license: llama2
model_name: 13B Legerdemain L2
inference: false
model_creator: CalderaAI
model_link: https://huggingface.co/CalderaAI/13B-Legerdemain-L2
model_type: llama
quantized_by: TheBloke
base_model: CalderaAI/13B-Legerdemain-L2
TheBloke's LLM work is generously supported by a grant from andreessen horowitz (a16z)
13B Legerdemain L2 - GGML
- Model creator: CalderaAI
- Original model: 13B Legerdemain L2
Description
This repo contains GGML format model files for CalderaAI's 13B Legerdemain L2.
Important note regarding GGML files.
The GGML format has now been superseded by GGUF. As of August 21st 2023, llama.cpp no longer supports GGML models. Third party clients and libraries are expected to still support it for a time, but many may also drop support.
Please use the GGUF models instead.
About GGML
GGML files are for CPU + GPU inference using llama.cpp and libraries and UIs which support this format, such as:
- text-generation-webui, the most popular web UI. Supports NVidia CUDA GPU acceleration.
- KoboldCpp, a powerful GGML web UI with GPU acceleration on all platforms (CUDA and OpenCL). Especially good for story telling.
- LM Studio, a fully featured local GUI with GPU acceleration on both Windows (NVidia and AMD), and macOS.
- LoLLMS Web UI, a great web UI with CUDA GPU acceleration via the c_transformers backend.
- ctransformers, a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server.
- llama-cpp-python, a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
Repositories available
- GPTQ models for GPU inference, with multiple quantisation parameter options.
- 2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference
- 2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference (deprecated)
- CalderaAI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions
Prompt template: Alpaca
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
{prompt}
### Response:
Compatibility
These quantised GGML files are compatible with llama.cpp between June 6th (commit 2d43387) and August 21st 2023.
For support with latest llama.cpp, please use GGUF files instead.
The final llama.cpp commit with support for GGML was: dadbed99e65252d79f81101a392d0d6497b86caa
As of August 23rd 2023 they are still compatible with all UIs, libraries and utilities which use GGML. This may change in the future.
Explanation of the new k-quant methods
Click to see details
The new methods available are:
- GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
- GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
- GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
- GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
- GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
- GGML_TYPE_Q8_K - "type-0" 8-bit quantization. Only used for quantizing intermediate results. The difference to the existing Q8_0 is that the block size is 256. All 2-6 bit dot products are implemented for this quantization type.
Refer to the Provided Files table below to see what files use which methods, and how.
Provided files
| Name | Quant method | Bits | Size | Max RAM required | Use case |
|---|---|---|---|---|---|
| 13b-legerdemain-l2.ggmlv3.q2_K.bin | q2_K | 2 | 5.51 GB | 8.01 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.vw and feed_forward.w2 tensors, GGML_TYPE_Q2_K for the other tensors. |
| 13b-legerdemain-l2.ggmlv3.q3_K_S.bin | q3_K_S | 3 | 5.66 GB | 8.16 GB | New k-quant method. Uses GGML_TYPE_Q3_K for all tensors |
| 13b-legerdemain-l2.ggmlv3.q3_K_M.bin | q3_K_M | 3 | 6.31 GB | 8.81 GB | New k-quant method. Uses GGML_TYPE_Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
| 13b-legerdemain-l2.ggmlv3.q3_K_L.bin | q3_K_L | 3 | 6.93 GB | 9.43 GB | New k-quant method. Uses GGML_TYPE_Q5_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else GGML_TYPE_Q3_K |
| 13b-legerdemain-l2.ggmlv3.q4_0.bin | q4_0 | 4 | 7.37 GB | 9.87 GB | Original quant method, 4-bit. |
| 13b-legerdemain-l2.ggmlv3.q4_K_S.bin | q4_K_S | 4 | 7.37 GB | 9.87 GB | New k-quant method. Uses GGML_TYPE_Q4_K for all tensors |
| 13b-legerdemain-l2.ggmlv3.q4_K_M.bin | q4_K_M | 4 | 7.87 GB | 10.37 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q4_K |
| 13b-legerdemain-l2.ggmlv3.q4_1.bin | q4_1 | 4 | 8.17 GB | 10.67 GB | Original quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models. |
| 13b-legerdemain-l2.ggmlv3.q5_0.bin | q5_0 | 5 | 8.97 GB | 11.47 GB | Original quant method, 5-bit. Higher accuracy, higher resource usage and slower inference. |
| 13b-legerdemain-l2.ggmlv3.q5_K_S.bin | q5_K_S | 5 | 8.97 GB | 11.47 GB | New k-quant method. Uses GGML_TYPE_Q5_K for all tensors |
| 13b-legerdemain-l2.ggmlv3.q5_K_M.bin | q5_K_M | 5 | 9.23 GB | 11.73 GB | New k-quant method. Uses GGML_TYPE_Q6_K for half of the attention.wv and feed_forward.w2 tensors, else GGML_TYPE_Q5_K |
| 13b-legerdemain-l2.ggmlv3.q5_1.bin | q5_1 | 5 | 9.78 GB | 12.28 GB | Original quant method, 5-bit. Even higher accuracy, resource usage and slower inference. |
| 13b-legerdemain-l2.ggmlv3.q6_K.bin | q6_K | 6 | 10.68 GB | 13.18 GB | New k-quant method. Uses GGML_TYPE_Q8_K for all tensors - 6-bit quantization |
| 13b-legerdemain-l2.ggmlv3.q8_0.bin | q8_0 | 8 | 13.79 GB | 16.29 GB | Original quant method, 8-bit. Almost indistinguishable from float16. High resource use and slow. Not recommended for most users. |
Note: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
How to run in llama.cpp
Make sure you are using llama.cpp from commit dadbed99e65252d79f81101a392d0d6497b86caa or earlier.
For compatibility with latest llama.cpp, please use GGUF files instead.
./main -t 10 -ngl 32 -m 13b-legerdemain-l2.ggmlv3.q4_K_M.bin --color -c 2048 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\nWrite a story about llamas\n\n### Response:"
Change -t 10 to the number of physical CPU cores you have. For example if your system has 8 cores/16 threads, use -t 8.
Change -ngl 32 to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
Change -c 2048 to the desired sequence length for this model. For example, -c 4096 for a Llama 2 model. For models that use RoPE, add --rope-freq-base 10000 --rope-freq-scale 0.5 for doubled context, or --rope-freq-base 10000 --rope-freq-scale 0.25 for 4x context.
If you want to have a chat-style conversation, replace the -p <PROMPT> argument with -i -ins
For other parameters and how to use them, please refer to the llama.cpp documentation
How to run in text-generation-webui
Further instructions here: text-generation-webui/docs/llama.cpp.md.
Discord
For further support, and discussions on these models and AI in general, join us at:
Thanks, and how to contribute.
Thanks to the chirper.ai team!
I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.
If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.
Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
- Patreon: https://patreon.com/TheBlokeAI
- Ko-Fi: https://ko-fi.com/TheBlokeAI
Special thanks to: Aemon Algiz.
Patreon special mentions: Russ Johnson, J, alfie_i, Alex, NimbleBox.ai, Chadd, Mandus, Nikolai Manek, Ken Nordquist, ya boyyy, Illia Dulskyi, Viktor Bowallius, vamX, Iucharbius, zynix, Magnesian, Clay Pascal, Pierre Kircher, Enrico Ros, Tony Hughes, Elle, Andrey, knownsqashed, Deep Realms, Jerry Meng, Lone Striker, Derek Yates, Pyrater, Mesiah Bishop, James Bentley, Femi Adebogun, Brandon Frisco, SuperWojo, Alps Aficionado, Michael Dempsey, Vitor Caleffi, Will Dee, Edmond Seymore, usrbinkat, LangChain4j, Kacper Wikieł, Luke Pendergrass, John Detwiler, theTransient, Nathan LeClaire, Tiffany J. Kim, biorpg, Eugene Pentland, Stanislav Ovsiannikov, Fred von Graf, terasurfer, Kalila, Dan Guido, Nitin Borwankar, 阿明, Ai Maven, John Villwock, Gabriel Puliatti, Stephen Murray, Asp the Wyvern, danny, Chris Smitley, ReadyPlayerEmma, S_X, Daniel P. Andersen, Olakabola, Jeffrey Morgan, Imad Khwaja, Caitlyn Gatomon, webtim, Alicia Loh, Trenton Dambrowitz, Swaroop Kallakuri, Erik Bjäreholt, Leonard Tan, Spiking Neurons AB, Luke @flexchar, Ajan Kanaga, Thomas Belote, Deo Leter, RoA, Willem Michiel, transmissions 11, subjectnull, Matthew Berman, Joseph William Delisle, David Ziegler, Michael Davis, Johann-Peter Hartmann, Talal Aujan, senxiiz, Artur Olbinski, Rainer Wilmers, Spencer Kim, Fen Risland, Cap'n Zoog, Rishabh Srivastava, Michael Levine, Geoffrey Montalvo, Sean Connelly, Alexandros Triantafyllidis, Pieter, Gabriel Tamborski, Sam, Subspace Studios, Junyu Yang, Pedro Madruga, Vadim, Cory Kujawski, K, Raven Klaugh, Randy H, Mano Prime, Sebastain Graf, Space Cruiser
Thank you to all my generous patrons and donaters!
And thank you again to a16z for their generous grant.
Original model card: CalderaAI's 13B Legerdemain L2
13B-Legerdemain-L2
13B-Legerdemain-L2 is the first model merge of its kind in a series of LLaMaV2 models mixed using a custom script built in-house by CalderaAI called Model-REVOLVER. M-REVOLVER is also the first in a series of custom scripts based on the concept of mixtuning - not only does the end user have contol over which models are mixed and their percentages on a per-layer basis, we tackle the problem of overcomplexity that arises from such a level of control; this model is the first of its series.
The Model-REVOLVER Process Designed by CalderaAI
M-REVOLVER (Rapid Evolution Via Optimized-List Viewer Evaluated Response) Per-layer merging between parent models is a nebulous inexact science, and therefore impractical to most users despite the raw power it offers. We propose an entirely new approach that gives the user a clear looking glass into the impact vastly different layer merge configurations between selected parent models of their choice will have on the potential offspring model - especially its inherited behaviors. We've developed solution MK.1 - A cyclic random pattern search in place that determines all layer merge ratios, combines test models, infers prompt completions, and deletes a prototype after data collection is saved. When the cyclic system has completed its entire run, nothing is left but the telemetry collected along with the cycle and layer merge ratios from every single prototype merge. This data is then used to empower the user to choose which offspring is most fit to their desired outcome. This final step is only initiated when all necessary data has been aggregated from all assembled-tested-erased prototypes sampled in the search space.
From here, the user is provided five 300 token prompt completions from each and every offspring contender that was created and tested during the cyclic process. The user simply browses each prototype's series of responses and selects their desired outcome model by entering the cycle number associated with the prompt completions they feel best suits their vision. That model is then instantly repatriated into the official offspring of its parent models and tokenizer files found to be most relevant are instantly auto-copied from the parent model dir to the offspring.
That's it - the user instantly has a complete model based on the behavior they decided on, suggested from one of many potentials; all with their own unique trait inheritence thanks to layer merge auto randomization inside an ordered system. One more thing - the user not only selects how many cycles to run, the user can edit prompts.txt which the system reads as a single prompt - this means if the user desires to use any multiline instruct format to observe all potential model outcomes from instruct, or desires simply their own prompt, it's up to them.. simply works.
Link to GitHub for M-REVOLVER are at the end of the model card. More advanced MergeTech toolsets and merge techniques are currently under internal testing and development by Caldera.
13B-Legerdemain-L2 Use
13B-Legerdemain-L2 is capable of following Alpaca instructions however it seems far more receptive to the by-the-book method as seen here:
Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
{instruction}
### Response:
{New Line}
The primary model of choice for this model was a story-only model called Holodeck by KoboldAI. Traits preserved seem to be detailed descriptiveness, verbosity, and characters with personality. The two other models selected were 13B-Nous-Hermes by NousResearch and 13B-orca-8k-3319 by OpenAssistant. I began the process by providing an incredibly obscene prompt and simply ignored each and every guardrail or censorship laden prompt completion and accepted the offensive ones in turn - intent wasn't to be crass but trigger censorship parts of the network to test if it's possible to completely undermine them. Second pass with offspring model and Orca was a simple milquetoast prompt to gauge vocabulary, word flow, and intelligence as I selected the most fit in that category. Result model seems a bit of a curiosity - different samplers and even a different UI (as I went from TGUI to KoboldAI) seem to uncover different facets of behavior. Godlike preset with Alpaca Instruct in TGUI worked fine. In KoboldAI some tweaking was necessary to get the same experience. If you choose to test this model, have fun - it's got a mind of its own.
Model-REVOLVER Git: