Downloads · 30 days
768
7% of all-time downloads
RichardErkhov/leia-llm_-_Leia-Swallow-13b-gguf
leia-llm_-_Leia-Swallow-13b-gguf is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
768
7% of all-time downloads
All-time downloads
11K
Public
Repo size
149 GB
Likes
0
Public
Click a slice to open those files.
.gguf173 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Leia-Swallow-13b - GGUF
| Name | Quant method | Size |
|---|---|---|
| Leia-Swallow-13b.Q2_K.gguf | Q2_K | 4.58GB |
| Leia-Swallow-13b.IQ3_XS.gguf | IQ3_XS | 5.06GB |
| Leia-Swallow-13b.IQ3_S.gguf | IQ3_S | 5.34GB |
| Leia-Swallow-13b.Q3_K_S.gguf | Q3_K_S | 5.34GB |
| Leia-Swallow-13b.IQ3_M.gguf | IQ3_M | 5.64GB |
| Leia-Swallow-13b.Q3_K.gguf | Q3_K | 5.97GB |
| Leia-Swallow-13b.Q3_K_M.gguf | Q3_K_M | 5.97GB |
| Leia-Swallow-13b.Q3_K_L.gguf | Q3_K_L | 6.52GB |
| Leia-Swallow-13b.IQ4_XS.gguf | IQ4_XS | 6.61GB |
| Leia-Swallow-13b.Q4_0.gguf | Q4_0 | 6.93GB |
| Leia-Swallow-13b.IQ4_NL.gguf | IQ4_NL | 6.98GB |
| Leia-Swallow-13b.Q4_K_S.gguf | Q4_K_S | 6.99GB |
| Leia-Swallow-13b.Q4_K.gguf | Q4_K | 7.4GB |
| Leia-Swallow-13b.Q4_K_M.gguf | Q4_K_M | 7.4GB |
| Leia-Swallow-13b.Q4_1.gguf | Q4_1 | 7.69GB |
| Leia-Swallow-13b.Q5_0.gguf | Q5_0 | 8.44GB |
| Leia-Swallow-13b.Q5_K_S.gguf | Q5_K_S | 8.44GB |
| Leia-Swallow-13b.Q5_K.gguf | Q5_K | 8.68GB |
| Leia-Swallow-13b.Q5_K_M.gguf | Q5_K_M | 8.68GB |
| Leia-Swallow-13b.Q5_1.gguf | Q5_1 | 9.19GB |
| Leia-Swallow-13b.Q6_K.gguf | Q6_K | 10.03GB |
| Leia-Swallow-13b.Q8_0.gguf | Q8_0 | 12.99GB |
license: apache-2.0 language:
LEIA is a training technique for autoregressive LLMs that effectively improves their performance in languages other than English by enhancing cross-lingual knowledge transfer from English to a target language. This model is constructed by applying LEIA to Swallow, a Japanese-English bilingual LLM based on LLaMA 2. The model achieves enhanced performance on four out of six Japanese question answering benchmarks and equivalent performance on the remaining two, as reported below.
Please refer to our paper or blog post (in Japanese) for further technical details.
The model is assessed using the following six question answering benchmarks:
| Model | X-CODAH | X-CSQA | JCommonsenseQA | NIILC | JEMHopQA | JAQKET v2 |
|---|---|---|---|---|---|---|
| Swallow | 43.3 | 41.8 | 89.3 | 64.1 | 50.6 | 88.9 |
| LEIA | 44.0 | 41.9 | 89.3 | 65.8 | 50.6 | 89.6 |
For further details of this experiment, please refer to our paper.