GithubMirror/modal-examples
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update trtllm_deepseek.py (#1349)

Browse Source 
* Update trtllm_deepseek.py

* bring back revision, comma

* ruff

---------

Co-authored-by: Charles Frye <cfrye59@gmail.com>
Co-authored-by: Charles Frye <charles@modal.com>
This commit is contained in:
ShivamB25
2025-08-31 09:08:02 +05:30
committed by GitHub
parent 60016b822e
commit cd75f2bee3
1 changed files with 3 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
misc/trtllm_deepseek.py
View File

@@ -91,8 +91,8 @@ hf_cache_vol = modal.Volume.from_name(f"{app_name}-hf-cache", create_if_missing=
HF_CACHE_PATH = Path("/hf_cache")
volumes = {HF_CACHE_PATH: hf_cache_vol}
MODEL_NAME = "nvidia/DeepSeek-R1-0528-FP4"
MODEL_REVISION = "91cfc7c35acd8ecfc769205989310208b8b81c9c" # in case repo updates!
MODEL_NAME = "nvidia/DeepSeek-R1-0528-FP4-v2"
MODEL_REVISION = "d12ff8db9876124d533b26bc24523c27907ce386" # in case repo updates!
MODELS_PATH = HF_CACHE_PATH / "models"
MODEL_PATH = MODELS_PATH / MODEL_NAME
@@ -118,9 +118,7 @@ def download_model():
MINUTES = 60 # seconds
tensorrt_image = (
tensorrt_image.run_commands(
"uv pip install --system --compile-bytecode hf-transfer==0.1.9 huggingface_hub==0.33.0"
)
tensorrt_image.uv_pip_install("hf-transfer==0.1.9", "huggingface_hub==0.33.0")
.env(
{
"HF_HUB_ENABLE_HF_TRANSFER": "1",