fix: propagate revision parameter to vLLM and PEFT loaders

- Add revision to load_vllm_kwargs in llama.py to fix config/weights mismatch
- Add revision to PEFT AutoConfig calls in loader.py (FastLanguageModel & FastModel)

Addresses reviewer feedback from @chatgpt-codex-connector and @Datta0

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
(cherry picked from commit 14f89e4531)

unsloth/models/llama.py

@@ -2471,6 +2471,7 @@ class FastLlamaModel:
                 disable_log_stats = disable_log_stats,
                 use_bitsandbytes = load_in_4bit,
                 unsloth_vllm_standby = unsloth_vllm_standby,
+                revision = revision,
                 fp8_mode = fp8_mode,
             )
             for allowed_arg in allowed_args:

unsloth/models/loader.py

@@ -553,6 +553,7 @@ class FastLanguageModel(FastLlamaModel):
             model_config = AutoConfig.from_pretrained(
                 model_name,
                 token = token,
+                revision = revision,
                 trust_remote_code = trust_remote_code,
             )
 
@@ -1304,6 +1305,7 @@ class FastModel(FastBaseModel):
             model_config = AutoConfig.from_pretrained(
                 model_name,
                 token = token,
+                revision = revision,
                 trust_remote_code = trust_remote_code,
             )