theroyallab/tabbyAPI
1
0
Fork 0
mirror of https://github.com/theroyallab/tabbyAPI.git synced 2026-03-15 00:07:28 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement lora support (#24)

Browse Source 
* Model: Implement basic lora support

* Add ability to load loras from config on launch
* Supports loading multiple loras and lora scaling
* Add function to unload loras

* Colab: Update for basic lora support

* Model: Test vram alloc after lora load, add docs

* Git: Add loras folder to .gitignore

* API: Add basic lora-related endpoints

* Add /loras/ endpoint for querying available loras
* Add /model/lora endpoint for querying currently loaded loras
* Add /model/lora/load endpoint for loading loras
* Add /model/lora/unload endpoint for unloading loras
* Move lora config-checking logic to main.py for better compat with API endpoints

* Revert bad CRLF line ending changes

* API: Add basic lora-related endpoints (fixed)

* Add /loras/ endpoint for querying available loras
* Add /model/lora endpoint for querying currently loaded loras
* Add /model/lora/load endpoint for loading loras
* Add /model/lora/unload endpoint for unloading loras
* Move lora config-checking logic to main.py for better compat with API endpoints

* Model: Unload loras first when unloading model

* API + Models: Cleanup lora endpoints and functions

Condenses down endpoint and model load code. Also makes the routes
behave the same way as model routes to help not confuse the end user.

Signed-off-by: kingbri <bdashore3@proton.me>

* Loras: Optimize load endpoint

Return successes and failures along with consolidating the request
to the rewritten load_loras function.

Signed-off-by: kingbri <bdashore3@proton.me>

---------

Co-authored-by: kingbri <bdashore3@proton.me>
Co-authored-by: DocShotgun <126566557+DocShotgun@users.noreply.github.com>
This commit is contained in:
DocShotgun
2023-12-08 20:36:40 -08:00
committed by kingbri
parent 161c9d2c19
commit 7380a3b79a
8 changed files with 197 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
TabbyAPI_Colab_Example.ipynb
View File

@@ -36,10 +36,17 @@
"# @markdown Select model:\n",
"repo_id = \"royallab/Noromaid-13b-v0.1.1-exl2\" # @param {type:\"string\"}\n",
"revision = \"4bpw\" # @param {type:\"string\"}\n",
"if revision == \"\": revision = \"main\"\n",
"# @markdown ---\n",
"# @markdown Select draft model (optional, for speculative decoding):\n",
"draft_repo_id = \"\" # @param {type:\"string\"}\n",
"draft_revision = \"\" # @param {type:\"string\"}\n",
"if draft_revision == \"\": draft_revision = \"main\"\n",
"# @markdown ---\n",
"# @markdown Select lora (optional):\n",
"lora_repo_id = \"\" # @param {type:\"string\"}\n",
"lora_revision = \"\" # @param {type:\"string\"}\n",
"if lora_revision == \"\": lora_revision = \"main\"\n",
"# @markdown ---\n",
"\n",
"# Install tabbyAPI\n",
@@ -62,8 +69,15 @@
"%cd /content/tabbyAPI/\n",
"\n",
"from huggingface_hub import snapshot_download\n",
"\n",
"snapshot_download(repo_id=repo_id, revision=revision, local_dir=f\"./models/{repo_id.replace('/', '_')}\")\n",
"if len(draft_repo_id) > 0: snapshot_download(repo_id=draft_repo_id, revision=draft_revision, local_dir=f\"./models/{draft_repo_id.replace('/', '_')}\")"
"model = repo_id.replace('/', '_')\n",
"\n",
"if len(draft_repo_id) > 0: snapshot_download(repo_id=draft_repo_id, revision=draft_revision, local_dir=f\"./models/{draft_repo_id.replace('/', '_')}\")\n",
"draft_model = draft_repo_id.replace('/', '_')\n",
"\n",
"if len(lora_repo_id) > 0: snapshot_download(repo_id=lora_repo_id, revision=lora_revision, local_dir=f\"./loras/{lora_repo_id.replace('/', '_')}\")\n",
"lora = lora_repo_id.replace('/', '_')"
]
},
{
@@ -77,9 +91,6 @@
"# @title # Configure and launch API { display-mode: \"form\" }\n",
"# @markdown ---\n",
"# @markdown Model parameters:\n",
"\n",
"model = repo_id.replace('/', '_')\n",
"draft_model = draft_repo_id.replace('/', '_')\n",
"ContextSize = 4096 # @param {type:\"integer\"}\n",
"RopeScale = 1.0 # @param {type:\"number\"}\n",
"RopeAlpha = 1.0 # @param {type:\"number\"}\n",
@@ -88,6 +99,9 @@
"DraftRopeScale = 1.0 # @param {type:\"number\"}\n",
"DraftRopeAlpha = 1.0 # @param {type:\"number\"}\n",
"# @markdown ---\n",
"# @markdown Lora parameters (optional, for loras):\n",
"LoraScaling = 1.0 # @param {type:\"number\"}\n",
"# @markdown ---\n",
"# @markdown Misc options:\n",
"CacheMode = \"FP16\" # @param [\"FP8\", \"FP16\"] {type:\"string\"}\n",
"UseDummyModels = False # @param {type:\"boolean\"}\n",
@@ -161,6 +175,16 @@
" # Rope parameters for draft models (default: 1.0)\n",
" draft_rope_scale: {DraftRopeScale}\n",
" draft_rope_alpha: {DraftRopeAlpha}\n",
"\n",
" # Options for loras\n",
" lora:\n",
" # Overrides the directory to look for loras (default: loras)\n",
" lora_dir: loras\n",
"\n",
" # List of loras to load and associated scaling factors (default: 1.0). Comment out unused entries or add more rows as needed.\n",
" loras:\n",
" - name: {lora}\n",
" scaling: {LoraScaling}\n",
"'''\n",
"with open(\"./config.yml\", \"w\") as file:\n",
" file.write(write)\n",
@@ -188,4 +212,4 @@
},
"nbformat": 4,
"nbformat_minor": 0
}
}