metadata

license: apache-2.0
datasets:
  - BeIR/nq
  - embedding-data/PAQ_pairs
  - sentence-transformers/msmarco-hard-negatives
  - leminda-ai/s2orc_small
  - lucadiliello/triviaqa
  - pietrolesci/agnews
  - mteb/amazon_reviews_multi
  - multiIR/ccnews2016-8multi
  - eli5
  - gooaq
  - quora
  - lucadiliello/searchqa
  - flax-sentence-embeddings/stackexchange_math_jsonl
  - yahoo_answers_qa
  - EdinburghNLP/xsum
  - wikihow
  - rajpurkar/squad_v2
  - nixiesearch/amazon-esci
  - osunlp/Mind2Web
  - derek-thomas/dataset-creator-askreddit
language:
  - en

nixie-querygen-v2

A Mistral-7B-v0.1 fine-tuned on query generation task. Main use cases:

synthetic query generation for downstream embedding fine-tuning tasks - when you have only documents and no queries/labels. Such task can be done with the nixietune toolkit, see the nixietune.qgen.generate recipe.
synthetic dataset expansion for further embedding training - when you DO have query-document pairs, but only a few. You can fine-tune the nixie-querygen-v2 on existing pairs, and then expand your document corpus with synthetic queries (which are still based on your few real ones). See nixietune.qgen.train recipe.

Training data

We used 200k query-document pairs sampled randomly from a diverse set of IR datasets:

Flavours

This repo has multiple versions of the model:

model-*.safetensors: Pytorch FP16 checkpoint, suitable for down-stream fine-tuning
ggml-model-f16.gguf: GGUF F16 non-quantized llama-cpp checkpoint, for CPU inference
ggml-model-q4.gguf: GGUF Q4_0 quantized llama-cpp checkpoint, for fast (and less precise) CPU inference.

Prompt formats

The model accepts the followinng prompt format:

<document next> [short|medium|long]? [question|regular]? query:

Some notes on format:

[short|medium|long] and [question|regular] fragments are optional and can be skipped.
the prompt suffix query: has no trailing space, be careful.

Inference example

With llama-cpp and Q4 model the inference can be done on a CPU:

$ ./main -m ~/models/nixie-querygen-v2/ggml-model-q4.gguf -p "git lfs track will \
begin tracking a new file or an existing file that is already checked in to your \
repository. When you run git lfs track and then commit that change, it will \
update the file, replacing it with the LFS pointer contents. short regular query:" -s 1

sampling: 
        repeat_last_n = 64, repeat_penalty = 1.100, frequency_penalty = 0.000, presence_penalty = 0.000
        top_k = 40, tfs_z = 1.000, top_p = 0.950, min_p = 0.050, typical_p = 1.000, temp = 0.800
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000
sampling order: 
CFG -> Penalties -> top_k -> tfs_z -> typical_p -> top_p -> min_p -> temp 
generate: n_ctx = 512, n_batch = 512, n_predict = -1, n_keep = 0


 git lfs track will begin tracking a new file or an existing file that is 
 already checked in to your repository. When you run git lfs track and then 
 commit that change, it will update the file, replacing it with the LFS 
 pointer contents. short regular query: git-lfs track [end of text]

Training config

The model is trained with the follwing nixietune config:

{
    "train_dataset": "/home/shutty/data/nixiesearch-datasets/query-doc/data/train",
    "eval_dataset": "/home/shutty/data/nixiesearch-datasets/query-doc/data/test",
    "seq_len": 512,
    "model_name_or_path": "mistralai/Mistral-7B-v0.1",
    "output_dir": "mistral-qgen",
    "num_train_epochs": 1,
    "seed": 33,
    "per_device_train_batch_size": 6,
    "per_device_eval_batch_size": 2,
    "bf16": true,
    "logging_dir": "logs",
    "gradient_checkpointing": true,
    "gradient_accumulation_steps": 1,
    "dataloader_num_workers": 14,
    "eval_steps": 0.03,
    "logging_steps": 0.03,
    "evaluation_strategy": "steps",
    "torch_compile": false,
    "report_to": [],
    "save_strategy": "epoch",
    "streaming": false,
    "do_eval": true,
    "label_names": [
        "labels"
    ]
}

License

Apache 2.0