mjf-su commited on
Commit
0f62cd3
·
1 Parent(s): ed8d2cf

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +14 -0
README.md CHANGED
@@ -7,6 +7,8 @@ tags:
7
  - sft
8
  - trl
9
  licence: license
 
 
10
  ---
11
 
12
  # Model Card for test-full-pipeline
@@ -14,6 +16,18 @@ licence: license
14
  This model is a fine-tuned version of [Qwen/Qwen3-VL-4B-Thinking](https://huggingface.co/Qwen/Qwen3-VL-4B-Thinking).
15
  It has been trained using [TRL](https://github.com/huggingface/trl).
16
 
 
 
 
 
 
 
 
 
 
 
 
 
17
  ## Quick start
18
 
19
  ```python
 
7
  - sft
8
  - trl
9
  licence: license
10
+ datasets:
11
+ - tom-jerry-123/Physical-AI-AV-US
12
  ---
13
 
14
  # Model Card for test-full-pipeline
 
16
  This model is a fine-tuned version of [Qwen/Qwen3-VL-4B-Thinking](https://huggingface.co/Qwen/Qwen3-VL-4B-Thinking).
17
  It has been trained using [TRL](https://github.com/huggingface/trl).
18
 
19
+ ## Architecture
20
+
21
+ This model is a **base vision-language architecture** that maps images directly to vehicle waypoints. It takes a forward-facing camera image as input and outputs a sequence of future 2-D waypoints `[x, y]` in the ego-vehicle frame.
22
+
23
+ **Egomotion context:** The model observes 2 seconds of past ego-vehicle motion, sampled at a 0.25 s period (9 historical poses including the current timestep).
24
+
25
+ **Prediction horizon:** The model predicts 6 seconds into the future, also at a 0.25 s period (24 future waypoints).
26
+
27
+ ## Training data
28
+
29
+ This model was trained on **shards 0–19** (the first 20 shards) of the [tom-jerry-123/Physical-AI-AV-US](https://huggingface.co/datasets/tom-jerry-123/Physical-AI-AV-US) dataset on Hugging Face.
30
+
31
  ## Quick start
32
 
33
  ```python