LiquidAI/LFM2.5-1.2B-Base fine-tuned with LoRA for QMD query
expansion using the v2 data recipe. This repository contains
the merged BF16 Transformers checkpoint at its root and QMD-ready GGUF
quantizations alongside it.
Repository formats
Merged BF16 Transformers checkpoint: load the repository directly with
AutoModelForCausalLM.from_pretrained("langleu/qmd-query-expansion-lfm2.5-1.2b-base-v2", torch_dtype=torch.bfloat16).
GGUF for QMD/llama.cpp:qmd-query-expansion-lfm2.5-1.2b-q8_0.gguf, qmd-query-expansion-lfm2.5-1.2b-q5_k_m.gguf, qmd-query-expansion-lfm2.5-1.2b-q4_k_m.gguf
The BF16 checkpoint is the unquantized merged fine-tune. The GGUF files are
derived from that same merged checkpoint.
This variant supports the production v2 QMD behavior:
an optional Query intent: line and /only:lex, /only:vec, and /only:hyde
directives.
Each emitted line begins with lex:, vec:, or hyde:.
Use with QMD
The Transformers checkpoint is not loaded by QMD directly;
select one of the GGUF files.
For a private repository, authenticate once with hf auth login, or set
HF_TOKEN. QMD's node-llama-cpp downloader reads the cached Hugging Face
token from ~/.cache/huggingface/token by default.
Training records were rebuilt from the pinned dataset's
messages field with the target LFM2.5 tokenizer. The dataset's preformatted
Qwen text was not used. Completion-only loss masked the user prompt.
Validation
Average QMD reward: 95.17%
Format compliance: 100.00%
Entity preservation: 98.31%
/only:* behavior: 100.00%
Hard failures: 0
BF16 and GGUF comparison
Format
Avg reward
Δ vs BF16
Format compliance
Entity
/only:*
BF16
95.13%
baseline
100.00%
100.00%
100.00%
Q8_0
94.91%
-0.22 pp
99.68%
96.61%
98.85%
Q5_K_M
95.17%
+0.04 pp
100.00%
98.31%
100.00%
Q4_K_M
94.76%
-0.37 pp
100.00%
98.31%
100.00%
Quality changes are reported in percentage points (pp), not relative
percent. Small positive GGUF deltas can occur because sampled generation is
not bit-for-bit deterministic.
Speed percentages are intentionally not reported. BF16 was measured with
batched Transformers inference, while QMD runs GGUF through llama.cpp one query
at a time. Their observed throughput and latency are useful operational
measurements, but dividing them would not be an apples-to-apples speedup.
Credits
Liquid AI for LFM2.5 and the LFM Open
License v1.0.
Tobi for QMD, the QMD query-expansion
datasets, evaluation/scoring design, and the Qwen3 query-expansion model.
OrcsRise for
the earlier LFM2 QMD fine-tuning work that informed the LFM target-module
recipe.
This derivative is governed by the LFM Open License v1.0. The included
LICENSE must be retained, including its attribution and commercial-use terms.
The upstream QMD dataset card did not declare an explicit dataset license at the
pinned revision. This repository records that fact and does not imply that a
license was granted. Users and redistributors are responsible for confirming
that their use is authorized.
