Sparse MoE as the New Dropout: Scaling Dense and Self-Slimmable Transformers
Paper • 2303.01610 • Published
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Quantization made by Richard Erkhov.
phi3-4x4b-v1 - GGUF
- Model creator: https://huggingface.co/Fizzarolli/
- Original model: https://huggingface.co/Fizzarolli/phi3-4x4b-v1/
| Name | Quant method | Size |
|---|---|---|
| phi3-4x4b-v1.Q2_K.gguf | Q2_K | 3.79GB |
| phi3-4x4b-v1.IQ3_XS.gguf | IQ3_XS | 4.23GB |
| phi3-4x4b-v1.IQ3_S.gguf | IQ3_S | 4.47GB |
| phi3-4x4b-v1.Q3_K_S.gguf | Q3_K_S | 4.47GB |
| phi3-4x4b-v1.IQ3_M.gguf | IQ3_M | 4.59GB |
| phi3-4x4b-v1.Q3_K.gguf | Q3_K | 4.97GB |
| phi3-4x4b-v1.Q3_K_M.gguf | Q3_K_M | 4.97GB |
| phi3-4x4b-v1.Q3_K_L.gguf | Q3_K_L | 5.39GB |
| phi3-4x4b-v1.IQ4_XS.gguf | IQ4_XS | 5.56GB |
| phi3-4x4b-v1.Q4_0.gguf | Q4_0 | 5.83GB |
| phi3-4x4b-v1.IQ4_NL.gguf | IQ4_NL | 5.87GB |
| phi3-4x4b-v1.Q4_K_S.gguf | Q4_K_S | 5.88GB |
| phi3-4x4b-v1.Q4_K.gguf | Q4_K | 6.25GB |
| phi3-4x4b-v1.Q4_K_M.gguf | Q4_K_M | 6.25GB |
| phi3-4x4b-v1.Q4_1.gguf | Q4_1 | 6.46GB |
| phi3-4x4b-v1.Q5_0.gguf | Q5_0 | 7.1GB |
| phi3-4x4b-v1.Q5_K_S.gguf | Q5_K_S | 7.1GB |
| phi3-4x4b-v1.Q5_K.gguf | Q5_K | 7.32GB |
| phi3-4x4b-v1.Q5_K_M.gguf | Q5_K_M | 7.32GB |
| phi3-4x4b-v1.Q5_1.gguf | Q5_1 | 7.74GB |
| phi3-4x4b-v1.Q6_K.gguf | Q6_K | 8.46GB |
| phi3-4x4b-v1.Q8_0.gguf | Q8_0 | 10.96GB |
Original model description:
license: mit tags: - phi3 - nlp - moe datasets: - BEE-spoke-data/gutenberg-en-v1-clean - NeelNanda/pile-10k
phi 3 4x4b
a continually pretrained phi3-mini sparse moe upcycle
benchmarks
ran locally
| Microsoft/phi-3-4k-instruct | Fizzarolli/phi3-4x4b-v1 | |
|---|---|---|
| MMLU acc. (0-shot) | 0.6799 | 0.6781 |
| Hellaswag acc. (0-shot) | 0.6053 | 0.5962 |
| ARC-E acc. (0-shot) | 0.8325 | 0.8367 |
| ARC-C acc. (0-shot) | 0.5546 | 0.5606 |
honestly i was expecting it to do worse :p, but those are all within a margin of error! so it didn't lose any performance, at least
open llm leaderboard
todo!
support me on ko-fi!
please i need money to stay alive and keep making models
notes
not trained on instruct data. it's pretty likely that it won't be much different from phi 3 if you use it like that, if not worse due to any forgetting of instruct formats during the continued training.
future experiments
- the datasets for this were literally chosen on a whim. perhaps experiment with a further filtered HuggingFaceFW/fineweb-edu?
- actually freeze the gate layers next time (see Chen et. al, 2023), oops
- MOAR TRAINING, this only went up to ~0.2 of an epoch because i ran out of dolar
- Downloads last month
- 1,014
Hardware compatibility
Log In to add your hardware
2-bit
3-bit
4-bit
5-bit
6-bit
8-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support