gardel/lightrag
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

移除kwargs中的hashing_kv参数取为变量

Browse Source
This commit is contained in:
magicyuan876
2024-12-06 15:23:18 +08:00
parent 8d9fc01b4c
commit efdd4b8b8e
1 changed files with 6 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
lightrag/llm.py
View File

@@ -73,11 +73,12 @@ async def openai_complete_if_cache(
messages.extend(history_messages) messages.extend(history_messages)
messages.append({"role": "user", "content": prompt}) messages.append({"role": "user", "content": prompt})
hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
# Handle cache # Handle cache
mode = kwargs.pop("mode", "default") mode = kwargs.pop("mode", "default")
args_hash = compute_args_hash(model, messages) args_hash = compute_args_hash(model, messages)
cached_response, quantized, min_val, max_val = await handle_cache( cached_response, quantized, min_val, max_val = await handle_cache(
kwargs.get("hashing_kv"), args_hash, prompt, mode hashing_kv, args_hash, prompt, mode
) )
if cached_response is not None: if cached_response is not None:
return cached_response return cached_response
@@ -219,12 +220,12 @@ async def bedrock_complete_if_cache(
# Add user prompt # Add user prompt
messages.append({"role": "user", "content": [{"text": prompt}]}) messages.append({"role": "user", "content": [{"text": prompt}]})
hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
# Handle cache # Handle cache
mode = kwargs.pop("mode", "default") mode = kwargs.pop("mode", "default")
args_hash = compute_args_hash(model, messages) args_hash = compute_args_hash(model, messages)
cached_response, quantized, min_val, max_val = await handle_cache( cached_response, quantized, min_val, max_val = await handle_cache(
kwargs.get("hashing_kv"), args_hash, prompt, mode hashing_kv, args_hash, prompt, mode
) )
if cached_response is not None: if cached_response is not None:
return cached_response return cached_response
@@ -250,12 +251,12 @@ async def bedrock_complete_if_cache(
args["inferenceConfig"][inference_params_map.get(param, param)] = ( args["inferenceConfig"][inference_params_map.get(param, param)] = (
kwargs.pop(param) kwargs.pop(param)
) )
hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
# Handle cache # Handle cache
mode = kwargs.pop("mode", "default") mode = kwargs.pop("mode", "default")
args_hash = compute_args_hash(model, messages) args_hash = compute_args_hash(model, messages)
cached_response, quantized, min_val, max_val = await handle_cache( cached_response, quantized, min_val, max_val = await handle_cache(
kwargs.get("hashing_kv"), args_hash, prompt, mode hashing_kv, args_hash, prompt, mode
) )
if cached_response is not None: if cached_response is not None:
return cached_response return cached_response