File size: 4,145 Bytes
f1e4f50 a3640b9 f1e4f50 2c9e240 f1e4f50 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 |
---
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](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|