hzxllll commited on
Commit
feebe03
·
verified ·
1 Parent(s): ee20c02

Use the model name throughout the loading instructions

Browse files
Files changed (1) hide show
  1. README.md +3 -3
README.md CHANGED
@@ -23,7 +23,7 @@ The action expert first learns from reference trajectories while the base brain
23
  | `qwen/` | Jointly trained Qwen3.6-35B-A3B brain, tokenizer and processor. The brain weights are stored in 16 BF16 safetensors shards. |
24
  | `learnable_query/learnable_query_embeddings.pt` | The 16 learnable query embeddings. Their contextual representations condition the action expert on the brain’s movement request. |
25
  | `action_head/` | Continuous-action expert weights and architecture configuration. |
26
- | `report.json` | Identifies the training step and weight export for the checkpoint loader. |
27
  | `training_config.json` | Training settings associated with this checkpoint, used by the evaluation launcher. |
28
 
29
  The full download is about 70.4 GB. The expert also uses a separate frozen Qwen3.5-4B visual backbone. Supply its local path and the benchmark scenes and episode files when configuring the runtime.
@@ -35,7 +35,7 @@ Start from a checkout of the [ZeroVLN code repository](https://github.com/hzx122
35
  ```bash
36
  source ./scripts/env/environment.local.sh
37
  export PATH="$(dirname "$PYTHON"):$PATH"
38
- export INITIAL_CHECKPOINT="$RL_ARTIFACT_ROOT/checkpoints/global_step_000152"
39
 
40
  # If access is restricted, authenticate with an account that can read this repository.
41
  hf download hzxllll/ZeroVLN --local-dir "$INITIAL_CHECKPOINT"
@@ -47,7 +47,7 @@ export TRAINING_CONFIG="$INITIAL_CHECKPOINT/training_config.json"
47
  --output-dir "$RL_ARTIFACT_ROOT/navigation-console" --port 5002
48
  ```
49
 
50
- Keep the directory name `global_step_000152` so the loader can match it to the step recorded in `report.json`. The preparation command makes a local compatibility link for the learnable query weights and can be run again. Open the console on port 5002, select a scene and instruction, then run an episode. The [code Quick start](https://github.com/hzx122/ZeroVLN#quick-start) pins the exact Hub revision.
51
 
52
  ## Evaluation
53
 
 
23
  | `qwen/` | Jointly trained Qwen3.6-35B-A3B brain, tokenizer and processor. The brain weights are stored in 16 BF16 safetensors shards. |
24
  | `learnable_query/learnable_query_embeddings.pt` | The 16 learnable query embeddings. Their contextual representations condition the action expert on the brain’s movement request. |
25
  | `action_head/` | Continuous-action expert weights and architecture configuration. |
26
+ | `report.json` | Export metadata used to validate the model files. |
27
  | `training_config.json` | Training settings associated with this checkpoint, used by the evaluation launcher. |
28
 
29
  The full download is about 70.4 GB. The expert also uses a separate frozen Qwen3.5-4B visual backbone. Supply its local path and the benchmark scenes and episode files when configuring the runtime.
 
35
  ```bash
36
  source ./scripts/env/environment.local.sh
37
  export PATH="$(dirname "$PYTHON"):$PATH"
38
+ export INITIAL_CHECKPOINT="$RL_ARTIFACT_ROOT/checkpoints/ZeroVLN"
39
 
40
  # If access is restricted, authenticate with an account that can read this repository.
41
  hf download hzxllll/ZeroVLN --local-dir "$INITIAL_CHECKPOINT"
 
47
  --output-dir "$RL_ARTIFACT_ROOT/navigation-console" --port 5002
48
  ```
49
 
50
+ The preparation command makes a local compatibility link for the learnable query weights and can be run again. Open the console on port 5002, select a scene and instruction, then run an episode. The [code Quick start](https://github.com/hzx122/ZeroVLN#quick-start) pins the exact Hub revision.
51
 
52
  ## Evaluation
53