hpcpony

quickstart: openai_gym AttributeError: 'EnvSpec' object has no attribute '_kwargs'. Did you mean: 'kwargs'?

2

tensorforce: 0.6.5 python: 3.10.4 ``` [user 0.6.5]$ python3 ../quick.py Traceback (most recent call last): File "/home/user/0.6.5/../quick.py", line 4, in environment = Environment.create( File "/opt/RL/Python_3.10.4_tensorforce/lib/python3.10/site-packages/tensorforce/environments/environment.py", line 204, in create return Environment.create(...

quickstart: openai_gym AttributeError: module 'gym.wrappers' has no attribute 'Monitor'

2

tensorforce: 0.6.5 ``` [user 0.6.5]$ python ../quick.py WARNING:root:Infinite min_value bound for state. Traceback (most recent call last): File "/home/user/0.6.5/../quick.py", line 24, in states = environment.reset() File "/opt/RL/Python_3.10.4_tensorforce/lib/python3.10/site-packages/tensorforce/environments/environment.py", line 529, in...

ch-fromhost --nvidia, security.selinux fails, 0.24?

5

charliecloud 0.24, anaconda3 2021.05 (python 3.8.8), CentOS 7.6. (non-root user). I haven't found a comprehensive guide for working with NVIDIA GPUs and charliecloud but I've managed to piece together enough...

Looks like convert-hf-to-gguf.py make the assumption that model part files are named "model-\*" when BLOOM (https://huggingface.co/bigscience/bloom/tree/main) uses the convention "model_\*". Looks like there's a similar issue with "pytorch_model-\*" vs. BLOOM's...

I tried starting with the conda install from the installation.md... ``` conda create -n hf conda activate hf conda install -c intel intel_extension_for_transformers . . . ``` ...but it's incomplete....

source build from release tar file?

1

I'd really like to build from source using one of the release tar files, but I'm having issues. Anybody know of any magic to make it possible? If you can't...

developer_document.md need elaboration on determining buffer sizes?

1

In the example for adding to gptneox_mem_req I see that n_layers comes from the num_hidden_layers in the config.json file, but where does the 512, 512, and 1024 come from? Maybe...

[Installation]: Build from source and flash-attention (0.7.2)

1

### Your current environment isolated system, can't provide environment. ### How you are installing vllm pip install . Is install from source using VLLM_FLASH_ATTN_SRC_DIR still supported? I don't see it...

### Your current environment N/A ### How you are installing vllm pip install . I was building vllm off-line with clones of CUTLASS and flash-attention. flash-attention (setup.py) does a "git...