Downloads · 30 days
87
4% of all-time downloads
RichardErkhov/afrizalha_-_Bakpia-V1-1.5B-Javanese-gguf
afrizalha_-_Bakpia-V1-1.5B-Javanese-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
87
4% of all-time downloads
All-time downloads
2.4K
Public
Repo size
16.3 GB
Likes
0
Public
Click a slice to open those files.
.gguf19.2 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Bakpia-V1-1.5B-Javanese - GGUF
| Name | Quant method | Size |
|---|---|---|
| Bakpia-V1-1.5B-Javanese.Q2_K.gguf | Q2_K | 0.63GB |
| Bakpia-V1-1.5B-Javanese.Q3_K_S.gguf | Q3_K_S | 0.71GB |
| Bakpia-V1-1.5B-Javanese.Q3_K.gguf | Q3_K | 0.77GB |
| Bakpia-V1-1.5B-Javanese.Q3_K_M.gguf | Q3_K_M | 0.77GB |
| Bakpia-V1-1.5B-Javanese.Q3_K_L.gguf | Q3_K_L | 0.82GB |
| Bakpia-V1-1.5B-Javanese.IQ4_XS.gguf | IQ4_XS | 0.84GB |
| Bakpia-V1-1.5B-Javanese.Q4_0.gguf | Q4_0 | 0.87GB |
| Bakpia-V1-1.5B-Javanese.IQ4_NL.gguf | IQ4_NL | 0.88GB |
| Bakpia-V1-1.5B-Javanese.Q4_K_S.gguf | Q4_K_S | 0.88GB |
| Bakpia-V1-1.5B-Javanese.Q4_K.gguf | Q4_K | 0.92GB |
| Bakpia-V1-1.5B-Javanese.Q4_K_M.gguf | Q4_K_M | 0.92GB |
| Bakpia-V1-1.5B-Javanese.Q4_1.gguf | Q4_1 | 0.95GB |
| Bakpia-V1-1.5B-Javanese.Q5_0.gguf | Q5_0 | 1.02GB |
| Bakpia-V1-1.5B-Javanese.Q5_K_S.gguf | Q5_K_S | 1.02GB |
| Bakpia-V1-1.5B-Javanese.Q5_K.gguf | Q5_K | 1.05GB |
| Bakpia-V1-1.5B-Javanese.Q5_K_M.gguf | Q5_K_M | 1.05GB |
| Bakpia-V1-1.5B-Javanese.Q5_1.gguf | Q5_1 | 1.1GB |
| Bakpia-V1-1.5B-Javanese.Q6_K.gguf | Q6_K | 1.19GB |
| Bakpia-V1-1.5B-Javanese.Q8_0.gguf | Q8_0 | 1.53GB |
language:
This repository contains the fp16 version of Bakpia V1 1.5B.
| Version | Base Model | URL | Training |
|---|---|---|---|
| V1 0.5B | Qwen 2 0.5B Instruct | fp16 | Epoch = 1, Batch = 16*8, lr = 5e-5, linear schedule |
| V1 1.5B | Qwen 2 1.5B Instruct | fp16 | Epoch = 1, Batch = 16*8, lr = 5e-5, linear schedule |
| V1 9B | Gemma 2 9B Instruct | fp16/4bit | Batch size = 16*8, lr = 4e-5, linear schedule |
Training data is accessible here.
This is the first version of Bakpia.
✨ Training
✨ Features
from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
tokenizer = AutoTokenizer.from_pretrained("afrizalha/Bakpia-V1-1.5B-Javanese")
model = AutoModelForCausalLM.from_pretrained("afrizalha/Bakpia-V1-1.5B-Javanese")
model.to("cuda")
template = """<|im_start|>system
<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
"""
input = template.format(prompt="Kados pundi kulo saged nyinaoni Basa Jawa kanthi sae?")
input = tokenizer([input], return_tensors = "pt").to("cuda")
outputs = model.generate(**input, max_new_tokens = 1024, streamer= TextStreamer(tokenizer), temperature=.5, use_cache=True, do_sample=True)