Hugging Face's logo

RESMP-DEV
/
LFM2.5-Encoder-230M-Code-BF16

Feature Extraction
Transformers
Safetensors
PyTorch
lfm2
code
embeddings
bf16
custom_code
Model card Files
xet
 Community
LFM2.5-Encoder-230M-Code-BF16
File size: 5,010 Bytes
3296c13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b23061b
 
 
3296c13
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
---
license: other
license_name: lfm1.0
license_link: LICENSE
base_model: LiquidAI/LFM2.5-Encoder-230M
pipeline_tag: feature-extraction
library_name: transformers
tags:
- code
- embeddings
- feature-extraction
- transformers
- pytorch
- bf16
---

# LFM2.5 Encoder 230M Code BF16

This is a modified RESMP.DEV research release derived from
[`LiquidAI/LFM2.5-Encoder-230M`](https://huggingface.co/LiquidAI/LFM2.5-Encoder-230M) at revision `0b649ad0c684378b03d4d8304f7577a662ab89bc`. It is
not an official Liquid AI release. We removed the masked-language-model head and
contrastively fine-tuned the full bidirectional encoder for multilingual code retrieval.

## Intended use

This BF16 checkpoint is the quality reference and the portable Transformers artifact.
The 230M release is particularly intended to make small-device and iPhone-class
experiments possible; deployment feasibility and power use still depend on the client
runtime and hardware.

## Held-out retrieval results

All rows use the same untouched 6,995-pair multilingual test set, 1,200-character query
and 4,000-character passage caps, query token cap 512, and passage token cap 2,048.
Higher is better. RTN is a matched quantization control; Nomic and Jina are external
service baselines, not architecture-matched controls.

| Model | MRR | R@1 | R@5 | R@10 | NDCG@10 | Python MRR | TypeScript MRR | Artifact |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| LFM2.5 230M BF16 | 0.3960 | 0.3212 | 0.4756 | 0.5375 | 0.4235 | 0.8360 | 0.2527 | 464.2 MB |
| LFM2.5 230M calibrated MXFP4 | 0.2110 | 0.1620 | 0.2572 | 0.3029 | 0.2262 | 0.7113 | 0.0693 | 225.5 MB |
| LFM2.5 230M RTN MXFP4 | 0.0806 | 0.0609 | 0.0931 | 0.1139 | 0.0844 | 0.4435 | 0.0191 | 225.4 MB |
| LFM2.5 230M calibrated MXFP8 | 0.3934 | 0.3197 | 0.4715 | 0.5342 | 0.4207 | 0.8394 | 0.2447 | 306.8 MB |
| LFM2.5 230M RTN MXFP8 | 0.3878 | 0.3162 | 0.4622 | 0.5245 | 0.4139 | 0.8322 | 0.2459 | 306.7 MB |
| Nomic v1.5 service | 0.5439 | 0.4968 | 0.5954 | 0.6236 | 0.5595 | 0.9289 | 0.3617 | service |
| Jina calibrated MXFP4 | 0.6645 | 0.6133 | 0.7221 | 0.7571 | 0.6832 | 0.9462 | 0.5057 | 1167.7 MB |

A separate BF16 cross-runtime run on `NVIDIA GeForce RTX 3090 Ti` with PyTorch `2.13.0+cu130` produced MRR 0.3963, 755.7 queries/s, 190.8 passages/s, and 736.2 MB peak CUDA allocation. CUDA throughput is reported separately and is not compared directly with Metal.

A paired 10,000-sample bootstrap estimates calibrated MXFP8 minus BF16 MRR at -0.0026, with a 95% interval of [-0.0042, -0.0010]. A point estimate whose interval crosses zero is not presented as a
quality win.

## Usage

```python
import torch
import torch.nn.functional as F
from transformers import AutoModel, AutoTokenizer

repo = "RESMP-DEV/LFM2.5-Encoder-230M-Code-BF16"
tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModel.from_pretrained(repo, trust_remote_code=True, dtype=torch.bfloat16)
texts = [
    "query: parse a TypeScript AST",
    "passage: export function parse(source: string) {{ ... }}",
]
batch = tokenizer(texts, padding=True, truncation=True, return_tensors="pt")
with torch.no_grad():
    hidden = model(**batch).last_hidden_state.float()
mask = batch["attention_mask"].unsqueeze(-1)
vectors = F.normalize((hidden * mask).sum(1) / mask.sum(1), dim=-1)
```

## Training and data receipts

Full-backbone symmetric in-batch InfoNCE training used 24,626
language-balanced pairs selected from the 42,626-row source training
split, two epochs, batch size 32, learning rate 2e-5, temperature 0.05, and seed 17. The
training report records the NVIDIA RTX A6000 runtime and validation history.

- `train`: 42,626 rows, SHA-256 `426ebfaad34b14d7627ba6e668ae36e08e548c9d057b0edc208bcfa6fe527629`
- `validation`: 5,319 rows, SHA-256 `9ac88b3138de4ca94c2ef3a87ccf19381fc76265c2bc9d65b4791983d0315096`
- `test`: 6,995 rows, SHA-256 `9ed10842a12132b6bfb5421df1e2f88dbcfbf6f6e960f36b22eb9ea6e3c72315`
- `calibration`: 4,096 rows, SHA-256 `ee9edaf80a6854c18053b96521090a51bdb76642abeb98618d7aed36e70b6de9`

The corpus combines pinned CodeSearchNet data with pinned permissively licensed code
repositories. Exact and token 8-gram near-duplicates were removed with test-before-
validation-before-train precedence. See `corpus_receipt.json`, `source_receipt.json`,
`training_report.json`, `quantization_report.json` when present, `benchmarks/`, and
`artifact_manifest.json` for machine-readable evidence.

## License and attribution

The weights retain the LFM Open License v1.0 in `LICENSE`, including its attribution and
commercial-use conditions. `MODIFICATIONS.md` identifies RESMP.DEV's changes. The
[training and quantization workbench](https://github.com/RESMP-DEV/calibrated-code-embeddings)
is separately MIT licensed.

## Citation

```bibtex
@article{liquidAI2026Encoders,
  author = {Liquid AI},
  title = {LFM2.5-Encoders: Fast at Long Context, Even on CPU},
  journal = {Liquid AI Blog},
  year = {2026},
  note = {www.liquid.ai/blog/lfm2-5-encoders},
}
```