初始化项目,由ModelHub XC社区提供模型

Model: ibm-granite/GneissWeb.7B_ablation_model_on_350B_FineWeb.Edu.seed3
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-04-19 08:24:02 +08:00
commit 297a1153ec
18 changed files with 147863 additions and 0 deletions

22
conversion_log.txt Normal file
View File

@@ -0,0 +1,22 @@
------------------------ Running Env: ------------------------
RUNTIME: 2025-02-03-13-03
CHECKPOINT_PATH: /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/checkpoints/step_335000_ckp
OUTPUT_PATH: /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/hf_model/step_335000_ckp
MODEL: llama2mod_starcoder_7b
TOKENIZER: /proj/data-eng/tokenizers/bigcode_starcoder
CONDA_INIT_PATH: /opt/share/miniconda/etc/profile.d/conda.sh
CONDA_ENV_PATH: /proj/data-eng/fsdp/train_env
ENV_FILE: /proj/data-eng/fsdp/env/train_v01.env
-------------------------------------------------------------------
== Converting checkpoint /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/checkpoints/step_335000_ckp ...
python fms_to_hf.py --model_variant llama2mod_starcoder_7b --compiled --load_path /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/checkpoints/step_335000_ckp/ --save_path /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/hf_model/step_335000_ckp --tokenizer_name_or_path /proj/data-eng/tokenizers/bigcode_starcoder
Initializing model...
Reading state dict from /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/checkpoints/step_335000_ckp/
Loading state dict into the model...
Converting to HF model..
Copying tokenizer...
Model converted to HF model, saving at /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/hf_model/step_335000_ckp
== MODEL CONVERSION `` IS DONE IN: 96(s) (log: /proj/data-eng/fsdp/experiments/Large_FW_Edu_350B_7B_S3/hf_model/step_335000_ckp/conversion_log.txt)