add some basic run.sh to facilitate testing recipes, fix some configs

recipes/gpt2/run.sh

@@ -0,0 +1,7 @@
+# naive script with commands from the readme
+# (useful to make sure the recipe still runs)
+
+eole convert HF --model_dir openai-community/gpt2 --output $EOLE_MODEL_DIR/openai_gpt2 --token $HF_TOKEN
+echo -e "The European Union was created in" > lm_input.txt
+eole predict -c inference.yaml
+eole tools eval_hellaswag -c inference.yaml

recipes/llama2/run.sh

@@ -0,0 +1,20 @@
+# naive script with commands from the readme
+# (useful to make sure the recipe still runs)
+
+eole convert HF --model_dir meta-llama/Llama-2-7b-chat-hf --output $EOLE_MODEL_DIR/llama2-7b-chat-hf --token $HF_TOKEN
+echo -e "<s>[INST] <<SYS>>
+You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.
+If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
+<</SYS>>
+What are some nice places to visit in France? [/INST]" | sed ':a;N;$!ba;s/\n/｟newline｠/g' > test_prompt.txt
+eole predict -c llama-inference.yaml -src test_prompt.txt -output test_output.txt
+eole predict -c llama-inference-tp-2gpu.yaml -src test_prompt.txt -output test_output.txt
+[ ! -d ./data ] && mkdir ./data
+# Alpaca
+wget -P ./data https://opennmt-models.s3.amazonaws.com/llama/alpaca_clean.txt
+# Vicuna
+wget -P ./data https://opennmt-models.s3.amazonaws.com/llama/sharegpt.txt
+# Open Assisstant
+wget -P ./data https://opennmt-models.s3.amazonaws.com/llama/osst1.flattened.txt
+eole train -c llama-finetune.yaml
+eole model lora --action merge --base_model ${EOLE_MODEL_DIR}/llama2-7b-chat-hf --lora_weights ./finetune/llama2-7b-chat-hf-finetune --output ./finetune/merged

recipes/llama3.1/llama-inference.yaml

@@ -5,7 +5,6 @@ model_path: "${EOLE_MODEL_DIR}/llama3.1-8b"
 seed: 42
 max_length: 256
 # max_length: 1
-gpu: 0
 batch_type: sents
 batch_size: 4
 world_size: 1

recipes/llama3/run.sh

@@ -0,0 +1,6 @@
+# naive script with commands from the readme
+# (useful to make sure the recipe still runs)
+
+eole convert HF --model_dir meta-llama/Meta-Llama-3-8B-Instruct --output $EOLE_MODEL_DIR/llama3-8b-instruct --token $HF_TOKEN
+echo -e "What are some nice places to visit in France?" | sed ':a;N;$!ba;s/\n/｟newline｠/g' > test_prompt.txt
+eole predict -c llama-inference.yaml -src test_prompt.txt -output test_output.txt

recipes/mistral/README.md

@@ -17,7 +17,7 @@ export HF_TOKEN=<your_hf_token>
 ### Download and convert model
 
 ```
-eole convert HF --model_dir TheBloke/Mistral-7B-Instruct-v0.2-AWQ --output ${EOLE_MODEL_DIR}/mistral-7b-instruct-v0.2-awq --token $HF_TOKEN
+eole convert HF --model_dir mistralai/Mistral-7B-v0.3 --output ${EOLE_MODEL_DIR}/mistral-7b-v0.3 --token $HF_TOKEN
 ```
 
 

recipes/mistral/mistral-7b-awq-gemm-inference.yaml

@@ -1,11 +1,11 @@
 transforms: [sentencepiece]
 transforms_configs:
  sentencepiece:
- src_subword_model: "$EOLE_MODEL_DIR/mistral-7b-v0.3/tokenizer.model"
- tgt_subword_model: "$EOLE_MODEL_DIR/mistral-7b-v0.3/tokenizer.model"
+ src_subword_model: "$EOLE_MODEL_DIR/mistral-7b-instruct-v0.2-awq/tokenizer.model"
+ tgt_subword_model: "$EOLE_MODEL_DIR/mistral-7b-instruct-v0.2-awq/tokenizer.model"
 
 # Model info
-model_path: "$EOLE_MODEL_DIR/mistral-7b-v0.3"
+model_path: "$EOLE_MODEL_DIR/mistral-7b-instruct-v0.2-awq"
 
 # Inference
 seed: 42
@@ -19,8 +19,7 @@ gpu_ranks: [0]
 # parallel_mode: "tensor_parallel"
 #quant_layers: ['gate_up_proj', 'down_proj', 'up_proj', 'linear_values', 'linear_query', 'linear_keys', 'final_linear']
 #quant_type: "bnb_NF4"
-# precision: fp16
-precision: fp16
+compute_dtype: fp16
 #top_k: 1
 #top_p: 0.6
 #temperature: 0.9

recipes/mistral/run.sh

@@ -0,0 +1,3 @@
+eole convert HF --model_dir mistralai/Mistral-7B-v0.3 --output ${EOLE_MODEL_DIR}/mistral-7b-v0.3 --token $HF_TOKEN
+echo -e "What are some nice places to visit in France?" | sed ':a;N;$!ba;s/\n/｟newline｠/g' > test_prompt.txt
+eole predict -c mistral-7b-awq-gemm-inference.yaml -src test_prompt.txt -output test_output.txt