bartowski's picture
Update README.md
aa1332c verified
metadata
base_model: Writer/Palmyra-Med-70B-32K
language:
  - en
license: other
license_name: writer-open-model-license
license_link: https://writer.com/legal/open-model-license/
pipeline_tag: text-generation
tags:
  - instruct
  - finetune
  - DPO
  - distillation
  - heathcare
  - medical
  - clinical
  - med
  - lifescience
  - Pharmaceutical
  - Pharma
  - Enterprise LLM
  - Enterprise
  - Enterprise ready
quantized_by: bartowski
extra_gated_prompt: >-
  By clicking "Agree", you agree to the [License
  Agreement](https://writer.com/legal/open-model-license/) and acknowledge
  Writer's [Privacy Policy](https://writer.com/legal/acceptable-use/).
extra_gated_fields:
  Name: text
  Email: text
  Organization or Affiliation: text
  Receive email updates and promotions on Writer products, services, and research?:
    type: select
    options:
      - 'Yes'
      - 'No'
  I acknowledge that this model is for non-commercial use only unless I acquire a separate license from Writer: checkbox
widget:
  - example_title: Palmyra-Med-70B-32k
    messages:
      - role: system
        content: >-
          You are a highly knowledgeable and experienced expert in the
          healthcare and biomedical field, possessing extensive medical
          knowledge and practical expertise.
      - role: user
        content: >-
          Does danzhi Xiaoyao San ameliorate depressive-like behavior by
          shifting toward serotonin via the downregulation of hippocampal
          indoleamine 2,3-dioxygenase?
    output:
      text: >-
        Danzhi Xiaoyao San (DXS) exhibited significant antidepressant-like
        effects in rats subjected to chronic unpredictable mild stress (CUMS), a
        commonly used model for inducing depressive-like behavior. The
        administration of DXS led to a notable reduction in the activity of
        indoleamine 2,3-dioxygenase (IDO) in the hippocampus. This reduction in
        IDO activity is crucial because IDO is an enzyme that catalyzes the
        degradation of tryptophan into kynurenine (Kyn), thereby diverting
        tryptophan away from the serotonin (5-HT) synthesis pathway. By
        downregulating IDO, DXS effectively shifts the balance of tryptophan
        metabolism towards increased serotonin production, thereby enhancing the
        availability of serotonin. This shift in the Kyn/5-HT pathway favoring
        serotonin is likely a key mechanism underlying the antidepressant
        effects observed with DXS treatment. Enhanced serotonin levels in the
        hippocampus are associated with improved mood and reduction of
        depressive symptoms, suggesting that DXS could be a promising
        therapeutic agent for managing depression.
model-index:
  - name: Palmyra-Med-70B-32k
    results: []

Llamacpp imatrix Quantizations of Palmyra-Med-70B-32K

Using llama.cpp release b3496 for quantization.

Original model: https://huggingface.co/Writer/Palmyra-Med-70B-32K

All quants made using imatrix option with dataset from here

Run them in LM Studio

Prompt format

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

Download a file (not the whole branch) from below:

Filename Quant type File Size Split Description
Palmyra-Med-70B-32K-Q8_0.gguf Q8_0 74.98GB true Extremely high quality, generally unneeded but max available quant.
Palmyra-Med-70B-32K-Q6_K.gguf Q6_K 57.89GB true Very high quality, near perfect, recommended.
Palmyra-Med-70B-32K-Q5_K_M.gguf Q5_K_M 49.95GB true High quality, recommended.
Palmyra-Med-70B-32K-Q5_K_S.gguf Q5_K_S 48.66GB false High quality, recommended.
Palmyra-Med-70B-32K-Q4_K_L.gguf Q4_K_L 43.30GB false Uses Q8_0 for embed and output weights. Good quality, recommended.
Palmyra-Med-70B-32K-Q4_K_M.gguf Q4_K_M 42.52GB false Good quality, default size for must use cases, recommended.
Palmyra-Med-70B-32K-Q4_K_S.gguf Q4_K_S 40.35GB false Slightly lower quality with more space savings, recommended.
Palmyra-Med-70B-32K-Q3_K_XL.gguf Q3_K_XL 38.06GB false Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
Palmyra-Med-70B-32K-IQ4_XS.gguf IQ4_XS 37.90GB false Decent quality, smaller than Q4_K_S with similar performance, recommended.
Palmyra-Med-70B-32K-Q3_K_L.gguf Q3_K_L 37.14GB false Lower quality but usable, good for low RAM availability.
Palmyra-Med-70B-32K-Q3_K_M.gguf Q3_K_M 34.27GB false Low quality.
Palmyra-Med-70B-32K-IQ3_M.gguf IQ3_M 31.94GB false Medium-low quality, new method with decent performance comparable to Q3_K_M.
Palmyra-Med-70B-32K-Q3_K_S.gguf Q3_K_S 30.91GB false Low quality, not recommended.
Palmyra-Med-70B-32K-IQ3_XS.gguf IQ3_XS 29.31GB false Lower quality, new method with decent performance, slightly better than Q3_K_S.
Palmyra-Med-70B-32K-Q2_K_L.gguf Q2_K_L 27.40GB false Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
Palmyra-Med-70B-32K-Q2_K.gguf Q2_K 26.38GB false Very low quality but surprisingly usable.
Palmyra-Med-70B-32K-IQ2_M.gguf IQ2_M 24.12GB false Relatively low quality, uses SOTA techniques to be surprisingly usable.
Palmyra-Med-70B-32K-IQ2_XS.gguf IQ2_XS 21.14GB false Low quality, uses SOTA techniques to be usable.
Palmyra-Med-70B-32K-IQ2_XXS.gguf IQ2_XXS 19.10GB false Very low quality, uses SOTA techniques to be usable.

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

Some say that this improves the quality, others don't notice any difference. If you use these models PLEASE COMMENT with your findings. I would like feedback that these are actually used and useful so I don't keep uploading quants no one is using.

Thanks!

Credits

Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset

Thank you ZeroWw for the inspiration to experiment with embed/output

Downloading using huggingface-cli

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download bartowski/Palmyra-Med-70B-32K-GGUF --include "Palmyra-Med-70B-32K-Q4_K_M.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download bartowski/Palmyra-Med-70B-32K-GGUF --include "Palmyra-Med-70B-32K-Q8_0/*" --local-dir ./

You can either specify a new local-dir (Palmyra-Med-70B-32K-Q8_0) or download them all in place (./)

Which file should I choose?

A great write up with charts showing various performances is provided by Artefact2 here

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.

If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.

Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.

If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.

If you want to get more into the weeds, you can check out this extremely useful feature chart:

llama.cpp feature matrix

But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.

These I-quants can also be used on CPU and Apple Metal, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.

The I-quants are not compatible with Vulcan, which is also AMD, so if you have an AMD card double check if you're using the rocBLAS build or the Vulcan build. At the time of writing this, LM Studio has a preview with ROCm support, and other inference engines have specific builds for ROCm.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski