README.md

---
library_name: transformers
license: mit
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
tags:
- generated_from_trainer
- conversational
- instruction-tuned
- smoltalk
datasets:
- HuggingFaceTB/smoltalk
metrics:
- MMLU
language:
- en

model-index:
- name: DeepSeek-R1-Distill-Qwen-1.5B-finetuned-smoltalk-everyday-conversations
  results:
  - task:
      name: Text Generation
      type: text-generation
    dataset:
      name: HuggingFaceTB/smoltalk
      type: HuggingFaceTB/smoltalk
    metrics:
    - name: MMLU-PEM (0-shot)
      type: MMLU-PEM (0-shot)
      value: 0.2749
---

# Model Card for DeepSeek-R1-SmolTalk

This model is a fine-tuned version of [`deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B`](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) on the [SmolTalk dataset](https://huggingface.co/datasets/HuggingFaceTB/smoltalk). It is optimized for small-scale, friendly, and engaging instruction-following dialogue.

## Model Details

### Model Description

This model builds on DeepSeek's distilled Qwen-1.5B architecture and is trained for conversational tasks using the SmolTalk dataset. The goal is to create a lightweight, instruction-following model suitable for use in chatbots or lightweight assistants with limited hardware resources.

- **Model type:** Instruction-tuned causal decoder (chat)
- **Language(s):** English
- **License:** MIT
- **Finetuned from model:** deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B


## Uses

### Direct Use

This model can be used as a lightweight assistant or chatbot in applications such as:

- Embedded conversational interfaces
- Educational or toy assistants
- Small devices or local applications

### Downstream Use

The model can be further fine-tuned or integrated into larger conversational systems, especially where resource efficiency is crucial.

### Out-of-Scope Use

- Not suitable for tasks requiring deep factual accuracy or reasoning
- Should not be used for sensitive or high-stakes decision making
- Not designed for multilingual use

## Bias, Risks, and Limitations

Due to the small model size and dataset limitations:

- May produce generic or incorrect outputs
- Can reflect biases present in the training dataset
- Not guaranteed to be safe for all user demographics or use cases

### Recommendations

- Use in controlled or sandboxed environments
- Consider integrating content moderation or rule-based filtering
- Do not deploy in contexts requiring factual correctness or ethical judgment

## How to Get Started with the Model

```Python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("avanishd/DeepSeek-R1-Distill-Qwen-1.5B-finetuned-smoltalk-everyday-conversations")
tokenizer = AutoTokenizer.from_pretrained("avanishd/DeepSeek-R1-Distill-Qwen-1.5B-finetuned-smoltalk-everyday-conversations")

input_text = "Hi there! What can you do?"
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Training Details

### Training Data

<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->

Used [SmolTalk dataset](https://huggingface.co/datasets/HuggingFaceTB/smoltalk), a dataset of lightweight, instruction-style conversations. The dataset is designed to help models learn concise, friendly, and helpful interactions.

### Training Procedure

<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->

#### Preprocessing [optional]

Used the DeepSeek tokenizer

#### LoRA Configuration

- rank: 6
- alpha: 12
- dropout: 0.05
- bias: none
- target: linear

#### Training Hyperparameters

The following hyperparameters were used during training:

- learning_rate: 2e-04
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- gradient_clipping: 0.3
- total_train_batch_size: 128
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED
- lr_scheduler_type: constant
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 1
- mixed_precision_training: bf16

#### Speeds, Sizes, Times [optional]

<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->

[More Information Needed]

## Evaluation

<!-- This section describes the evaluation protocols and provides the results. -->

### Testing Data, Factors & Metrics

#### Testing Data

<!-- This should link to a Dataset Card if possible. -->

[More Information Needed]

#### Factors

<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->

[More Information Needed]

#### Metrics

<!-- These are the evaluation metrics being used, ideally with a description of why. -->

[More Information Needed]

### Results

[More Information Needed]

#### Summary



## Model Examination [optional]

<!-- Relevant interpretability work for the model goes here -->

[More Information Needed]

## Environmental Impact

<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->

Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).

- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]

## Technical Specifications [optional]

### Model Architecture and Objective

[More Information Needed]

### Compute Infrastructure

[More Information Needed]

#### Hardware

[More Information Needed]

#### Software

[More Information Needed]

## Citation [optional]

<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

[More Information Needed]

**APA:**

[More Information Needed]

## Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->

[More Information Needed]

## More Information [optional]

[More Information Needed]

## Model Card Authors [optional]

[More Information Needed]

## Model Card Contact

[More Information Needed]

fill this model card