metadata

base_model: speakleash/Bielik-4.5B-v3.0-Instruct
license: apache-2.0
model_creator: speakleash
model_name: Bielik-4.5B-v3.0-Instruct
quantized_by: Second State Inc.
language:
  - pl
library_name: transformers
inference:
  parameters:
    temperature: 0.4

Bielik-4.5B-v3.0-Instruct-GGUF

Original Model

speakleash/Bielik-4.5B-v3.0-Instruct

Run with LlamaEdge

LlamaEdge version: v0.18.3

Prompt template

Prompt type: chatml

Prompt string

<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant

Context size: 32000

Run as LlamaEdge service

wasmedge --dir .:. --nn-preload default:GGML:AUTO:Bielik-4.5B-v3.0-Instruct-Q5_K_M.gguf \
  llama-api-server.wasm \
  --model-name Bielik-4.5B-v3.0-Instruct \
  --prompt-template chatml \
  --ctx-size 32000

Run as LlamaEdge command app

wasmedge --dir .:. --nn-preload default:GGML:AUTO:Bielik-4.5B-v3.0-Instruct-Q5_K_M.gguf \
  llama-chat.wasm \
  --prompt-template chatml \
  --ctx-size 32000

Quantized GGUF Models

Name	Quant method	Bits	Size	Use case
Bielik-4.5B-v3.0-Instruct-Q2_K.gguf	Q2_K	2	1.77 GB	smallest, significant quality loss - not recommended for most purposes
Bielik-4.5B-v3.0-Instruct-Q3_K_L.gguf	Q3_K_L	3	2.50 GB	small, substantial quality loss
Bielik-4.5B-v3.0-Instruct-Q3_K_M.gguf	Q3_K_M	3	2.30 GB	very small, high quality loss
Bielik-4.5B-v3.0-Instruct-Q3_K_S.gguf	Q3_K_S	3	2.08 GB	very small, high quality loss
Bielik-4.5B-v3.0-Instruct-Q4_0.gguf	Q4_0	4	2.70 GB	legacy; small, very high quality loss - prefer using Q3_K_M
Bielik-4.5B-v3.0-Instruct-Q4_K_M.gguf	Q4_K_M	4	2.88 GB	medium, balanced quality - recommended
Bielik-4.5B-v3.0-Instruct-Q4_K_S.gguf	Q4_K_S	4	2.72 GB	small, greater quality loss
Bielik-4.5B-v3.0-Instruct-Q5_0.gguf	Q5_0	5	3.29 GB	legacy; medium, balanced quality - prefer using Q4_K_M
Bielik-4.5B-v3.0-Instruct-Q5_K_M.gguf	Q5_K_M	5	3.38 GB	large, very low quality loss - recommended
Bielik-4.5B-v3.0-Instruct-Q5_K_S.gguf	Q5_K_S	5	3.29 GB	large, low quality loss - recommended
Bielik-4.5B-v3.0-Instruct-Q6_K.gguf	Q6_K	6	3.91 GB	very large, extremely low quality loss
Bielik-4.5B-v3.0-Instruct-Q8_0.gguf	Q8_0	8	5.06 GB	very large, extremely low quality loss - not recommended
Bielik-4.5B-v3.0-Instruct-f16.gguf	f16	16	9.52 GB

Quantized with llama.cpp b5201