Instructions to use AShi846/all-MiniLM-L6-v2_rag_ft_e-7 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AShi846/all-MiniLM-L6-v2_rag_ft_e-7 with sentence-transformers:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("AShi846/all-MiniLM-L6-v2_rag_ft_e-7")

sentences = [
    "Suppose we use the Simplex method to solve the following linear program: \\begin{align*} \\textbf{maximize} \\hspace{0.8cm} & \\hspace{0.4cm}4x_1 - 6x_2 + 4x_3 \\\\ \\textbf{subject to}\\hspace{0.6cm} & x_1 - 3x_2 + x_3  + s_1  = 1 \\\\ \\hspace{0.8cm} & \\hspace{1.90cm}x_1  + s_2  = 8 \\\\ \\hspace{0.8cm} & \\hspace{0.65cm} 3x_2  + 2x_3 + s_3 = 6 \\\\ \\hspace{0.8cm} &\\hspace{-0.35cm}  x_1,\\: x_2, \\: x_3, \\:s_1, \\:s_2, \\:s_3 \\geq 0 \\end{align*} At the current step, we have the following Simplex tableau: \\begin{align*} \\hspace{1cm} x_1 &= 1 + 3x_2 - x_3 - s_1 \\\\ s_2 &= 7 -3x_2 + x_3  + s_1  \\\\ s_3 &= 6 - 3x_2  - 2x_3 \\\\ \\cline{1-2} z &= 4 + 6 x_2   -  4s_1 \\end{align*} Write the tableau obtained by executing one iteration (pivot) of the Simplex method starting from the above tableau.",
    "We want to assign each job to a machine. For any job $j$ which the LP assigns to a single machine $i$, i.e., $x^*_{ij} = 1$, it clearly makes sense to listen to the LP and assign $j$ to $i$. To assign the other jobs, we will use the graph $H$, which has the following properties: \\begin{itemize} \\item $H$ is acyclic, i.e., a forest. \\item Vertices corresponding to the already-assigned jobs are isolated. \\item Vertices corresponding to the yet-unassigned jobs have degree at least two. \\end{itemize} While $H$ contains an edge, we iteratively do the following: \\begin{itemize} \\item Choose a machine $i$ whose vertex $a_i$ has degree one in $H$, i.e., there is only a single edge $\\{a_i, b_j\\}$ incident to $a_i$. (Such a machine exists because $H$ is nonempty, and a nonempty forest contains a degree-one vertex; since job-vertices have degrees either zero or at least two, a degree-one vertex must be a machine-vertex.) \\item Assign job $j$ to machine $i$. \\item Remove all edges incident to $j$ from $H$. (Note that this preserves all the above properties of $H$.) \\end{itemize} Obviously, this procedure will terminate. Note that it will assign all jobs to machines, because as long as there is an unassigned job $j$, the vertex $b_j$ has at least two incident edges in $H$. It remains to reason that for every machine $i$, the sum of processing times of jobs assigned to $i$ (this is called the \\textit{makespan} of $i$) is at most $T + \\max_{j \\in J} p_j$. This is because there are two kinds of jobs assigned to $i$: \\begin{itemize} \\item Assigned in the beginning: the jobs $j$ which had $x^*_{ij} = 1$. The total processing time of these jobs is \\[ \\sum_{j\\in J: x^*_{ij} = 1} p_j = \\sum_{j\\in J: x^*_{ij} = 1} x^*_{ij} p_j \\le \\sum_{j\\in J: i \\in N(j)} x^*_{ij} p_j \\le T. \\] \\item Assigned later: at most one other job. For note that we only assign a job to $i$ when $a_i$ is of degree-one, and once we do, $a_i$ becomes isolated and no more jobs can be assigned to $i$. This job has processing time at most $\\max_{j \\in J} p_j$. \\end{itemize} Another perspective on the solution is the following. By looking at the makespan bound $T + \\max_j p_j$ that we should satisfy, we can notice that it will be fine if we assign the integral jobs to the machines that the LP has selected and the non-integral jobs in such a way that each machine gets at most one. In other words, we just need to prove that the graph $H$ has a matching which matches all non-integral jobs (any such matching will be fine). Our iterative procedure is one way to prove that such a matching exists.",
    "1",
    "It breaks backward compatibility, because the signature changes and \"Document\" is not a special kind of \"String\" thus callers will have to be updated"
]
embeddings = model.encode(sentences)

similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4]

Notebooks
Google Colab
Kaggle

SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("AShi846/all-MiniLM-L6-v2_rag_ft_e-7")
# Run inference
sentences = [
    'Show a code snippet which represents the kernel of a Spectre\n  attack (use any convenient programming language or assembly).\n',
    "No, this will force other members to sit through an explanation that's not relevant to them, the colleague was right.",
    'Whatever your personal convictions, you have to adapt to the convention used by the project.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Training Details

Training Dataset

Unnamed Dataset

Size: 475 training samples
Columns: sentence_0, sentence_1, and label
Approximate statistics based on the first 475 samples:
sentence_0 sentence_1 label
type string string float
details
min: 5 tokens
mean: 135.81 tokens
max: 256 tokens

min: 3 tokens
mean: 110.0 tokens
max: 256 tokens

min: 0.1
mean: 0.1
max: 0.1

	sentence_0	sentence_1	label
type	string	string	float
details	min: 5 tokens mean: 135.81 tokens max: 256 tokens	min: 3 tokens mean: 110.0 tokens max: 256 tokens	min: 0.1 mean: 0.1 max: 0.1

Samples:

sentence_0	sentence_1	label
`Now let $\xv$ be a random vector distributed according to the uniform distribution over the finite centered dataset $\xv_1, . . . , \xv_N$ from above. % Consider the problem of finding a unit vector, $\wv \in \R^D$, such that the random variable $\wv^ op \xx$ has \emph{maximal} variance. What does it mean for the data vectors $\xv_1, . . . , \xv_N$ to be centered, as for principle component analysis (PCA) to be meaningful? Use the notation $\x_{nd}$ for individual entries.`	`False.`	`0.1`
In the following problem Alice holds a string $x = \langle x_1, x_2, \ldots, x_n \rangle$ and Bob holds a string $y = \langle y_1, y_2, \ldots, y_n\rangle$. Both strings are of length $n$ and $x_i, y_i \in {1,2,\ldots, n}$ for $i=1,2, \ldots, n$. The goal is for Alice and Bob to use little communication to estimate the quantity \begin{align} Q = \sum_{i=1}^n (x_i + y_i)^2,. \end{align} A trivial solution is for Alice to transfer all of her string $x$ to Bob who then computes $Q$ exactly. However this requires Alice to send $\Theta(n \log n)$ bits of information to Bob. In the following, we use randomization and approximation to achieve a huge improvement on the number of bits transferred from Alice to Bob. Indeed, for a small parameter $\epsilon > 0$, your task is to devise and analyze a protocol of the following type: \begin{itemize} \item On input $x$, Alice uses a randomized algorithm to compute a message $m$ that consists of $O(\log (n)/\epsilon^2)$ bits. She then transmits ...	`The algorithm is as follows: \begin{itemize} \item If $x_1 \geq W$, then do the exchange on day $1$ and receive $x_1$ Swiss francs. \item Otherwise, do the exchange on day $2$ and receive $x_2$ Swiss francs. \end{itemize} We now analyze its competitiveness. If $x_1 \geq W$, then our algorithm gets at least $W$ Swiss francs. Optimum is at most $W^2$ and so we are $1/W$ competitive. Otherwise if $x_1 < W$ then we get $x_2 \geq 1$ Swiss francs which is $x_2/ \max(x_2, x_1) \geq 1/W$ competitive.`	`0.1`
`What does it mean that a processor implements precise exceptions?`	`When preparing benchmarks, your colleague should make sure that the inputs are representative of the real world, and complex enough to potentially detect performance bottlenecks. In this example, the inputs strings should be long enough.`	`0.1`

Loss: CosineSimilarityLoss with these parameters:

{
    "loss_fct": "torch.nn.modules.loss.MSELoss"
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
per_device_eval_batch_size: 16
num_train_epochs: 7
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 7
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
tp_size: 0
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Framework Versions

Python: 3.12.8
Sentence Transformers: 3.4.1
Transformers: 4.51.3
PyTorch: 2.6.0+cu126
Accelerate: 1.3.0
Datasets: 3.2.0
Tokenizers: 0.21.0

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}