ccore commited on
Commit
15f353a
·
verified ·
1 Parent(s): 521105a

Update October 27 Q4 and Q8 checkpoints and Rome comparison

Browse files

Refresh both GGUF downloads, describe improved one-shot scene generation with unchanged parameters, and announce the October 28 FP16, dataset, and technical report release.

.gitattributes CHANGED
@@ -46,3 +46,4 @@ images/dragonrider-aerie.png filter=lfs diff=lfs merge=lfs -text
46
  Qwen3.5-4B-3D-Coder-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
47
  Qwen3.5-4B-3D-Coder-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
48
  images/dif.png filter=lfs diff=lfs merge=lfs -text
 
 
46
  Qwen3.5-4B-3D-Coder-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
47
  Qwen3.5-4B-3D-Coder-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
48
  images/dif.png filter=lfs diff=lfs merge=lfs -text
49
+ images/roma3.png filter=lfs diff=lfs merge=lfs -text
Qwen3.5-4B-3D-Coder-Q4_K_M.gguf CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ca7442573cdede7b62e80eabc1d78dfa1e0fade28058e414042ef714adadb36b
3
  size 2708804096
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32c37c83d3ca055d30aefa679f1a1670d3e2eb889ee7cc805792d53f219cfae8
3
  size 2708804096
Qwen3.5-4B-3D-Coder-Q8_0.gguf CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fee6c198e151d8ba4b032c98e44d810a0a2dc8e947ee9f9f933137a3a5528b91
3
  size 4482402816
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c30b02629456aa2bc32e522de0f58eb1aafd8518a7634128fed67d1486fc2a9a
3
  size 4482402816
README.md CHANGED
@@ -20,29 +20,25 @@ This is an early preview of a 4B model I am still training, with a particular fo
20
 
21
  This is an **independent fine-tune** by [CCORE](https://huggingface.co/ccore), based on [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B). **It is not an official Qwen release.**
22
 
23
- ## Update — October 26, 2026
24
 
25
- Both **Q4_K_M** and **Q8_0** have been updated. If you downloaded an earlier version, please download the files again to get this update.
26
 
27
- - Fixed the **layer-related issue** reported by some users.
28
- - In my latest tests, this update appears to produce more successful **one-shot scene generations**.
29
 
30
- The image below shows one example comparison from my tests.
31
 
32
- [![Example comparison of two voxel-style Rome scenes from testing](images/dif.png)](https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder/blob/main/images/dif.png)
33
 
34
- ## Training and release schedule
35
 
36
- Training is expected to finish on **October 28, 2026, at 07:30 UTC**.
37
 
38
- On **October 26 and 27, 2026**, I will update the **Q8_0 GGUF** to the latest available training checkpoint.
39
 
40
- After training finishes, I will publish:
41
 
42
- - The **full FP16 model weights**.
43
- - The **full set of quantized versions**.
44
- - The **training dataset**, so others can use it to train or fine-tune other models.
45
- - An **evaluation report with a variety of prompts**, showing how the model performed on them throughout training.
46
 
47
  ## Early examples
48
 
 
20
 
21
  This is an **independent fine-tune** by [CCORE](https://huggingface.co/ccore), based on [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B). **It is not an official Qwen release.**
22
 
23
+ ## October 27 checkpoint update
24
 
25
+ Both **Q4_K_M** and **Q8_0** now contain the updated October 27 checkpoint. If you downloaded an earlier version, please download the files again to get this update.
26
 
27
+ In my latest tests, **one-shot success has improved substantially**, along with **overall scene quality**. Some scenes can still require another attempt or a correction.
 
28
 
29
+ **Generation parameters remain unchanged** for both quantizations. Use the settings listed below. The earlier fix for the layer-related issue reported by some users is also included.
30
 
31
+ The comparison below shows the Rome scene across the **October 25, 26, and 27** previews.
32
 
33
+ [![Rome scene progression across the October 25, 26, and 27 previews](images/roma3.png)](https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder/blob/main/images/roma3.png)
34
 
35
+ ## Release plans — October 28, 2026
36
 
37
+ **Today, October 28, 2026**, I will release the **full FP16 model weights**, the **training dataset**, and a **technical report**.
38
 
39
+ The report will include a variety of prompts and their results across training checkpoints, showing how the model's behavior and scene quality evolved throughout training.
40
 
41
+ The **full set of quantized versions** is also planned.
 
 
 
42
 
43
  ## Early examples
44
 
images/{dif.png → roma3.png} RENAMED
File without changes