|
|
--- |
|
|
library_name: transformers |
|
|
pipeline_tag: text-generation |
|
|
tags: |
|
|
- 192k context |
|
|
- reasoning |
|
|
- thinking |
|
|
- qwen3 |
|
|
- uncensored |
|
|
- llama-cpp |
|
|
- gguf-my-repo |
|
|
base_model: DavidAU/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored |
|
|
--- |
|
|
|
|
|
# Triangle104/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored-Q4_K_M-GGUF |
|
|
This model was converted to GGUF format from [`DavidAU/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored`](https://huggingface.co/DavidAU/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space. |
|
|
Refer to the [original model card](https://huggingface.co/DavidAU/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored) for more details on the model. |
|
|
|
|
|
--- |
|
|
This repo is for Goekdeniz-Guelmez's excellent |
|
|
"Josiefied-Qwen3-8B-abliterated-v1", modified from 32k (32768) context |
|
|
to 192 k (196608) context modified using YARN as per tech notes at Qwen |
|
|
repo. |
|
|
|
|
|
|
|
|
ORG model repo for this fine tune: |
|
|
|
|
|
|
|
|
[ https://huggingface.co/Goekdeniz-Guelmez/Josiefied-Qwen3-8B-abliterated-v1 ] |
|
|
|
|
|
|
|
|
Max context on this version is : 192k (196608) |
|
|
|
|
|
|
|
|
Suggest min context limit of : 8k to 16k for "thinking" / "output". |
|
|
|
|
|
|
|
|
This model can output 2k to over 13k. |
|
|
|
|
|
|
|
|
To improve long form output performance (especially creative): |
|
|
|
|
|
|
|
|
Temp 1+, 2+ or higher. |
|
|
Top k 100+ |
|
|
Rep pen 1.02-1.09 |
|
|
|
|
|
|
|
|
Use Jinja Template or CHATML template. |
|
|
|
|
|
|
|
|
Please refer the QWEN model card for details, benchmarks, how to use, settings, turning reasoning on/off/ system roles etc etc : |
|
|
|
|
|
|
|
|
[ https://huggingface.co/Qwen/Qwen3-8B ] |
|
|
|
|
|
--- |
|
|
## Use with llama.cpp |
|
|
Install llama.cpp through brew (works on Mac and Linux) |
|
|
|
|
|
```bash |
|
|
brew install llama.cpp |
|
|
|
|
|
``` |
|
|
Invoke the llama.cpp server or the CLI. |
|
|
|
|
|
### CLI: |
|
|
```bash |
|
|
llama-cli --hf-repo Triangle104/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored-Q4_K_M-GGUF --hf-file qwen3-8b-192k-context-6x-josiefied-uncensored-q4_k_m.gguf -p "The meaning to life and the universe is" |
|
|
``` |
|
|
|
|
|
### Server: |
|
|
```bash |
|
|
llama-server --hf-repo Triangle104/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored-Q4_K_M-GGUF --hf-file qwen3-8b-192k-context-6x-josiefied-uncensored-q4_k_m.gguf -c 2048 |
|
|
``` |
|
|
|
|
|
Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well. |
|
|
|
|
|
Step 1: Clone llama.cpp from GitHub. |
|
|
``` |
|
|
git clone https://github.com/ggerganov/llama.cpp |
|
|
``` |
|
|
|
|
|
Step 2: Move into the llama.cpp folder and build it with `LLAMA_CURL=1` flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux). |
|
|
``` |
|
|
cd llama.cpp && LLAMA_CURL=1 make |
|
|
``` |
|
|
|
|
|
Step 3: Run inference through the main binary. |
|
|
``` |
|
|
./llama-cli --hf-repo Triangle104/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored-Q4_K_M-GGUF --hf-file qwen3-8b-192k-context-6x-josiefied-uncensored-q4_k_m.gguf -p "The meaning to life and the universe is" |
|
|
``` |
|
|
or |
|
|
``` |
|
|
./llama-server --hf-repo Triangle104/Qwen3-8B-192k-Context-6X-Josiefied-Uncensored-Q4_K_M-GGUF --hf-file qwen3-8b-192k-context-6x-josiefied-uncensored-q4_k_m.gguf -c 2048 |
|
|
``` |
|
|
|
