Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pipeline_tag: text-generation
|
| 6 |
+
datasets:
|
| 7 |
+
- Delta-Vector/Hydrus-Preview-Tulu-3-SFT-Mix
|
| 8 |
+
base_model:
|
| 9 |
+
- arcee-ai/GLM-4-32B-Base-32K
|
| 10 |
+
library_name: transformers
|
| 11 |
+
tags:
|
| 12 |
+
- instruct
|
| 13 |
+
- code
|
| 14 |
+
- chemistry
|
| 15 |
+
- GLM
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
<div align="center">
|
| 19 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/66c26b6fb01b19d8c3c2467b/j9LpDp1Wup-m-IJ_XMh23.png" alt="Image" width="400" />
|
| 20 |
+
<br>
|
| 21 |
+
<small><em>Promise I will never go blonde like Kanye</em></small>
|
| 22 |
+
</div>
|
| 23 |
+
|
| 24 |
+
---
|
| 25 |
+
# Overview
|
| 26 |
+
|
| 27 |
+
Didn't really have any cool README ideas for this so we're just going with just whatever song i'm listening to rn and it happened to be `Baby i'm bleeding`
|
| 28 |
+
|
| 29 |
+
Nevertheless, This is a finetune from the 32K context extended (or fixed?) Arcee GLM4 base - Trained shrimply with just the Tulu-SFT-Mixture *but* I removed Safety alignment examples. Came out pretty well, It uses chatML due to the GLM4(and other formats like it, Such as Dan-chat) giving me a headache. It's a decently competant assistant although I haven't done any testing on how well the model performs at longer-contexts, nor have i done any RL afterwards to fix up it's edges.
|
| 30 |
+
|
| 31 |
+
Think it should be a decent base for any future finetunes, I felt that GLM4 really wasn't given the proper time of day and it's a way better base then any Qwen3 model.
|
| 32 |
+
|
| 33 |
+
# Quants
|
| 34 |
+
|
| 35 |
+
GGUF: https://huggingface.co/mradermacher/GLM-Tulu-ChatML-GGUF
|
| 36 |
+
|
| 37 |
+
Imatrix GGUF: https://huggingface.co/mradermacher/GLM-Tulu-ChatML-i1-GGUF
|
| 38 |
+
|
| 39 |
+
# Prompting
|
| 40 |
+
|
| 41 |
+
The model was trained with ChatML formatting
|
| 42 |
+
|
| 43 |
+
```
|
| 44 |
+
"""<|im_start|>system
|
| 45 |
+
system prompt<|im_end|>
|
| 46 |
+
<|im_start|>user
|
| 47 |
+
Hi there!<|im_end|>
|
| 48 |
+
<|im_start|>assistant
|
| 49 |
+
Nice to meet you!<|im_end|>
|
| 50 |
+
<|im_start|>user
|
| 51 |
+
Can I ask a question?<|im_end|>
|
| 52 |
+
<|im_start|>assistant
|
| 53 |
+
"""
|
| 54 |
+
```
|
| 55 |
+
|
| 56 |
+
# Configs
|
| 57 |
+
|
| 58 |
+
WandB : https://wandb.ai/new-eden/Training-A100/runs/05kktve8?nw=nwuserdeltavector
|
| 59 |
+
|
| 60 |
+
This train took 15 hours on 8xB200s provided by Deepinfra and Cognitive Computations, Config is linked in the WandB
|
| 61 |
+
|
| 62 |
+
# Credits
|
| 63 |
+
|
| 64 |
+
Thank you to Lucy, Auri, NyxKrage, Creators of the Tulu-SFT-Mix and everyone at Anthracite & Allura
|
| 65 |
+
|