About the test results of charades_sta.

[jiangerxiaozhao2]

Great work!
I got strange results. Could you help me?

../outputs/charades_sta-recursive_grounding-4_turns.jsonl num examples: 3720
turns: 1
mean iou: 0.2698
[email protected]/0.5/0.7: 0.3430/0.0022/0.0000
turns: 2
mean iou: 0.2698
[email protected]/0.5/0.7: 0.3430/0.0022/0.0000
turns: 3
mean iou: 0.2698
[email protected]/0.5/0.7: 0.3430/0.0022/0.0000
turns: 4
mean iou: 0.2698
[email protected]/0.5/0.7: 0.3430/0.0022/0.0000

The log file is:
charades_sta-recursive_grounding-4_turns.log

Only change to current code is

HawkEye/models/blip2/blip2.py

Line 28 in 635b1c5

tokenizer = BertTokenizer.from_pretrained(self.bert_path, truncation_side=truncation_side, local_files_only=True)

local_files_only=True->False for download necessary files

[yellow-binary-tree]

Yea, the problem is the model is always predicting "Throughout the entire video." at the first turn and terminates the recursive grounding process.

Why is there a CUDA out of memory error traceback in your log in line 102-119? Maybe the model is not correctly loaded?

[jiangerxiaozhao2]

I ran the code again and there was no "out of memory" error, but the result was the same.
charades_sta-recursive_grounding-4_turns.log

After debugging the program, I found that the prediction for each video is “llm_message = '('”.

[yellow-binary-tree]

Here is my recursive grounding log:
charades_sta-recursive_grounding-4_turns.log

I compared my recursive grounding log (which works as expected) with yours. The main difference is the bos_token and eos_token used:

For hawkeye model we use 0 as bos_token_id and 1 for eos_token_id. However, in your recursive log you used 1 as bos_token_id and 2 for eos_token _id. This may cause the model to generate nothing as it wrongly identifies the bos_token as the eos_token. You can try to reset the token_ids in LlamaConfig manually by adding following code after this line and see if it works:

HawkEye/models/hawkeye_it.py

Line 112 in 635b1c5

llama_config = LlamaConfig.from_pretrained(llama_model_path)

llama_config.bos_token_id = 0
llama_config.eos_token_id = 1

As for how this problem arises, all special tokens of Hawkeye were inherited from vicuna-v0, which were further inherited from LLaMA (v1). I can vaguely recall that when llama was first released in early 2023, its special token was once inconsistent in different released versions and caused some confusion at that time.

Please let me know if this solves your problem so I can fix this bug.

[jiangerxiaozhao2]

I tried, but it didn't work.
Fix two bugs:
llama_config.bos_token_id = 0
llama_config.eos_token_id = 1
And fix other two diffenrences:
llama_config.pad_token_id = -1
"transformers_version": "4.35.2"
The new log is:
charades_sta-recursive_grounding-4_turns.log