Skip to content

Commit

Permalink
Add files via upload
Browse files Browse the repository at this point in the history
  • Loading branch information
jeffmeloy authored Aug 22, 2023
1 parent 09a3866 commit d899f04
Show file tree
Hide file tree
Showing 2 changed files with 14 additions and 13 deletions.
1 change: 0 additions & 1 deletion get_py2dataset_params.py
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@
b. Write the default model configuration to the MODEL_CONFIG_FILE in the specified directory.
c. Write the default model configuration to the MODEL_CONFIG_FILE in the current working directory if the output_dir argument is not provided or invalid.
"""

import os
import json
import logging
Expand Down
26 changes: 14 additions & 12 deletions py2dataset_model_config.yaml
Original file line number Diff line number Diff line change
@@ -1,13 +1,13 @@
prompt_template: "You are a genius mathematician and expert Python programmer. Please review the provided Python code context and use your intelligence and expertise to provide the best answer to the question below, which will be used to train people and AI models.
prompt_template: "Assuming the role of a master mathematician and expert Python developer, please analyze the given Python code. Provide a clear, concise response to the question and explain how the code implements its mathematical and logical functions, making it understandable for both humans and AI models.
\n### Instruction:\nGiven this context:\n'{context}'\nAnswer the following question and provide your reasoning step by step: {query}\n### Response:"
inference_model:
model_import_path: "ctransformers.AutoModelForCausalLM"
model_inference_function: "from_pretrained"
model_params:
## USABLE OUTPUT WITH CURRENT PROMPT TEMPLATE
#model_path: "TheBloke/WizardCoder-Guanaco-15B-V1.1-GGML"
#model_type: "gpt_bigcode"
#local_files_only: false
model_path: "TheBloke/WizardCoder-Guanaco-15B-V1.1-GGML"
model_type: "gpt_bigcode"
local_files_only: false
#model_path: "TheBloke/Starcoderplus-Guanaco-GPT4-15B-V1.0-GGML"
#model_type: "gpt_bigcode"
#local_files_only: false
Expand All @@ -17,9 +17,9 @@ inference_model:
#model_path: "./models/wizardcoder-guanaco-15b-v1.1.ggmlv1.q4_0.bin"
#model_type: "gpt_bigcode"
#local_files_only: true
model_path: "./models/starcoderplus-guanaco-gpt4.ggmlv1.q4_0.bin"
model_type: "gpt_bigcode"
local_files_only: true
#model_path: "./models/starcoderplus-guanaco-gpt4.ggmlv1.q4_0.bin"
#model_type: "gpt_bigcode"
#local_files_only: true
#model_path: "./models/octocoder.ggmlv1.q4_1.bin"
#model_type: "gpt_bigcode"
#local_files_only: true
Expand All @@ -30,9 +30,11 @@ inference_model:
#model_path: "TheBloke/stablecode-instruct-alpha-3b-GGML"
#model_type: "gpt_neox"
#local_files_only: false
lib: "avx2"
threads: 30
batch_size: 32
## MODEL CONFIGURATION PARAMETERS
#avx2 and gpu_layers are not compatible with ctransformers
#lib: "avx2"
threads: 16
batch_size: 256
max_new_tokens: 2048
gpu_layers: 0
reset: true
gpu_layers: 32
reset: false

0 comments on commit d899f04

Please sign in to comment.