adding exercise

Author: jayant-yadav

docs/DL_exercises.md

@@ -47,10 +47,10 @@
 
 ???+ question "Submit a Slurm job"
 
-    - Close the [cifar10 resnet repository](https://github.com/akamaster/pytorch_resnet_cifar10?tab=readme-ov-file) and edit the run.sh by adding appropriate slurm sbatch commands.
+    - Submit `sentiment_analysis.py` with appropriate flags for its slurm job.
 
     ??? tip "Answer"
-        - edit a file using you prefered editor, named `my_bio_worksflow.sh`, for example, with the content
+        - edit a file using you prefered editor, named `sentiment_analysis_batch.sh`, for example, with the content
         
         ```bash
             #!/bin/bash -l
@@ -59,33 +59,53 @@
             #SBATCH -p node
             #SBATCH -N 1
             #SBATCH -t 01:00:00
-            #SBATCH -J cifar_demo
+            #SBATCH -J sentiment_analysis
             #SBATCH -M snowy
             #SBATCH --gres=gpu:1
 
-            module load python_ML_packages/3.9.5-gpu
+            source .....
 
             python -c "import torch; print(torch.__version__); print(torch.version.cuda); print(torch.cuda.get_device_properties(0)); print(torch.randn(1).cuda())"
 
-            #for model in resnet20 resnet32 resnet44 resnet56 resnet110 resnet1202
-            for model in resnet20 resnet110
-            do
-                echo "python -u trainer.py  --arch=$model  --save-dir=save_$model |& tee -a log_$model"
-                python -u trainer.py  --arch=$model  --save-dir=save_$model |& tee -a log_$model
-            done
+            echo "running sentiment_analysis.py"
 
+            python .....sentiment_analysis.py
         ```
 
-        - make the job script executable
+        - make the job script executable, if not already.
         ```bash
-        $ chmod a+x run.sh
+        $ chmod a+x sentiment_analysis_batch.sh
         ```
         
         - submit the job
         ```bash
-        $ sbatch run.sh
+        $ sbatch sentiment_analysis_batch.sh
         ```
-        
+
+    - Similarly run `sentiment_analysis.ipynb` jupyter notebook. Add matplotblib and seaborn to your environment. Start a jupyter server on snowy compute node and then tunnel to the host from your local.
+
+        ??? tip "Answer"
+                
+                * Install matplotlib and seaborn and start an interactive snowy session:  
+
+                ```console
+                source torch_env/bin/acivate
+                pip install matplotlib seaborn
+                interactive -A uppmax2025-3-5 -M snowy -p node -N 1 -t 1:01:00 --gres=gpu:1
+                source torch_env/bin/acivate
+                jupyter notebook --ip 0.0.0.0 --no-browser
+                ```
+                * Then tunnel:  
+                `ssh -L 8888:s123:8888 [email protected]`
+                
+                * Copy the localhost url and paste it in your browser
+
+
+### Cache management
+
+* By default HF will install the models and temp files to yur `$HOME` folde which is rather limited to 32 GB and 300k files. 
+* To avoid that, you can set `HF_HOME` and `HF_HUB_CACHE` variables to your project folder. Follow instructions on https://huggingface.co/docs/transformers/en/installation?cpu-only=PyTorch#cache-directory. 
+
 <!-- ## Doing installations