File size: 6,630 Bytes
b363ebd
 
45a3d2d
 
 
 
 
 
 
 
 
909d29e
 
45a3d2d
 
909d29e
 
 
 
 
 
 
 
 
 
 
5bef2cd
909d29e
 
b363ebd
 
45a3d2d
b363ebd
45a3d2d
b363ebd
 
 
 
 
45a3d2d
b363ebd
45a3d2d
 
 
 
b363ebd
 
 
 
 
 
45a3d2d
b363ebd
45a3d2d
 
 
b363ebd
45a3d2d
b363ebd
45a3d2d
b363ebd
 
 
45a3d2d
 
 
b363ebd
 
 
45a3d2d
b363ebd
45a3d2d
 
 
b363ebd
 
 
45a3d2d
 
 
b363ebd
 
 
45a3d2d
 
b363ebd
7efc087
 
45a3d2d
 
 
 
 
 
b363ebd
 
 
 
 
 
 
45a3d2d
b363ebd
 
 
 
 
 
 
45a3d2d
 
 
b363ebd
45a3d2d
 
 
 
 
b363ebd
 
 
45a3d2d
 
 
 
 
 
 
 
 
 
 
 
 
 
b363ebd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
45a3d2d
 
 
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
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
---
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