You can deploy this model and start earning money today!

RichardErkhov/
hubentu_-_cmd2cwl_Llama-3.2-3B-gguf

Name: hubentu_-_cmd2cwl_Llama-3.2-3B-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

cmd2cwl_Llama-3.2-3B - GGUF

Model creator: https://huggingface.co/hubentu/
Original model: https://huggingface.co/hubentu/cmd2cwl_Llama-3.2-3B/

Name	Quant method	Size
cmd2cwl_Llama-3.2-3B.Q2_K.gguf	Q2_K	1.27GB
cmd2cwl_Llama-3.2-3B.IQ3_XS.gguf	IQ3_XS	1.38GB
cmd2cwl_Llama-3.2-3B.IQ3_S.gguf	IQ3_S	1.44GB
cmd2cwl_Llama-3.2-3B.Q3_K_S.gguf	Q3_K_S	1.44GB
cmd2cwl_Llama-3.2-3B.IQ3_M.gguf	IQ3_M	1.49GB
cmd2cwl_Llama-3.2-3B.Q3_K.gguf	Q3_K	1.57GB
cmd2cwl_Llama-3.2-3B.Q3_K_M.gguf	Q3_K_M	1.57GB
cmd2cwl_Llama-3.2-3B.Q3_K_L.gguf	Q3_K_L	1.69GB
cmd2cwl_Llama-3.2-3B.IQ4_XS.gguf	IQ4_XS	1.71GB
cmd2cwl_Llama-3.2-3B.Q4_0.gguf	Q4_0	1.79GB
cmd2cwl_Llama-3.2-3B.IQ4_NL.gguf	IQ4_NL	1.79GB
cmd2cwl_Llama-3.2-3B.Q4_K_S.gguf	Q4_K_S	1.8GB
cmd2cwl_Llama-3.2-3B.Q4_K.gguf	Q4_K	1.88GB
cmd2cwl_Llama-3.2-3B.Q4_K_M.gguf	Q4_K_M	1.88GB
cmd2cwl_Llama-3.2-3B.Q4_1.gguf	Q4_1	1.95GB
cmd2cwl_Llama-3.2-3B.Q5_0.gguf	Q5_0	2.11GB
cmd2cwl_Llama-3.2-3B.Q5_K_S.gguf	Q5_K_S	2.11GB
cmd2cwl_Llama-3.2-3B.Q5_K.gguf	Q5_K	2.16GB
cmd2cwl_Llama-3.2-3B.Q5_K_M.gguf	Q5_K_M	2.16GB
cmd2cwl_Llama-3.2-3B.Q5_1.gguf	Q5_1	2.28GB
cmd2cwl_Llama-3.2-3B.Q6_K.gguf	Q6_K	2.46GB
cmd2cwl_Llama-3.2-3B.Q8_0.gguf	Q8_0	3.19GB

Original model description:

base_model: unsloth/Llama-3.2-3B language:

en license: apache-2.0 tags:
text-generation-inference
transformers
unsloth
llama
trl
sft

Model Information

The cmd2cwl model is an instruction fine-tuned version of the unsloth/Llama-3.2-3B. This model has been trained on a custom dataset consisting of help documentation from various command-line tools and corresponding CWL (Common Workflow Language) scripts. Its purpose is to assist users in converting command-line tool documentation into clean and well-structured CWL scripts, enhancing automation and workflow reproducibility.

Example

Task

python

question = """
Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
Write a cwl script for md5sum with docker image alpine.

### Input:

    With no FILE, or when FILE is -, read standard input.

    -b, --binary         read in binary mode
    -c, --check          read MD5 sums from the FILEs and check them
    --tag            create a BSD-style checksum
    -t, --text           read in text mode (default)
    -z, --zero           end each output line with NUL, not newline,
    and disable file name escaping

    The following five options are useful only when verifying checksums:
    --ignore-missing  don't fail or report status for missing files
    --quiet          don't print OK for each successfully verified file
    --status         don't output anything, status code shows success
    --strict         exit non-zero for improperly formatted checksum lines
    -w, --warn           warn about improperly formatted checksum lines

    --help     display this help and exit
    --version  output version information and exit

    The sums are computed as described in RFC 1321.  When checking, the input
    should be a former output of this program.  The default mode is to print a
    line with checksum, a space, a character indicating input mode ('*' for binary,
    ' ' for text or where binary is insignificant), and name for each FILE.


### Response:
"""

Using unsloth

python

from unsloth import FastLanguageModel
from transformers import TextStreamer

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "hubentu/cmd2cwl_Llama-3.2-3B",
    load_in_4bit = False,
)
FastLanguageModel.for_inference(model)

inputs = tokenizer(
    [question],
    return_tensors = "pt").to("cuda")

text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer)

Using AutoModelForCausalLM

python

from transformers import AutoTokenizer, AutoModelForCausalLM
from transformers import TextStreamer

model = AutoModelForCausalLM.from_pretrained("hubentu/cmd2cwl_Llama-3.2-3B")
tokenizer = AutoTokenizer.from_pretrained("hubentu/cmd2cwl_Llama-3.2-3B")
model.to('cuda')

text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_length=8192)

Using generator

python

from transformers import pipeline
generator = pipeline('text-generation', model="checkpoints/cmd2cwl_Llama-3.2-3B", device='cuda')
resp = generator(question, max_length=8192)
print(resp[0]['generated_text'].split("### Response:\n")[-1])

Output

cwlVersion: v1.0
class: CommandLineTool
baseCommand:
- md5sum
requirements:
- class: DockerRequirement
  dockerPull: alpine:latest
label: md5sum
doc: Compute and check MD5 checksums
inputs:
  files:
    label: files
    doc: Input files
    type: File[]
    inputBinding:
      separate: true
outputs:
  md5:
    label: md5
    doc: MD5 checksums
    type: string[]
    outputBinding:
      glob: $(inputs.files.name)

Uploaded model

Developed by: hubentu
License: apache-2.0
Finetuned from model : unsloth/Llama-3.2-3B

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.

RichardErkhov/hubentu_-_cmd2cwl_Llama-3.2-3B-gguf

Original model description:

Model Information

Example

Task

Using unsloth

Using AutoModelForCausalLM

Using generator

Output

Uploaded model

RichardErkhov/
hubentu_-_cmd2cwl_Llama-3.2-3B-gguf