3547–3615 python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:5177–5245
43 lines in 4 placespython230 tokens
python/freetoken/models/llama/config.py:1–43 python/freetoken/models/mistral/config.py:1–43 python/freetoken/models/qwen2/config.py:1–43 python/freetoken/models/qwen3/config.py:1–43
38 lines in 2 placesc578 tokens
python/freetoken/kernel/csrc/ggml/ggml-cpu/quants.c:734–771 python/freetoken/kernel/csrc/ggml/ggml-cpu/quants.c:813–851
71 lines in 2 placesc-header555 tokens
python/freetoken/kernel/csrc/ggml/ggml-cpu/simd-mappings.h:269–339 python/freetoken/kernel/csrc/ggml/ggml-cpu/simd-mappings.h:384–454
9 lines in 11 placespython52 tokens
python/freetoken/models/gemma4/model.py:84–92 python/freetoken/models/glm4_moe/model.py:75–83 python/freetoken/models/glm5_next/model.py:164–172 python/freetoken/models/glm_moe_dsa/model.py:70–78 python/freetoken/models/llama/model.py:75–83and 6 more
43 lines in 3 placespython255 tokens
python/freetoken/models/llama/attention.py:17–59 python/freetoken/models/qwen3/attention.py:17–59 python/freetoken/models/qwen3_moe/attention.py:17–59
54 lines in 2 placesc503 tokens
python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:1885–1938 python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:1999–2053

Largest clones over all files, tests and data included

9 lines in 21 placesc73 tokens

1830void ggml_vec_dot_q3_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int …
1831    assert(n % QK_K == 0);
1832    assert(nrc == 1);
1833    UNUSED(nrc);
1834    UNUSED(bx);
1835    UNUSED(by);
1836    UNUSED(bs);
1837
1838    const uint32_t kmask1 = 0x03030303;

2102void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int …
2103    assert(n % QK_K == 0);
2104    assert(nrc == 1);
2105    UNUSED(nrc);
2106    UNUSED(bx);
2107    UNUSED(by);
2108    UNUSED(bs);
2109
2110    const block_q4_K * GGML_RESTRICT x = vx;

69 lines in 3 placespython497 tokens

python/freetoken/models/llama/attention.py:17–85python/freetoken/models/mistral/attention.py:17–85python/freetoken/models/qwen2/attention.py:17–85

17class LlamaAttention(BaseOP):
18    def __init__(
19        self,
20        config: ModelConfig,
21        layer_id: int,
22        *,
23        has_attn_bias: bool = False,
24        has_qk_norm: bool = False,
25        prefix: str = "",
26    ):

First 10 of 69 lines.

59 lines in 4 placespython323 tokens

python/freetoken/models/llama/weight.py:1–59python/freetoken/models/mistral/weight.py:1–59python/freetoken/models/qwen2/weight.py:1–59python/freetoken/models/qwen3/weight.py:1–59

1from __future__ import annotations
2
3from typing import Iterator
4
5import safetensors
6import torch
7from freetoken.distributed import get_tp_info
8from freetoken.models.loader import MergeRule, iter_merged_tensors, iter_weight_files, shard_tensor
9from freetoken.utils import cached_load_hf_config
10from tqdm import tqdm

First 10 of 59 lines.

69 lines in 2 placesc758 tokens
python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:3547–3615 python/freetoken/kernel/csrc/ggml/src/ggml-quants.c:5177–5245
43 lines in 4 placespython230 tokens
python/freetoken/models/llama/config.py:1–43 python/freetoken/models/mistral/config.py:1–43 python/freetoken/models/qwen2/config.py:1–43 python/freetoken/models/qwen3/config.py:1–43

By format, code only

Format	Files	Lines	Clones	Duplicated lines	Duplication
python	471	109,627	542	6,816	6.22%
c	4	11,322	298	3,311	29.24%
c-header	27	13,078	46	668	5.11%
cpp	17	4,659	17	133	2.85%
bash	5	1,033	0	0	0%

By format, all files

Format	Files	Lines	Clones	Duplicated lines	Duplication
python	643	155,092	697	8,320	5.36%
c	4	11,322	298	3,311	29.24%
c-header	30	13,580	46	668	4.92%
cpp	18	4,734	17	133	2.81%
diff	1	1,400	11	93	6.64%
yaml	9	804	4	60	7.46%
markup	4	47	2	21	44.68%
markdown	18	1,718	0	0	0%
bash	14	1,138	0	0	0%
toml	2	167	0	0	0%

Trending appearances

Day	Rank	Stars	Files	Lines	Clones	All files	Code	Health	Commit
Nov 2, 2026	#8	15,546 (+242)	749	190,196	1,075	6.63%	7.82%	C 63.7	`ee43b2a`

Want this for your own project? Install jscpd and run jscpd ., or measure it the way the baselines were measured.