Instructions to use Tristan/pythia-70m_default with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Tristan/pythia-70m_default with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Tristan/pythia-70m_default")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Tristan/pythia-70m_default")
model = AutoModelForCausalLM.from_pretrained("Tristan/pythia-70m_default")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Tristan/pythia-70m_default with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Tristan/pythia-70m_default"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Tristan/pythia-70m_default",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Tristan/pythia-70m_default

SGLang

How to use Tristan/pythia-70m_default with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Tristan/pythia-70m_default" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Tristan/pythia-70m_default",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Tristan/pythia-70m_default" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Tristan/pythia-70m_default",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Tristan/pythia-70m_default with Docker Model Runner:
```
docker model run hf.co/Tristan/pythia-70m_default
```

Tristan commited on May 21, 2024

Commit

e689af1

verified ·

1 Parent(s): 36ae1c1

Training in progress, epoch 1

Browse files

Files changed (6) hide show

config.json +1 -1
eval_job_output.txt +8 -92
logs/events.out.tfevents.1716245528.sphinx2 +3 -0
model.safetensors +1 -1
train_job_output.txt +0 -0
training_args.bin +2 -2

config.json CHANGED Viewed

@@ -23,7 +23,7 @@
   "rotary_pct": 0.25,
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
-  "transformers_version": "4.39.3",
   "use_cache": true,
   "use_parallel_residual": true,
   "vocab_size": 50304

   "rotary_pct": 0.25,
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
+  "transformers_version": "4.40.2",
   "use_cache": true,
   "use_parallel_residual": true,
   "vocab_size": 50304

eval_job_output.txt CHANGED Viewed

@@ -1,4 +1,4 @@
-slurm submission log: 2024-05-11 22:52:02.402564
 created following sbatch script:
 ###############################
@@ -7,24 +7,24 @@ created following sbatch script:
 #SBATCH --account=nlp
 #SBATCH --cpus-per-task=16
-#SBATCH --dependency=afterok:
 #SBATCH --gres=gpu:1
-#SBATCH --job-name=tthrush-job-4888498
 #SBATCH --mem=60G
-#SBATCH --nodelist=sphinx2
 #SBATCH --open-mode=append
-#SBATCH --output=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/eval_job_output.txt
 #SBATCH --partition=sphinx
 #SBATCH --time=14-0
 # activate your desired anaconda environment
-. /nlp/scr/tthrush/miniconda3/etc/profile.d/conda.sh ; conda activate pretraining-coreset-selection
 # cd to working directory
 cd .
 # launch commands
-srun --unbuffered run_as_child_processes 'lm_eval --model hf --model_args pretrained=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default,revision=main,dtype=float16,trust_remote_code=True --tasks xnli_en,xnli_fr,sciq,piqa,lambada,arc_easy --device cuda --output_path /juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/perf'
 ###############################
@@ -34,91 +34,7 @@ submission to slurm complete!
 ###############################
 slurm submission output
-sbatch: error: Batch job submission failed: Job dependency problem
-###############################
-slurm submission log: 2024-05-11 22:53:20.065335
-created following sbatch script:
-###############################
-#!/bin/bash
-#SBATCH --account=nlp
-#SBATCH --cpus-per-task=16
-#SBATCH --dependency=afterok:7599822
-#SBATCH --gres=gpu:1
-#SBATCH --job-name=tthrush-job-2562954
-#SBATCH --mem=60G
-#SBATCH --nodelist=sphinx2
-#SBATCH --open-mode=append
-#SBATCH --output=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/eval_job_output.txt
-#SBATCH --partition=sphinx
-#SBATCH --time=14-0
-# activate your desired anaconda environment
-. /nlp/scr/tthrush/miniconda3/etc/profile.d/conda.sh ; conda activate pretraining-coreset-selection
-# cd to working directory
-cd .
-# launch commands
-srun --unbuffered run_as_child_processes 'lm_eval --model hf --model_args pretrained=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default,revision=main,dtype=float16,trust_remote_code=True --tasks xnli_en,xnli_fr,sciq,piqa,lambada,arc_easy --device cuda --output_path /juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/perf'
-###############################
-submission to slurm complete!
-###############################
-slurm submission output
-Submitted batch job 7599823
-###############################
-slurm submission log: 2024-05-11 23:09:48.287855
-created following sbatch script:
-###############################
-#!/bin/bash
-#SBATCH --account=nlp
-#SBATCH --cpus-per-task=16
-#SBATCH --dependency=afterok:7599868
-#SBATCH --gres=gpu:1
-#SBATCH --job-name=tthrush-job-4073620
-#SBATCH --mem=60G
-#SBATCH --nodelist=sphinx2
-#SBATCH --open-mode=append
-#SBATCH --output=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/eval_job_output.txt
-#SBATCH --partition=sphinx
-#SBATCH --time=14-0
-# activate your desired anaconda environment
-. /nlp/scr/tthrush/miniconda3/etc/profile.d/conda.sh ; conda activate pretraining-coreset-selection
-# cd to working directory
-cd .
-# launch commands
-srun --unbuffered run_as_child_processes 'lm_eval --model hf --model_args pretrained=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default,revision=main,dtype=float16,trust_remote_code=True --tasks xnli_en,xnli_fr,sciq,piqa,lambada,arc_easy --device cuda --output_path /juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/llms_5/pythia-70m_default/perf'
-###############################
-submission to slurm complete!
-###############################
-slurm submission output
-Submitted batch job 7599869

+slurm submission log: 2024-05-20 10:11:59.906526
 created following sbatch script:
 ###############################
 #SBATCH --account=nlp
 #SBATCH --cpus-per-task=16
+#SBATCH --dependency=afterok:7635790
 #SBATCH --gres=gpu:1
+#SBATCH --job-name=tthrush-job-1011450
 #SBATCH --mem=60G
+#SBATCH --nodelist=sphinx1
 #SBATCH --open-mode=append
+#SBATCH --output=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/70m_llms_more_data_test_normalized/pythia-70m_default_1/eval_job_output.txt
 #SBATCH --partition=sphinx
 #SBATCH --time=14-0
 # activate your desired anaconda environment
+. /nlp/scr/tthrush/miniconda3/envs/pretraining-coreset-selection/etc/profile.d/conda.sh ; conda activate pretraining-coreset-selection
 # cd to working directory
 cd .
 # launch commands
+srun --unbuffered run_as_child_processes 'lm_eval --model hf --model_args pretrained=/juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/70m_llms_more_data_test_normalized/pythia-70m_default_1,revision=main,dtype=float16,trust_remote_code=True --tasks xnli_en,xnli_fr,sciq,piqa,lambada,arc_easy,default --device cuda --output_path /juice5/scr5/tthrush/pretraining-coreset-selection/llm_pretraining/70m_llms_more_data_test_normalized/pythia-70m_default_1/perf'
 ###############################
 ###############################
 slurm submission output
+Submitted batch job 7635791

logs/events.out.tfevents.1716245528.sphinx2 ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:8b4de350dcd464831a416cf5a86a0b0f2b7fd0003923c3a64438cc31390713a6
+size 95455

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b951b996bf316cd5f1f805ff62c7a17ff38b1796cef1aa7ed93aab71b0e1ab3
 size 281715176

 version https://git-lfs.github.com/spec/v1
+oid sha256:cf7b47385f852fe85e4a2837781a49a0e236a5e632a260688f4ae5c333f57f3a
 size 281715176

train_job_output.txt CHANGED Viewed

The diff for this file is too large to render. See raw diff

training_args.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6cf3be7689dc496c68d9042ba7ef79c5166b2cf3d4a5717ebfc52a3a1503f499
-size 5112

 version https://git-lfs.github.com/spec/v1
+oid sha256:0433006cce1a247878d083c2265b5f8ae1fd4d1878e57fb6a9ba2f55d3ef8231
+size 5240