Spaces:
Runtime error
Runtime error
updated interface and error handling
Browse files
app.py
CHANGED
|
@@ -8,6 +8,8 @@ import gradio as gr
|
|
| 8 |
from huggingface_hub import HfApi
|
| 9 |
from huggingface_hub import ModelCard
|
| 10 |
|
|
|
|
|
|
|
| 11 |
from textwrap import dedent
|
| 12 |
|
| 13 |
HF_PATH = "https://huggingface.co/"
|
|
@@ -69,8 +71,14 @@ QUANTIZATIONS = ["q0f16",
|
|
| 69 |
"q4f16_awq"]
|
| 70 |
|
| 71 |
def button_click(hf_model_id, conv_template, quantization, oauth_token: gr.OAuthToken | None):
|
| 72 |
-
if
|
| 73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
|
| 75 |
api = HfApi(token=oauth_token.token)
|
| 76 |
model_dir_name = hf_model_id.split("/")[1]
|
|
@@ -81,13 +89,17 @@ def button_click(hf_model_id, conv_template, quantization, oauth_token: gr.OAuth
|
|
| 81 |
|
| 82 |
api.snapshot_download(repo_id=hf_model_id, local_dir=f"./dist/models/{model_dir_name}")
|
| 83 |
|
| 84 |
-
|
| 85 |
" --quantization " + quantization + \
|
| 86 |
-
" -o dist/" + mlc_model_name)
|
|
|
|
|
|
|
| 87 |
|
| 88 |
-
|
| 89 |
" --quantization " + quantization + " --conv-template " + conv_template + \
|
| 90 |
-
" -o dist/" + mlc_model_name + "/")
|
|
|
|
|
|
|
| 91 |
|
| 92 |
# push to HF
|
| 93 |
user_name = api.whoami()["name"]
|
|
@@ -131,11 +143,20 @@ def button_click(hf_model_id, conv_template, quantization, oauth_token: gr.OAuth
|
|
| 131 |
|
| 132 |
os.system("rm -rf dist/")
|
| 133 |
|
| 134 |
-
return "Successful"
|
| 135 |
|
| 136 |
with gr.Blocks() as demo:
|
| 137 |
gr.LoginButton()
|
| 138 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 139 |
conv = gr.Dropdown(CONV_TEMPLATES, label="Conversation Template")
|
| 140 |
quant = gr.Dropdown(QUANTIZATIONS, label="Quantization Method")
|
| 141 |
btn = gr.Button("Convert to MLC")
|
|
|
|
| 8 |
from huggingface_hub import HfApi
|
| 9 |
from huggingface_hub import ModelCard
|
| 10 |
|
| 11 |
+
from gradio_huggingfacehub_search import HuggingfaceHubSearch
|
| 12 |
+
|
| 13 |
from textwrap import dedent
|
| 14 |
|
| 15 |
HF_PATH = "https://huggingface.co/"
|
|
|
|
| 71 |
"q4f16_awq"]
|
| 72 |
|
| 73 |
def button_click(hf_model_id, conv_template, quantization, oauth_token: gr.OAuthToken | None):
|
| 74 |
+
if oauth_token.token == None:
|
| 75 |
+
return "Log in to Huggingface to use this"
|
| 76 |
+
elif not hf_model_id:
|
| 77 |
+
return "Enter a Huggingface model ID"
|
| 78 |
+
elif not conv_template:
|
| 79 |
+
return "Select a conversation template"
|
| 80 |
+
elif not quantization:
|
| 81 |
+
return "Select a quantization method"
|
| 82 |
|
| 83 |
api = HfApi(token=oauth_token.token)
|
| 84 |
model_dir_name = hf_model_id.split("/")[1]
|
|
|
|
| 89 |
|
| 90 |
api.snapshot_download(repo_id=hf_model_id, local_dir=f"./dist/models/{model_dir_name}")
|
| 91 |
|
| 92 |
+
convert_weight_result = subprocess.run(["mlc_llm convert_weight ./dist/models/" + model_dir_name + "/" + \
|
| 93 |
" --quantization " + quantization + \
|
| 94 |
+
" -o dist/" + mlc_model_name], shell=True, capture_output=True, text=True)
|
| 95 |
+
if convert_weight_result.stderr:
|
| 96 |
+
return convert_weight_result.stderr
|
| 97 |
|
| 98 |
+
gen_config_result = subprocess.run(["mlc_llm gen_config ./dist/models/" + model_dir_name + "/" + \
|
| 99 |
" --quantization " + quantization + " --conv-template " + conv_template + \
|
| 100 |
+
" -o dist/" + mlc_model_name + "/"], shell=True, capture_output=True, text=True)
|
| 101 |
+
if gen_config_result.stderr:
|
| 102 |
+
return gen_config_result.stderr
|
| 103 |
|
| 104 |
# push to HF
|
| 105 |
user_name = api.whoami()["name"]
|
|
|
|
| 143 |
|
| 144 |
os.system("rm -rf dist/")
|
| 145 |
|
| 146 |
+
return "Successful, please find your compiled LLM model on your personal account"
|
| 147 |
|
| 148 |
with gr.Blocks() as demo:
|
| 149 |
gr.LoginButton()
|
| 150 |
+
gr.Markdown(
|
| 151 |
+
"""
|
| 152 |
+
# Compile your LLM model with MLC-LLM and run it locally!
|
| 153 |
+
### This space takes in Huggingface model ID, and converts it for you using your selected conversation template and quantization method!
|
| 154 |
+
""")
|
| 155 |
+
model_id = HuggingfaceHubSearch(
|
| 156 |
+
label="HF Model ID",
|
| 157 |
+
placeholder="Search for your model on Huggingface",
|
| 158 |
+
search_type="model",
|
| 159 |
+
)
|
| 160 |
conv = gr.Dropdown(CONV_TEMPLATES, label="Conversation Template")
|
| 161 |
quant = gr.Dropdown(QUANTIZATIONS, label="Quantization Method")
|
| 162 |
btn = gr.Button("Convert to MLC")
|