phi-4-deepseek-reasoning

Model Card for Model ID

Phi-4 trained on reasoning outputs on complex logic, math and coding challenges derived from nvidia/Llama-Nemotron-Post-Training-Dataset filtered to include high length reasoning answers generated by DeepSeek R1.

Model Details

Model Description

Phi-4 trained on reasoning outputs on complex logic, math and coding challenges derived from nvidia/Llama-Nemotron-Post-Training-Dataset filtered to include high length reasoning answers generated by DeepSeek R1. The training was on 10,000 samples done on an RTX 5090 (yes managed to make unsloth work on a 5090) with context length of 16384 and took around 10 hours using unsloth 4-bit quants and transfomers SFT Trainer. You do not need to add a system prompt but it can help in some use cases. The model will automatically go into thinking mode when presented with complex tasks.

Recommended Settings of temperature = 1.5 (you can test with 1 to 1.5) , min_p = 0.1, repeat penalty 1.2 or 1.3 to mitigate extremely long reasoning around the same concept.

Try the following prompt or similar structured prompts containing complex connections and the model will automatically go into thinking mode and generate long reasoning chains akin to DeepSeek.

Prompt:

This prompt was generated using Claude 3.7 Sonnet and not included in the train or test dataset, use similarly structred prompts and see the magic!

1. Network Packet Routing Optimization Challenge

You're designing a system to optimize packet routing in a network with multiple possible paths. The network consists of nodes connected by bidirectional links, each with different bandwidth capacities and latency values.

Your task is to find the most efficient routing path between a given source and destination node that satisfies specific constraints on bandwidth, latency, and hop count.

Input Specification

The first line contains four space-separated integers: n, m, b_min, and l_max (2 ≤ n ≤ 100, 1 ≤ m ≤ 5000, 1 ≤ b_min ≤ 1000, 1 ≤ l_max ≤ 10000)

• n: number of nodes in the network (numbered from 1 to n)
• m: number of links between nodes
• b_min: minimum required bandwidth for the path
• l_max: maximum allowed total latency for the path

The next m lines each contain four integers u, v, b, l (1 ≤ u, v ≤ n, u ≠ v, 1 ≤ b ≤ 1000, 1 ≤ l ≤ 1000):

• u, v: nodes connected by this link
• b: bandwidth capacity of the link
• l: latency of the link

The last line contains two integers s and t (1 ≤ s, t ≤ n, s ≠ t) - the source and destination nodes.

Constraints and Notes

1. The bandwidth of a path is the minimum bandwidth among all links in the path
2. The latency of a path is the sum of latencies of all links in the path
3. A valid path must have bandwidth ≥ b_min and latency ≤ l_max
4. Among all valid paths, you must choose the one with the highest bandwidth
5. If there are multiple paths with the same highest bandwidth, choose the one with the lowest latency
6. If there are still multiple paths, choose the one with the fewest hops (links)
7. If no valid path exists, output "NO PATH"

Output

If a valid path exists, the first line should contain three space-separated integers: the bandwidth of the chosen path, the total latency of the chosen path, and the number of hops.

The second line should contain the sequence of nodes in the path, starting with s and ending with t.

If no valid path exists, output "NO PATH" (without quotes).

Examples

Example 1:

5 6 50 100
1 2 100 20
2 3 80 30
3 5 70 10
1 4 60 10
4 5 90 30
1 3 50 5
1 5

Output:

70 60 3
1 2 3 5

Example 2:

4 5 80 50
1 2 80 20
2 3 120 15
3 4 90 10
1 3 100 30
2 4 70 25
1 4

Output:

90 40 2
1 3 4

Example 3:

3 3 100 100
1 2 150 40
2 3 180 70
1 3 120 30
1 3

Output:

120 30 1
1 3

Your solution should efficiently find the optimal path that satisfies all constraints, handling potentially complex network topologies with multiple possible routes between source and destination.

• Developed by: [More Information Needed]
• Funded by [optional]: [More Information Needed]
• Shared by [optional]: [More Information Needed]
• Model type: [More Information Needed]
• Language(s) (NLP): [More Information Needed]
• License: [More Information Needed]
• **Finetuned from model [optional]:**unsloth/phi-4

Uses

Complex reasoning requiring challenging thinking and coding (mostly python).

Direct Use

<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->

[More Information Needed]

Downstream Use [optional]

<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->

[More Information Needed]

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->

[More Information Needed]

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

[More Information Needed]

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

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. -->

[More Information Needed]

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]

[More Information Needed]

Training Hyperparameters

• Training regime: [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->

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 presented in Lacoste et al. (2019).

• 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]